http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/docs/tutorials/spark-samsara/classify-a-doc-from-the-shell.md ---------------------------------------------------------------------- diff --git a/website/docs/tutorials/spark-samsara/classify-a-doc-from-the-shell.md b/website/docs/tutorials/spark-samsara/classify-a-doc-from-the-shell.md new file mode 100644 index 0000000..0a237d1 --- /dev/null +++ b/website/docs/tutorials/spark-samsara/classify-a-doc-from-the-shell.md @@ -0,0 +1,258 @@ +--- +layout: page +title: Text Classification Example +theme: + name: mahout2 +--- + +# Building a text classifier in Mahout's Spark Shell + +This tutorial will take you through the steps used to train a Multinomial Naive Bayes model and create a text classifier based on that model using the ```mahout spark-shell```. + +## Prerequisites +This tutorial assumes that you have your Spark environment variables set for the ```mahout spark-shell``` see: [Playing with Mahout's Shell](http://mahout.apache.org/users/sparkbindings/play-with-shell.html). As well we assume that Mahout is running in cluster mode (i.e. with the ```MAHOUT_LOCAL``` environment variable **unset**) as we'll be reading and writing to HDFS. + +## Downloading and Vectorizing the Wikipedia dataset +*As of Mahout v. 0.10.0, we are still reliant on the MapReduce versions of ```mahout seqwiki``` and ```mahout seq2sparse``` to extract and vectorize our text. A* [*Spark implementation of seq2sparse*](https://issues.apache.org/jira/browse/MAHOUT-1663) *is in the works for Mahout v. 0.11.* However, to download the Wikipedia dataset, extract the bodies of the documentation, label each document and vectorize the text into TF-IDF vectors, we can simpmly run the [wikipedia-classifier.sh](https://github.com/apache/mahout/blob/master/examples/bin/classify-wikipedia.sh) example. + + Please select a number to choose the corresponding task to run + 1. CBayes (may require increased heap space on yarn) + 2. BinaryCBayes + 3. clean -- cleans up the work area in /tmp/mahout-work-wiki + Enter your choice : + +Enter (2). This will download a large recent XML dump of the Wikipedia database, into a ```/tmp/mahout-work-wiki``` directory, unzip it and place it into HDFS. It will run a [MapReduce job to parse the wikipedia set](http://mahout.apache.org/users/classification/wikipedia-classifier-example.html), extracting and labeling only pages with category tags for [United States] and [United Kingdom] (~11600 documents). It will then run ```mahout seq2sparse``` to convert the documents into TF-IDF vectors. The script will also a build and test a [Naive Bayes model using MapReduce](http://mahout.apache.org/users/classification/bayesian.html). When it is completed, you should see a confusion matrix on your screen. For this tutorial, we will ignore the MapReduce model, and build a new model using Spark based on the vectorized text output by ```seq2sparse```. + +## Getting Started + +Launch the ```mahout spark-shell```. There is an example script: ```spark-document-classifier.mscala``` (.mscala denotes a Mahout-Scala script which can be run similarly to an R script). We will be walking through this script for this tutorial but if you wanted to simply run the script, you could just issue the command: + + mahout> :load /path/to/mahout/examples/bin/spark-document-classifier.mscala + +For now, lets take the script apart piece by piece. You can cut and paste the following code blocks into the ```mahout spark-shell```. + +## Imports + +Our Mahout Naive Bayes imports: + + import org.apache.mahout.classifier.naivebayes._ + import org.apache.mahout.classifier.stats._ + import org.apache.mahout.nlp.tfidf._ + +Hadoop imports needed to read our dictionary: + + import org.apache.hadoop.io.Text + import org.apache.hadoop.io.IntWritable + import org.apache.hadoop.io.LongWritable + +## Read in our full set from HDFS as vectorized by seq2sparse in classify-wikipedia.sh + + val pathToData = "/tmp/mahout-work-wiki/" + val fullData = drmDfsRead(pathToData + "wikipediaVecs/tfidf-vectors") + +## Extract the category of each observation and aggregate those observations by category + + val (labelIndex, aggregatedObservations) = SparkNaiveBayes.extractLabelsAndAggregateObservations( + fullData) + +## Build a Muitinomial Naive Bayes model and self test on the training set + + val model = SparkNaiveBayes.train(aggregatedObservations, labelIndex, false) + val resAnalyzer = SparkNaiveBayes.test(model, fullData, false) + println(resAnalyzer) + +printing the ```ResultAnalyzer``` will display the confusion matrix. + +## Read in the dictionary and document frequency count from HDFS + + val dictionary = sdc.sequenceFile(pathToData + "wikipediaVecs/dictionary.file-0", + classOf[Text], + classOf[IntWritable]) + val documentFrequencyCount = sdc.sequenceFile(pathToData + "wikipediaVecs/df-count", + classOf[IntWritable], + classOf[LongWritable]) + + // setup the dictionary and document frequency count as maps + val dictionaryRDD = dictionary.map { + case (wKey, wVal) => wKey.asInstanceOf[Text] + .toString() -> wVal.get() + } + + val documentFrequencyCountRDD = documentFrequencyCount.map { + case (wKey, wVal) => wKey.asInstanceOf[IntWritable] + .get() -> wVal.get() + } + + val dictionaryMap = dictionaryRDD.collect.map(x => x._1.toString -> x._2.toInt).toMap + val dfCountMap = documentFrequencyCountRDD.collect.map(x => x._1.toInt -> x._2.toLong).toMap + +## Define a function to tokenize and vectorize new text using our current dictionary + +For this simple example, our function ```vectorizeDocument(...)``` will tokenize a new document into unigrams using native Java String methods and vectorize using our dictionary and document frequencies. You could also use a [Lucene](https://lucene.apache.org/core/) analyzer for bigrams, trigrams, etc., and integrate Apache [Tika](https://tika.apache.org/) to extract text from different document types (PDF, PPT, XLS, etc.). Here, however we will keep it simple, stripping and tokenizing our text using regexs and native String methods. + + def vectorizeDocument(document: String, + dictionaryMap: Map[String,Int], + dfMap: Map[Int,Long]): Vector = { + val wordCounts = document.replaceAll("[^\\p{L}\\p{Nd}]+", " ") + .toLowerCase + .split(" ") + .groupBy(identity) + .mapValues(_.length) + val vec = new RandomAccessSparseVector(dictionaryMap.size) + val totalDFSize = dfMap(-1) + val docSize = wordCounts.size + for (word <- wordCounts) { + val term = word._1 + if (dictionaryMap.contains(term)) { + val tfidf: TermWeight = new TFIDF() + val termFreq = word._2 + val dictIndex = dictionaryMap(term) + val docFreq = dfCountMap(dictIndex) + val currentTfIdf = tfidf.calculate(termFreq, + docFreq.toInt, + docSize, + totalDFSize.toInt) + vec.setQuick(dictIndex, currentTfIdf) + } + } + vec + } + +## Setup our classifier + + val labelMap = model.labelIndex + val numLabels = model.numLabels + val reverseLabelMap = labelMap.map(x => x._2 -> x._1) + + // instantiate the correct type of classifier + val classifier = model.isComplementary match { + case true => new ComplementaryNBClassifier(model) + case _ => new StandardNBClassifier(model) + } + +## Define an argmax function + +The label with the highest score wins the classification for a given document. + + def argmax(v: Vector): (Int, Double) = { + var bestIdx: Int = Integer.MIN_VALUE + var bestScore: Double = Integer.MIN_VALUE.asInstanceOf[Int].toDouble + for(i <- 0 until v.size) { + if(v(i) > bestScore){ + bestScore = v(i) + bestIdx = i + } + } + (bestIdx, bestScore) + } + +## Define our TF(-IDF) vector classifier + + def classifyDocument(clvec: Vector) : String = { + val cvec = classifier.classifyFull(clvec) + val (bestIdx, bestScore) = argmax(cvec) + reverseLabelMap(bestIdx) + } + +## Two sample news articles: United States Football and United Kingdom Football + + // A random United States football article + // http://www.reuters.com/article/2015/01/28/us-nfl-superbowl-security-idUSKBN0L12JR20150128 + val UStextToClassify = new String("(Reuters) - Super Bowl security officials acknowledge" + + " the NFL championship game represents a high profile target on a world stage but are" + + " unaware of any specific credible threats against Sunday's showcase. In advance of" + + " one of the world's biggest single day sporting events, Homeland Security Secretary" + + " Jeh Johnson was in Glendale on Wednesday to review security preparations and tour" + + " University of Phoenix Stadium where the Seattle Seahawks and New England Patriots" + + " will battle. Deadly shootings in Paris and arrest of suspects in Belgium, Greece and" + + " Germany heightened fears of more attacks around the world and social media accounts" + + " linked to Middle East militant groups have carried a number of threats to attack" + + " high-profile U.S. events. There is no specific credible threat, said Johnson, who" + + " has appointed a federal coordination team to work with local, state and federal" + + " agencies to ensure safety of fans, players and other workers associated with the" + + " Super Bowl. I'm confident we will have a safe and secure and successful event." + + " Sunday's game has been given a Special Event Assessment Rating (SEAR) 1 rating, the" + + " same as in previous years, except for the year after the Sept. 11, 2001 attacks, when" + + " a higher level was declared. But security will be tight and visible around Super" + + " Bowl-related events as well as during the game itself. All fans will pass through" + + " metal detectors and pat downs. Over 4,000 private security personnel will be deployed" + + " and the almost 3,000 member Phoenix police force will be on Super Bowl duty. Nuclear" + + " device sniffing teams will be deployed and a network of Bio-Watch detectors will be" + + " set up to provide a warning in the event of a biological attack. The Department of" + + " Homeland Security (DHS) said in a press release it had held special cyber-security" + + " and anti-sniper training sessions. A U.S. official said the Transportation Security" + + " Administration, which is responsible for screening airline passengers, will add" + + " screeners and checkpoint lanes at airports. Federal air marshals, behavior detection" + + " officers and dog teams will help to secure transportation systems in the area. We" + + " will be ramping it (security) up on Sunday, there is no doubt about that, said Federal"+ + " Coordinator Matthew Allen, the DHS point of contact for planning and support. I have" + + " every confidence the public safety agencies that represented in the planning process" + + " are going to have their best and brightest out there this weekend and we will have" + + " a very safe Super Bowl.") + + // A random United Kingdom football article + // http://www.reuters.com/article/2015/01/26/manchester-united-swissquote-idUSL6N0V52RZ20150126 + val UKtextToClassify = new String("(Reuters) - Manchester United have signed a sponsorship" + + " deal with online financial trading company Swissquote, expanding the commercial" + + " partnerships that have helped to make the English club one of the richest teams in" + + " world soccer. United did not give a value for the deal, the club's first in the sector," + + " but said on Monday it was a multi-year agreement. The Premier League club, 20 times" + + " English champions, claim to have 659 million followers around the globe, making the" + + " United name attractive to major brands like Chevrolet cars and sportswear group Adidas." + + " Swissquote said the global deal would allow it to use United's popularity in Asia to" + + " help it meet its targets for expansion in China. Among benefits from the deal," + + " Swissquote's clients will have a chance to meet United players and get behind the scenes" + + " at the Old Trafford stadium. Swissquote is a Geneva-based online trading company that" + + " allows retail investors to buy and sell foreign exchange, equities, bonds and other asset" + + " classes. Like other retail FX brokers, Swissquote was left nursing losses on the Swiss" + + " franc after Switzerland's central bank stunned markets this month by abandoning its cap" + + " on the currency. The fallout from the abrupt move put rival and West Ham United shirt" + + " sponsor Alpari UK into administration. Swissquote itself was forced to book a 25 million" + + " Swiss francs ($28 million) provision for its clients who were left out of pocket" + + " following the franc's surge. United's ability to grow revenues off the pitch has made" + + " them the second richest club in the world behind Spain's Real Madrid, despite a" + + " downturn in their playing fortunes. United Managing Director Richard Arnold said" + + " there was still lots of scope for United to develop sponsorships in other areas of" + + " business. The last quoted statistics that we had showed that of the top 25 sponsorship" + + " categories, we were only active in 15 of those, Arnold told Reuters. I think there is a" + + " huge potential still for the club, and the other thing we have seen is there is very" + + " significant growth even within categories. United have endured a tricky transition" + + " following the retirement of manager Alex Ferguson in 2013, finishing seventh in the" + + " Premier League last season and missing out on a place in the lucrative Champions League." + + " ($1 = 0.8910 Swiss francs) (Writing by Neil Maidment, additional reporting by Jemima" + + " Kelly; editing by Keith Weir)") + +## Vectorize and classify our documents + + val usVec = vectorizeDocument(UStextToClassify, dictionaryMap, dfCountMap) + val ukVec = vectorizeDocument(UKtextToClassify, dictionaryMap, dfCountMap) + + println("Classifying the news article about superbowl security (united states)") + classifyDocument(usVec) + + println("Classifying the news article about Manchester United (united kingdom)") + classifyDocument(ukVec) + +## Tie everything together in a new method to classify text + + def classifyText(txt: String): String = { + val v = vectorizeDocument(txt, dictionaryMap, dfCountMap) + classifyDocument(v) + } + +## Now we can simply call our classifyText(...) method on any String + + classifyText("Hello world from Queens") + classifyText("Hello world from London") + +## Model persistance + +You can save the model to HDFS: + + model.dfsWrite("/path/to/model") + +And retrieve it with: + + val model = NBModel.dfsRead("/path/to/model") + +The trained model can now be embedded in an external application. \ No newline at end of file
http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/docs/tutorials/spark-samsara/how-to-build-an-app.md ---------------------------------------------------------------------- diff --git a/website/docs/tutorials/spark-samsara/how-to-build-an-app.md b/website/docs/tutorials/spark-samsara/how-to-build-an-app.md new file mode 100644 index 0000000..0ad232e --- /dev/null +++ b/website/docs/tutorials/spark-samsara/how-to-build-an-app.md @@ -0,0 +1,256 @@ +--- +layout: page +title: Mahout Samsara In Core +theme: + name: mahout2 +--- +# How to create and App using Mahout + +This is an example of how to create a simple app using Mahout as a Library. The source is available on Github in the [3-input-cooc project](https://github.com/pferrel/3-input-cooc) with more explanation about what it does (has to do with collaborative filtering). For this tutorial we'll concentrate on the app rather than the data science. + +The app reads in three user-item interactions types and creats indicators for them using cooccurrence and cross-cooccurrence. The indicators will be written to text files in a format ready for search engine indexing in search engine based recommender. + +## Setup +In order to build and run the CooccurrenceDriver you need to install the following: + +* Install the Java 7 JDK from Oracle. Mac users look here: [Java SE Development Kit 7u72](http://www.oracle.com/technetwork/java/javase/downloads/jdk7-downloads-1880260.html). +* Install sbt (simple build tool) 0.13.x for [Mac](http://www.scala-sbt.org/release/tutorial/Installing-sbt-on-Mac.html), [Linux](http://www.scala-sbt.org/release/tutorial/Installing-sbt-on-Linux.html) or [manual instalation](http://www.scala-sbt.org/release/tutorial/Manual-Installation.html). +* Install [Spark 1.1.1](https://spark.apache.org/docs/1.1.1/spark-standalone.html). Don't forget to setup SPARK_HOME +* Install [Mahout 0.10.0](http://mahout.apache.org/general/downloads.html). Don't forget to setup MAHOUT_HOME and MAHOUT_LOCAL + +Why install if you are only using them as a library? Certain binaries and scripts are required by the libraries to get information about the environment like discovering where jars are located. + +Spark requires a set of jars on the classpath for the client side part of an app and another set of jars must be passed to the Spark Context for running distributed code. The example should discover all the neccessary classes automatically. + +## Application +Using Mahout as a library in an application will require a little Scala code. Scala has an App trait so we'll create an object, which inherits from ```App``` + + + object CooccurrenceDriver extends App { + } + + +This will look a little different than Java since ```App``` does delayed initialization, which causes the body to be executed when the App is launched, just as in Java you would create a main method. + +Before we can execute something on Spark we'll need to create a context. We could use raw Spark calls here but default values are setup for a Mahout context by using the Mahout helper function. + + implicit val mc = mahoutSparkContext(masterUrl = "local", + appName = "CooccurrenceDriver") + +We need to read in three files containing different interaction types. The files will each be read into a Mahout IndexedDataset. This allows us to preserve application-specific user and item IDs throughout the calculations. + +For example, here is data/purchase.csv: + + u1,iphone + u1,ipad + u2,nexus + u2,galaxy + u3,surface + u4,iphone + u4,galaxy + +Mahout has a helper function that reads the text delimited files SparkEngine.indexedDatasetDFSReadElements. The function reads single element tuples (user-id,item-id) in a distributed way to create the IndexedDataset. Distributed Row Matrices (DRM) and Vectors are important data types supplied by Mahout and IndexedDataset is like a very lightweight Dataframe in R, it wraps a DRM with HashBiMaps for row and column IDs. + +One important thing to note about this example is that we read in all datasets before we adjust the number of rows in them to match the total number of users in the data. This is so the math works out [(A'A, A'B, A'C)](http://mahout.apache.org/users/algorithms/intro-cooccurrence-spark.html) even if some users took one action but not another there must be the same number of rows in all matrices. + + /** + * Read files of element tuples and create IndexedDatasets one per action. These + * share a userID BiMap but have their own itemID BiMaps + */ + def readActions(actionInput: Array[(String, String)]): Array[(String, IndexedDataset)] = { + var actions = Array[(String, IndexedDataset)]() + + val userDictionary: BiMap[String, Int] = HashBiMap.create() + + // The first action named in the sequence is the "primary" action and + // begins to fill up the user dictionary + for ( actionDescription <- actionInput ) {// grab the path to actions + val action: IndexedDataset = SparkEngine.indexedDatasetDFSReadElements( + actionDescription._2, + schema = DefaultIndexedDatasetElementReadSchema, + existingRowIDs = userDictionary) + userDictionary.putAll(action.rowIDs) + // put the name in the tuple with the indexedDataset + actions = actions :+ (actionDescription._1, action) + } + + // After all actions are read in the userDictonary will contain every user seen, + // even if they may not have taken all actions . Now we adjust the row rank of + // all IndxedDataset's to have this number of rows + // Note: this is very important or the cooccurrence calc may fail + val numUsers = userDictionary.size() // one more than the cardinality + + val resizedNameActionPairs = actions.map { a => + //resize the matrix by, in effect by adding empty rows + val resizedMatrix = a._2.create(a._2.matrix, userDictionary, a._2.columnIDs).newRowCardinality(numUsers) + (a._1, resizedMatrix) // return the Tuple of (name, IndexedDataset) + } + resizedNameActionPairs // return the array of Tuples + } + + +Now that we have the data read in we can perform the cooccurrence calculation. + + // actions.map creates an array of just the IndeedDatasets + val indicatorMatrices = SimilarityAnalysis.cooccurrencesIDSs( + actions.map(a => a._2)) + +All we need to do now is write the indicators. + + // zip a pair of arrays into an array of pairs, reattaching the action names + val indicatorDescriptions = actions.map(a => a._1).zip(indicatorMatrices) + writeIndicators(indicatorDescriptions) + + +The ```writeIndicators``` method uses the default write function ```dfsWrite```. + + /** + * Write indicatorMatrices to the output dir in the default format + * for indexing by a search engine. + */ + def writeIndicators( indicators: Array[(String, IndexedDataset)]) = { + for (indicator <- indicators ) { + // create a name based on the type of indicator + val indicatorDir = OutputPath + indicator._1 + indicator._2.dfsWrite( + indicatorDir, + // Schema tells the writer to omit LLR strengths + // and format for search engine indexing + IndexedDatasetWriteBooleanSchema) + } + } + + +See the Github project for the full source. Now we create a build.sbt to build the example. + + name := "cooccurrence-driver" + + organization := "com.finderbots" + + version := "0.1" + + scalaVersion := "2.10.4" + + val sparkVersion = "1.1.1" + + libraryDependencies ++= Seq( + "log4j" % "log4j" % "1.2.17", + // Mahout's Spark code + "commons-io" % "commons-io" % "2.4", + "org.apache.mahout" % "mahout-math-scala_2.10" % "0.10.0", + "org.apache.mahout" % "mahout-spark_2.10" % "0.10.0", + "org.apache.mahout" % "mahout-math" % "0.10.0", + "org.apache.mahout" % "mahout-hdfs" % "0.10.0", + // Google collections, AKA Guava + "com.google.guava" % "guava" % "16.0") + + resolvers += "typesafe repo" at " http://repo.typesafe.com/typesafe/releases/"; + + resolvers += Resolver.mavenLocal + + packSettings + + packMain := Map( + "cooc" -> "CooccurrenceDriver") + + +## Build +Building the examples from project's root folder: + + $ sbt pack + +This will automatically set up some launcher scripts for the driver. To run execute + + $ target/pack/bin/cooc + +The driver will execute in Spark standalone mode and put the data in /path/to/3-input-cooc/data/indicators/*indicator-type* + +## Using a Debugger +To build and run this example in a debugger like IntelliJ IDEA. Install from the IntelliJ site and add the Scala plugin. + +Open IDEA and go to the menu File->New->Project from existing sources->SBT->/path/to/3-input-cooc. This will create an IDEA project from ```build.sbt``` in the root directory. + +At this point you may create a "Debug Configuration" to run. In the menu choose Run->Edit Configurations. Under "Default" choose "Application". In the dialog hit the elipsis button "..." to the right of "Environment Variables" and fill in your versions of JAVA_HOME, SPARK_HOME, and MAHOUT_HOME. In configuration editor under "Use classpath from" choose root-3-input-cooc module. + + + +Now choose "Application" in the left pane and hit the plus sign "+". give the config a name and hit the elipsis button to the right of the "Main class" field as shown. + + + + +After setting breakpoints you are now ready to debug the configuration. Go to the Run->Debug... menu and pick your configuration. This will execute using a local standalone instance of Spark. + +##The Mahout Shell + +For small script-like apps you may wish to use the Mahout shell. It is a Scala REPL type interactive shell built on the Spark shell with Mahout-Samsara extensions. + +To make the CooccurrenceDriver.scala into a script make the following changes: + +* You won't need the context, since it is created when the shell is launched, comment that line out. +* Replace the logger.info lines with println +* Remove the package info since it's not needed, this will produce the file in ```path/to/3-input-cooc/bin/CooccurrenceDriver.mscala```. + +Note the extension ```.mscala``` to indicate we are using Mahout's scala extensions for math, otherwise known as [Mahout-Samsara](http://mahout.apache.org/users/environment/out-of-core-reference.html) + +To run the code make sure the output does not exist already + + $ rm -r /path/to/3-input-cooc/data/indicators + +Launch the Mahout + Spark shell: + + $ mahout spark-shell + +You'll see the Mahout splash: + + MAHOUT_LOCAL is set, so we don't add HADOOP_CONF_DIR to classpath. + + _ _ + _ __ ___ __ _| |__ ___ _ _| |_ + | '_ ` _ \ / _` | '_ \ / _ \| | | | __| + | | | | | | (_| | | | | (_) | |_| | |_ + |_| |_| |_|\__,_|_| |_|\___/ \__,_|\__| version 0.10.0 + + + Using Scala version 2.10.4 (Java HotSpot(TM) 64-Bit Server VM, Java 1.7.0_72) + Type in expressions to have them evaluated. + Type :help for more information. + 15/04/26 09:30:48 WARN NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable + Created spark context.. + Mahout distributed context is available as "implicit val sdc". + mahout> + +To load the driver type: + + mahout> :load /path/to/3-input-cooc/bin/CooccurrenceDriver.mscala + Loading ./bin/CooccurrenceDriver.mscala... + import com.google.common.collect.{HashBiMap, BiMap} + import org.apache.log4j.Logger + import org.apache.mahout.math.cf.SimilarityAnalysis + import org.apache.mahout.math.indexeddataset._ + import org.apache.mahout.sparkbindings._ + import scala.collection.immutable.HashMap + defined module CooccurrenceDriver + mahout> + +To run the driver type: + + mahout> CooccurrenceDriver.main(args = Array("")) + +You'll get some stats printed: + + Total number of users for all actions = 5 + purchase indicator matrix: + Number of rows for matrix = 4 + Number of columns for matrix = 5 + Number of rows after resize = 5 + view indicator matrix: + Number of rows for matrix = 4 + Number of columns for matrix = 5 + Number of rows after resize = 5 + category indicator matrix: + Number of rows for matrix = 5 + Number of columns for matrix = 7 + Number of rows after resize = 5 + +If you look in ```path/to/3-input-cooc/data/indicators``` you should find folders containing the indicator matrices. http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/docs/tutorials/spark-samsara/play-with-shell.md ---------------------------------------------------------------------- diff --git a/website/docs/tutorials/spark-samsara/play-with-shell.md b/website/docs/tutorials/spark-samsara/play-with-shell.md new file mode 100644 index 0000000..6b5e4a0 --- /dev/null +++ b/website/docs/tutorials/spark-samsara/play-with-shell.md @@ -0,0 +1,199 @@ +--- +layout: page +title: Mahout Samsara In Core +theme: + name: mahout2 +--- +# Playing with Mahout's Spark Shell + +This tutorial will show you how to play with Mahout's scala DSL for linear algebra and its Spark shell. **Please keep in mind that this code is still in a very early experimental stage**. + +_(Edited for 0.10.2)_ + +## Intro + +We'll use an excerpt of a publicly available [dataset about cereals](http://lib.stat.cmu.edu/DASL/Datafiles/Cereals.html). The dataset tells the protein, fat, carbohydrate and sugars (in milligrams) contained in a set of cereals, as well as a customer rating for the cereals. Our aim for this example is to fit a linear model which infers the customer rating from the ingredients. + + +Name | protein | fat | carbo | sugars | rating +:-----------------------|:--------|:----|:------|:-------|:--------- +Apple Cinnamon Cheerios | 2 | 2 | 10.5 | 10 | 29.509541 +Cap'n'Crunch | 1 | 2 | 12 | 12 | 18.042851 +Cocoa Puffs | 1 | 1 | 12 | 13 | 22.736446 +Froot Loops | 2 | 1 | 11 | 13 | 32.207582 +Honey Graham Ohs | 1 | 2 | 12 | 11 | 21.871292 +Wheaties Honey Gold | 2 | 1 | 16 | 8 | 36.187559 +Cheerios | 6 | 2 | 17 | 1 | 50.764999 +Clusters | 3 | 2 | 13 | 7 | 40.400208 +Great Grains Pecan | 3 | 3 | 13 | 4 | 45.811716 + + +## Installing Mahout & Spark on your local machine + +We describe how to do a quick toy setup of Spark & Mahout on your local machine, so that you can run this example and play with the shell. + + 1. Download [Apache Spark 1.6.2](http://d3kbcqa49mib13.cloudfront.net/spark-1.6.2-bin-hadoop2.6.tgz) and unpack the archive file + 1. Change to the directory where you unpacked Spark and type ```sbt/sbt assembly``` to build it + 1. Create a directory for Mahout somewhere on your machine, change to there and checkout the master branch of Apache Mahout from GitHub ```git clone https://github.com/apache/mahout mahout``` + 1. Change to the ```mahout``` directory and build mahout using ```mvn -DskipTests clean install``` + +## Starting Mahout's Spark shell + + 1. Goto the directory where you unpacked Spark and type ```sbin/start-all.sh``` to locally start Spark + 1. Open a browser, point it to [http://localhost:8080/](http://localhost:8080/) to check whether Spark successfully started. Copy the url of the spark master at the top of the page (it starts with **spark://**) + 1. Define the following environment variables: <pre class="codehilite">export MAHOUT_HOME=[directory into which you checked out Mahout] +export SPARK_HOME=[directory where you unpacked Spark] +export MASTER=[url of the Spark master] +</pre> + 1. Finally, change to the directory where you unpacked Mahout and type ```bin/mahout spark-shell```, +you should see the shell starting and get the prompt ```mahout> ```. Check +[FAQ](http://mahout.apache.org/users/sparkbindings/faq.html) for further troubleshooting. + +## Implementation + +We'll use the shell to interactively play with the data and incrementally implement a simple [linear regression](https://en.wikipedia.org/wiki/Linear_regression) algorithm. Let's first load the dataset. Usually, we wouldn't need Mahout unless we processed a large dataset stored in a distributed filesystem. But for the sake of this example, we'll use our tiny toy dataset and "pretend" it was too big to fit onto a single machine. + +*Note: You can incrementally follow the example by copy-and-pasting the code into your running Mahout shell.* + +Mahout's linear algebra DSL has an abstraction called *DistributedRowMatrix (DRM)* which models a matrix that is partitioned by rows and stored in the memory of a cluster of machines. We use ```dense()``` to create a dense in-memory matrix from our toy dataset and use ```drmParallelize``` to load it into the cluster, "mimicking" a large, partitioned dataset. + +<div class="codehilite"><pre> +val drmData = drmParallelize(dense( + (2, 2, 10.5, 10, 29.509541), // Apple Cinnamon Cheerios + (1, 2, 12, 12, 18.042851), // Cap'n'Crunch + (1, 1, 12, 13, 22.736446), // Cocoa Puffs + (2, 1, 11, 13, 32.207582), // Froot Loops + (1, 2, 12, 11, 21.871292), // Honey Graham Ohs + (2, 1, 16, 8, 36.187559), // Wheaties Honey Gold + (6, 2, 17, 1, 50.764999), // Cheerios + (3, 2, 13, 7, 40.400208), // Clusters + (3, 3, 13, 4, 45.811716)), // Great Grains Pecan + numPartitions = 2); +</pre></div> + +Have a look at this matrix. The first four columns represent the ingredients +(our features) and the last column (the rating) is the target variable for +our regression. [Linear regression](https://en.wikipedia.org/wiki/Linear_regression) +assumes that the **target variable** `\(\mathbf{y}\)` is generated by the +linear combination of **the feature matrix** `\(\mathbf{X}\)` with the +**parameter vector** `\(\boldsymbol{\beta}\)` plus the + **noise** `\(\boldsymbol{\varepsilon}\)`, summarized in the formula +`\(\mathbf{y}=\mathbf{X}\boldsymbol{\beta}+\boldsymbol{\varepsilon}\)`. +Our goal is to find an estimate of the parameter vector +`\(\boldsymbol{\beta}\)` that explains the data very well. + +As a first step, we extract `\(\mathbf{X}\)` and `\(\mathbf{y}\)` from our data matrix. We get *X* by slicing: we take all rows (denoted by ```::```) and the first four columns, which have the ingredients in milligrams as content. Note that the result is again a DRM. The shell will not execute this code yet, it saves the history of operations and defers the execution until we really access a result. **Mahout's DSL automatically optimizes and parallelizes all operations on DRMs and runs them on Apache Spark.** + +<div class="codehilite"><pre> +val drmX = drmData(::, 0 until 4) +</pre></div> + +Next, we extract the target variable vector *y*, the fifth column of the data matrix. We assume this one fits into our driver machine, so we fetch it into memory using ```collect```: + +<div class="codehilite"><pre> +val y = drmData.collect(::, 4) +</pre></div> + +Now we are ready to think about a mathematical way to estimate the parameter vector *β*. A simple textbook approach is [ordinary least squares (OLS)](https://en.wikipedia.org/wiki/Ordinary_least_squares), which minimizes the sum of residual squares between the true target variable and the prediction of the target variable. In OLS, there is even a closed form expression for estimating `\(\boldsymbol{\beta}\)` as +`\(\left(\mathbf{X}^{\top}\mathbf{X}\right)^{-1}\mathbf{X}^{\top}\mathbf{y}\)`. + +The first thing which we compute for this is `\(\mathbf{X}^{\top}\mathbf{X}\)`. The code for doing this in Mahout's scala DSL maps directly to the mathematical formula. The operation ```.t()``` transposes a matrix and analogous to R ```%*%``` denotes matrix multiplication. + +<div class="codehilite"><pre> +val drmXtX = drmX.t %*% drmX +</pre></div> + +The same is true for computing `\(\mathbf{X}^{\top}\mathbf{y}\)`. We can simply type the math in scala expressions into the shell. Here, *X* lives in the cluster, while is *y* in the memory of the driver, and the result is a DRM again. +<div class="codehilite"><pre> +val drmXty = drmX.t %*% y +</pre></div> + +We're nearly done. The next step we take is to fetch `\(\mathbf{X}^{\top}\mathbf{X}\)` and +`\(\mathbf{X}^{\top}\mathbf{y}\)` into the memory of our driver machine (we are targeting +features matrices that are tall and skinny , +so we can assume that `\(\mathbf{X}^{\top}\mathbf{X}\)` is small enough +to fit in). Then, we provide them to an in-memory solver (Mahout provides +the an analog to R's ```solve()``` for that) which computes ```beta```, our +OLS estimate of the parameter vector `\(\boldsymbol{\beta}\)`. + +<div class="codehilite"><pre> +val XtX = drmXtX.collect +val Xty = drmXty.collect(::, 0) + +val beta = solve(XtX, Xty) +</pre></div> + +That's it! We have a implemented a distributed linear regression algorithm +on Apache Spark. I hope you agree that we didn't have to worry a lot about +parallelization and distributed systems. The goal of Mahout's linear algebra +DSL is to abstract away the ugliness of programming a distributed system +as much as possible, while still retaining decent performance and +scalability. + +We can now check how well our model fits its training data. +First, we multiply the feature matrix `\(\mathbf{X}\)` by our estimate of +`\(\boldsymbol{\beta}\)`. Then, we look at the difference (via L2-norm) of +the target variable `\(\mathbf{y}\)` to the fitted target variable: + +<div class="codehilite"><pre> +val yFitted = (drmX %*% beta).collect(::, 0) +(y - yFitted).norm(2) +</pre></div> + +We hope that we could show that Mahout's shell allows people to interactively and incrementally write algorithms. We have entered a lot of individual commands, one-by-one, until we got the desired results. We can now refactor a little by wrapping our statements into easy-to-use functions. The definition of functions follows standard scala syntax. + +We put all the commands for ordinary least squares into a function ```ols```. + +<div class="codehilite"><pre> +def ols(drmX: DrmLike[Int], y: Vector) = + solve(drmX.t %*% drmX, drmX.t %*% y)(::, 0) + +</pre></div> + +Note that DSL declares implicit `collect` if coersion rules require an in-core argument. Hence, we can simply +skip explicit `collect`s. + +Next, we define a function ```goodnessOfFit``` that tells how well a model fits the target variable: + +<div class="codehilite"><pre> +def goodnessOfFit(drmX: DrmLike[Int], beta: Vector, y: Vector) = { + val fittedY = (drmX %*% beta).collect(::, 0) + (y - fittedY).norm(2) +} +</pre></div> + +So far we have left out an important aspect of a standard linear regression +model. Usually there is a constant bias term added to the model. Without +that, our model always crosses through the origin and we only learn the +right angle. An easy way to add such a bias term to our model is to add a +column of ones to the feature matrix `\(\mathbf{X}\)`. +The corresponding weight in the parameter vector will then be the bias term. + +Here is how we add a bias column: + +<div class="codehilite"><pre> +val drmXwithBiasColumn = drmX cbind 1 +</pre></div> + +Now we can give the newly created DRM ```drmXwithBiasColumn``` to our model fitting method ```ols``` and see how well the resulting model fits the training data with ```goodnessOfFit```. You should see a large improvement in the result. + +<div class="codehilite"><pre> +val betaWithBiasTerm = ols(drmXwithBiasColumn, y) +goodnessOfFit(drmXwithBiasColumn, betaWithBiasTerm, y) +</pre></div> + +As a further optimization, we can make use of the DSL's caching functionality. We use ```drmXwithBiasColumn``` repeatedly as input to a computation, so it might be beneficial to cache it in memory. This is achieved by calling ```checkpoint()```. In the end, we remove it from the cache with uncache: + +<div class="codehilite"><pre> +val cachedDrmX = drmXwithBiasColumn.checkpoint() + +val betaWithBiasTerm = ols(cachedDrmX, y) +val goodness = goodnessOfFit(cachedDrmX, betaWithBiasTerm, y) + +cachedDrmX.uncache() + +goodness +</pre></div> + + +Liked what you saw? Checkout Mahout's overview for the [Scala and Spark bindings]({{ BASE_PATH }}/distributed/spark-bindings). \ No newline at end of file http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/docs/tutorials/spark-samsara/spark-naive-bayes.md ---------------------------------------------------------------------- diff --git a/website/docs/tutorials/spark-samsara/spark-naive-bayes.md b/website/docs/tutorials/spark-samsara/spark-naive-bayes.md new file mode 100644 index 0000000..8823812 --- /dev/null +++ b/website/docs/tutorials/spark-samsara/spark-naive-bayes.md @@ -0,0 +1,132 @@ +--- +layout: default +title: Spark Naive Bayes +theme: + name: retro-mahout +--- + +# Spark Naive Bayes + + +## Intro + +Mahout currently has two flavors of Naive Bayes. The first is standard Multinomial Naive Bayes. The second is an implementation of Transformed Weight-normalized Complement Naive Bayes as introduced by Rennie et al. [[1]](http://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf). We refer to the former as Bayes and the latter as CBayes. + +Where Bayes has long been a standard in text classification, CBayes is an extension of Bayes that performs particularly well on datasets with skewed classes and has been shown to be competitive with algorithms of higher complexity such as Support Vector Machines. + + +## Implementations +The mahout `math-scala` library has an implemetation of both Bayes and CBayes which is further optimized in the `spark` module. Currently the Spark optimized version provides CLI drivers for training and testing. Mahout Spark-Naive-Bayes models can also be trained, tested and saved to the filesystem from the Mahout Spark Shell. + +## Preprocessing and Algorithm + +As described in [[1]](http://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf) Mahout Naive Bayes is broken down into the following steps (assignments are over all possible index values): + +- Let `\(\vec{d}=(\vec{d_1},...,\vec{d_n})\)` be a set of documents; `\(d_{ij}\)` is the count of word `\(i\)` in document `\(j\)`. +- Let `\(\vec{y}=(y_1,...,y_n)\)` be their labels. +- Let `\(\alpha_i\)` be a smoothing parameter for all words in the vocabulary; let `\(\alpha=\sum_i{\alpha_i}\)`. +- **Preprocessing**(via seq2Sparse) TF-IDF transformation and L2 length normalization of `\(\vec{d}\)` + 1. `\(d_{ij} = \sqrt{d_{ij}}\)` + 2. `\(d_{ij} = d_{ij}\left(\log{\frac{\sum_k1}{\sum_k\delta_{ik}+1}}+1\right)\)` + 3. `\(d_{ij} =\frac{d_{ij}}{\sqrt{\sum_k{d_{kj}^2}}}\)` +- **Training: Bayes**`\((\vec{d},\vec{y})\)` calculate term weights `\(w_{ci}\)` as: + 1. `\(\hat\theta_{ci}=\frac{d_{ic}+\alpha_i}{\sum_k{d_{kc}}+\alpha}\)` + 2. `\(w_{ci}=\log{\hat\theta_{ci}}\)` +- **Training: CBayes**`\((\vec{d},\vec{y})\)` calculate term weights `\(w_{ci}\)` as: + 1. `\(\hat\theta_{ci} = \frac{\sum_{j:y_j\neq c}d_{ij}+\alpha_i}{\sum_{j:y_j\neq c}{\sum_k{d_{kj}}}+\alpha}\)` + 2. `\(w_{ci}=-\log{\hat\theta_{ci}}\)` + 3. `\(w_{ci}=\frac{w_{ci}}{\sum_i \lvert w_{ci}\rvert}\)` +- **Label Assignment/Testing:** + 1. Let `\(\vec{t}= (t_1,...,t_n)\)` be a test document; let `\(t_i\)` be the count of the word `\(t\)`. + 2. Label the document according to `\(l(t)=\arg\max_c \sum\limits_{i} t_i w_{ci}\)` + +As we can see, the main difference between Bayes and CBayes is the weight calculation step. Where Bayes weighs terms more heavily based on the likelihood that they belong to class `\(c\)`, CBayes seeks to maximize term weights on the likelihood that they do not belong to any other class. + +## Running from the command line + +Mahout provides CLI drivers for all above steps. Here we will give a simple overview of Mahout CLI commands used to preprocess the data, train the model and assign labels to the training set. An [example script](https://github.com/apache/mahout/blob/master/examples/bin/classify-20newsgroups.sh) is given for the full process from data acquisition through classification of the classic [20 Newsgroups corpus](https://mahout.apache.org/users/classification/twenty-newsgroups.html). + +- **Preprocessing:** +For a set of Sequence File Formatted documents in PATH_TO_SEQUENCE_FILES the [mahout seq2sparse](https://mahout.apache.org/users/basics/creating-vectors-from-text.html) command performs the TF-IDF transformations (-wt tfidf option) and L2 length normalization (-n 2 option) as follows: + + $ mahout seq2sparse + -i ${PATH_TO_SEQUENCE_FILES} + -o ${PATH_TO_TFIDF_VECTORS} + -nv + -n 2 + -wt tfidf + +- **Training:** +The model is then trained using `mahout spark-trainnb`. The default is to train a Bayes model. The -c option is given to train a CBayes model: + + $ mahout spark-trainnb + -i ${PATH_TO_TFIDF_VECTORS} + -o ${PATH_TO_MODEL} + -ow + -c + +- **Label Assignment/Testing:** +Classification and testing on a holdout set can then be performed via `mahout spark-testnb`. Again, the -c option indicates that the model is CBayes: + + $ mahout spark-testnb + -i ${PATH_TO_TFIDF_TEST_VECTORS} + -m ${PATH_TO_MODEL} + -c + +## Command line options + +- **Preprocessing:** *note: still reliant on MapReduce seq2sparse* + + Only relevant parameters used for Bayes/CBayes as detailed above are shown. Several other transformations can be performed by `mahout seq2sparse` and used as input to Bayes/CBayes. For a full list of `mahout seq2Sparse` options see the [Creating vectors from text](https://mahout.apache.org/users/basics/creating-vectors-from-text.html) page. + + $ mahout seq2sparse + --output (-o) output The directory pathname for output. + --input (-i) input Path to job input directory. + --weight (-wt) weight The kind of weight to use. Currently TF + or TFIDF. Default: TFIDF + --norm (-n) norm The norm to use, expressed as either a + float or "INF" if you want to use the + Infinite norm. Must be greater or equal + to 0. The default is not to normalize + --overwrite (-ow) If set, overwrite the output directory + --sequentialAccessVector (-seq) (Optional) Whether output vectors should + be SequentialAccessVectors. If set true + else false + --namedVector (-nv) (Optional) Whether output vectors should + be NamedVectors. If set true else false + +- **Training:** + + $ mahout spark-trainnb + --input (-i) input Path to job input directory. + --output (-o) output The directory pathname for output. + --trainComplementary (-c) Train complementary? Default is false. + --master (-ma) Spark Master URL (optional). Default: "local". + Note that you can specify the number of + cores to get a performance improvement, + for example "local[4]" + --help (-h) Print out help + +- **Testing:** + + $ mahout spark-testnb + --input (-i) input Path to job input directory. + --model (-m) model The path to the model built during training. + --testComplementary (-c) Test complementary? Default is false. + --master (-ma) Spark Master URL (optional). Default: "local". + Note that you can specify the number of + cores to get a performance improvement, + for example "local[4]" + --help (-h) Print out help + +## Examples +1. [20 Newsgroups classification](https://github.com/apache/mahout/blob/master/examples/bin/classify-20newsgroups.sh) +2. [Document classification with Naive Bayes in the Mahout shell](https://github.com/apache/mahout/blob/master/examples/bin/spark-document-classifier.mscala) + + +## References + +[1]: Jason D. M. Rennie, Lawerence Shih, Jamie Teevan, David Karger (2003). [Tackling the Poor Assumptions of Naive Bayes Text Classifiers](http://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf). Proceedings of the Twentieth International Conference on Machine Learning (ICML-2003). + + + http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/old_site_migration/completed/bankmarketing-example.md ---------------------------------------------------------------------- diff --git a/website/old_site_migration/completed/bankmarketing-example.md b/website/old_site_migration/completed/bankmarketing-example.md new file mode 100644 index 0000000..846a4ce --- /dev/null +++ b/website/old_site_migration/completed/bankmarketing-example.md @@ -0,0 +1,53 @@ +--- +layout: default +title: +theme: + name: retro-mahout +--- + +Notice: Licensed to the Apache Software Foundation (ASF) under one + or more contributor license agreements. See the NOTICE file + distributed with this work for additional information + regarding copyright ownership. The ASF licenses this file + to you under the Apache License, Version 2.0 (the + "License"); you may not use this file except in compliance + with the License. You may obtain a copy of the License at + . + http://www.apache.org/licenses/LICENSE-2.0 + . + Unless required by applicable law or agreed to in writing, + software distributed under the License is distributed on an + "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + KIND, either express or implied. See the License for the + specific language governing permissions and limitations + under the License. + +#Bank Marketing Example + +### Introduction + +This page describes how to run Mahout's SGD classifier on the [UCI Bank Marketing dataset](http://mlr.cs.umass.edu/ml/datasets/Bank+Marketing). +The goal is to predict if the client will subscribe a term deposit offered via a phone call. The features in the dataset consist +of information such as age, job, marital status as well as information about the last contacts from the bank. + +### Code & Data + +The bank marketing example code lives under + +*mahout-examples/src/main/java/org.apache.mahout.classifier.sgd.bankmarketing* + +The data can be found at + +*mahout-examples/src/main/resources/bank-full.csv* + +### Code details + +This example consists of 3 classes: + + - BankMarketingClassificationMain + - TelephoneCall + - TelephoneCallParser + +When you run the main method of BankMarketingClassificationMain it parses the dataset using the TelephoneCallParser and trains +a logistic regression model with 20 runs and 20 passes. The TelephoneCallParser uses Mahout's feature vector encoder +to encode the features in the dataset into a vector. Afterwards the model is tested and the learning rate and AUC is printed accuracy is printed to standard output. \ No newline at end of file http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/old_site_migration/completed/breiman-example.md ---------------------------------------------------------------------- diff --git a/website/old_site_migration/completed/breiman-example.md b/website/old_site_migration/completed/breiman-example.md new file mode 100644 index 0000000..d8d049e --- /dev/null +++ b/website/old_site_migration/completed/breiman-example.md @@ -0,0 +1,67 @@ +--- +layout: default +title: Breiman Example +theme: + name: retro-mahout +--- + +#Breiman Example + +#### Introduction + +This page describes how to run the Breiman example, which implements the test procedure described in [Leo Breiman's paper](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.23.3999&rep=rep1&type=pdf). The basic algorithm is as follows : + + * repeat *I* iterations + * in each iteration do + * keep 10% of the dataset apart as a testing set + * build two forests using the training set, one with *m = int(log2(M) + 1)* (called Random-Input) and one with *m = 1* (called Single-Input) + * choose the forest that gave the lowest oob error estimation to compute +the test set error + * compute the test set error using the Single Input Forest (test error), +this demonstrates that even with *m = 1*, Decision Forests give comparable +results to greater values of *m* + * compute the mean testset error using every tree of the chosen forest +(tree error). This should indicate how well a single Decision Tree performs + * compute the mean test error for all iterations + * compute the mean tree error for all iterations + + +#### Running the Example + +The current implementation is compatible with the [UCI repository](http://archive.ics.uci.edu/ml/) file format. We'll show how to run this example on two datasets: + +First, we deal with [Glass Identification](http://archive.ics.uci.edu/ml/datasets/Glass+Identification): download the [dataset](http://archive.ics.uci.edu/ml/machine-learning-databases/glass/glass.data) file called **glass.data** and store it onto your local machine. Next, we must generate the descriptor file **glass.info** for this dataset with the following command: + + bin/mahout org.apache.mahout.classifier.df.tools.Describe -p /path/to/glass.data -f /path/to/glass.info -d I 9 N L + +Substitute */path/to/* with the folder where you downloaded the dataset, the argument "I 9 N L" indicates the nature of the variables. Here it means 1 +ignored (I) attribute, followed by 9 numerical(N) attributes, followed by +the label (L). + +Finally, we build and evaluate our random forest classifier as follows: + + bin/mahout org.apache.mahout.classifier.df.BreimanExample -d /path/to/glass.data -ds /path/to/glass.info -i 10 -t 100 +which builds 100 trees (-t argument) and repeats the test 10 iterations (-i +argument) + +The example outputs the following results: + + * Selection error: mean test error for the selected forest on all iterations + * Single Input error: mean test error for the single input forest on all +iterations + * One Tree error: mean single tree error on all iterations + * Mean Random Input Time: mean build time for random input forests on all +iterations + * Mean Single Input Time: mean build time for single input forests on all +iterations + +We can repeat this for a [Sonar](http://archive.ics.uci.edu/ml/datasets/Connectionist+Bench+%28Sonar,+Mines+vs.+Rocks%29) usecase: download the [dataset](http://archive.ics.uci.edu/ml/machine-learning-databases/undocumented/connectionist-bench/sonar/sonar.all-data) file called **sonar.all-data** and store it onto your local machine. Generate the descriptor file **sonar.info** for this dataset with the following command: + + bin/mahout org.apache.mahout.classifier.df.tools.Describe -p /path/to/sonar.all-data -f /path/to/sonar.info -d 60 N L + +The argument "60 N L" means 60 numerical(N) attributes, followed by the label (L). Analogous to the previous case, we run the evaluation as follows: + + bin/mahout org.apache.mahout.classifier.df.BreimanExample -d /path/to/sonar.all-data -ds /path/to/sonar.info -i 10 -t 100 + + + http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/old_site_migration/completed/classification/bayesian.md ---------------------------------------------------------------------- diff --git a/website/old_site_migration/completed/classification/bayesian.md b/website/old_site_migration/completed/classification/bayesian.md new file mode 100644 index 0000000..51a5c74 --- /dev/null +++ b/website/old_site_migration/completed/classification/bayesian.md @@ -0,0 +1,147 @@ +--- +layout: default +title: +theme: + name: retro-mahout +--- + +# Naive Bayes + + +## Intro + +Mahout currently has two Naive Bayes implementations. The first is standard Multinomial Naive Bayes. The second is an implementation of Transformed Weight-normalized Complement Naive Bayes as introduced by Rennie et al. [[1]](http://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf). We refer to the former as Bayes and the latter as CBayes. + +Where Bayes has long been a standard in text classification, CBayes is an extension of Bayes that performs particularly well on datasets with skewed classes and has been shown to be competitive with algorithms of higher complexity such as Support Vector Machines. + + +## Implementations +Both Bayes and CBayes are currently trained via MapReduce Jobs. Testing and classification can be done via a MapReduce Job or sequentially. Mahout provides CLI drivers for preprocessing, training and testing. A Spark implementation is currently in the works ([MAHOUT-1493](https://issues.apache.org/jira/browse/MAHOUT-1493)). + +## Preprocessing and Algorithm + +As described in [[1]](http://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf) Mahout Naive Bayes is broken down into the following steps (assignments are over all possible index values): + +- Let `\(\vec{d}=(\vec{d_1},...,\vec{d_n})\)` be a set of documents; `\(d_{ij}\)` is the count of word `\(i\)` in document `\(j\)`. +- Let `\(\vec{y}=(y_1,...,y_n)\)` be their labels. +- Let `\(\alpha_i\)` be a smoothing parameter for all words in the vocabulary; let `\(\alpha=\sum_i{\alpha_i}\)`. +- **Preprocessing**(via seq2Sparse) TF-IDF transformation and L2 length normalization of `\(\vec{d}\)` + 1. `\(d_{ij} = \sqrt{d_{ij}}\)` + 2. `\(d_{ij} = d_{ij}\left(\log{\frac{\sum_k1}{\sum_k\delta_{ik}+1}}+1\right)\)` + 3. `\(d_{ij} =\frac{d_{ij}}{\sqrt{\sum_k{d_{kj}^2}}}\)` +- **Training: Bayes**`\((\vec{d},\vec{y})\)` calculate term weights `\(w_{ci}\)` as: + 1. `\(\hat\theta_{ci}=\frac{d_{ic}+\alpha_i}{\sum_k{d_{kc}}+\alpha}\)` + 2. `\(w_{ci}=\log{\hat\theta_{ci}}\)` +- **Training: CBayes**`\((\vec{d},\vec{y})\)` calculate term weights `\(w_{ci}\)` as: + 1. `\(\hat\theta_{ci} = \frac{\sum_{j:y_j\neq c}d_{ij}+\alpha_i}{\sum_{j:y_j\neq c}{\sum_k{d_{kj}}}+\alpha}\)` + 2. `\(w_{ci}=-\log{\hat\theta_{ci}}\)` + 3. `\(w_{ci}=\frac{w_{ci}}{\sum_i \lvert w_{ci}\rvert}\)` +- **Label Assignment/Testing:** + 1. Let `\(\vec{t}= (t_1,...,t_n)\)` be a test document; let `\(t_i\)` be the count of the word `\(t\)`. + 2. Label the document according to `\(l(t)=\arg\max_c \sum\limits_{i} t_i w_{ci}\)` + +As we can see, the main difference between Bayes and CBayes is the weight calculation step. Where Bayes weighs terms more heavily based on the likelihood that they belong to class `\(c\)`, CBayes seeks to maximize term weights on the likelihood that they do not belong to any other class. + +## Running from the command line + +Mahout provides CLI drivers for all above steps. Here we will give a simple overview of Mahout CLI commands used to preprocess the data, train the model and assign labels to the training set. An [example script](https://github.com/apache/mahout/blob/master/examples/bin/classify-20newsgroups.sh) is given for the full process from data acquisition through classification of the classic [20 Newsgroups corpus](https://mahout.apache.org/users/classification/twenty-newsgroups.html). + +- **Preprocessing:** +For a set of Sequence File Formatted documents in PATH_TO_SEQUENCE_FILES the [mahout seq2sparse](https://mahout.apache.org/users/basics/creating-vectors-from-text.html) command performs the TF-IDF transformations (-wt tfidf option) and L2 length normalization (-n 2 option) as follows: + + mahout seq2sparse + -i ${PATH_TO_SEQUENCE_FILES} + -o ${PATH_TO_TFIDF_VECTORS} + -nv + -n 2 + -wt tfidf + +- **Training:** +The model is then trained using `mahout trainnb` . The default is to train a Bayes model. The -c option is given to train a CBayes model: + + mahout trainnb + -i ${PATH_TO_TFIDF_VECTORS} + -o ${PATH_TO_MODEL}/model + -li ${PATH_TO_MODEL}/labelindex + -ow + -c + +- **Label Assignment/Testing:** +Classification and testing on a holdout set can then be performed via `mahout testnb`. Again, the -c option indicates that the model is CBayes. The -seq option tells `mahout testnb` to run sequentially: + + mahout testnb + -i ${PATH_TO_TFIDF_TEST_VECTORS} + -m ${PATH_TO_MODEL}/model + -l ${PATH_TO_MODEL}/labelindex + -ow + -o ${PATH_TO_OUTPUT} + -c + -seq + +## Command line options + +- **Preprocessing:** + + Only relevant parameters used for Bayes/CBayes as detailed above are shown. Several other transformations can be performed by `mahout seq2sparse` and used as input to Bayes/CBayes. For a full list of `mahout seq2Sparse` options see the [Creating vectors from text](https://mahout.apache.org/users/basics/creating-vectors-from-text.html) page. + + mahout seq2sparse + --output (-o) output The directory pathname for output. + --input (-i) input Path to job input directory. + --weight (-wt) weight The kind of weight to use. Currently TF + or TFIDF. Default: TFIDF + --norm (-n) norm The norm to use, expressed as either a + float or "INF" if you want to use the + Infinite norm. Must be greater or equal + to 0. The default is not to normalize + --overwrite (-ow) If set, overwrite the output directory + --sequentialAccessVector (-seq) (Optional) Whether output vectors should + be SequentialAccessVectors. If set true + else false + --namedVector (-nv) (Optional) Whether output vectors should + be NamedVectors. If set true else false + +- **Training:** + + mahout trainnb + --input (-i) input Path to job input directory. + --output (-o) output The directory pathname for output. + --alphaI (-a) alphaI Smoothing parameter. Default is 1.0 + --trainComplementary (-c) Train complementary? Default is false. + --labelIndex (-li) labelIndex The path to store the label index in + --overwrite (-ow) If present, overwrite the output directory + before running job + --help (-h) Print out help + --tempDir tempDir Intermediate output directory + --startPhase startPhase First phase to run + --endPhase endPhase Last phase to run + +- **Testing:** + + mahout testnb + --input (-i) input Path to job input directory. + --output (-o) output The directory pathname for output. + --overwrite (-ow) If present, overwrite the output directory + before running job + + + --model (-m) model The path to the model built during training + --testComplementary (-c) Test complementary? Default is false. + --runSequential (-seq) Run sequential? + --labelIndex (-l) labelIndex The path to the location of the label index + --help (-h) Print out help + --tempDir tempDir Intermediate output directory + --startPhase startPhase First phase to run + --endPhase endPhase Last phase to run + + +## Examples + +Mahout provides an example for Naive Bayes classification: + +1. [Classify 20 Newsgroups](twenty-newsgroups.html) + +## References + +[1]: Jason D. M. Rennie, Lawerence Shih, Jamie Teevan, David Karger (2003). [Tackling the Poor Assumptions of Naive Bayes Text Classifiers](http://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf). Proceedings of the Twentieth International Conference on Machine Learning (ICML-2003). + + http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/old_site_migration/completed/classification/class-discovery.md ---------------------------------------------------------------------- diff --git a/website/old_site_migration/completed/classification/class-discovery.md b/website/old_site_migration/completed/classification/class-discovery.md new file mode 100644 index 0000000..a24cc14 --- /dev/null +++ b/website/old_site_migration/completed/classification/class-discovery.md @@ -0,0 +1,155 @@ +--- +layout: default +title: Class Discovery +theme: + name: retro-mahout +--- +<a name="ClassDiscovery-ClassDiscovery"></a> +# Class Discovery + +See http://www.cs.bham.ac.uk/~wbl/biblio/gecco1999/GP-417.pdf + +CDGA uses a Genetic Algorithm to discover a classification rule for a given +dataset. +A dataset can be seen as a table: + +<table> +<tr><th> </th><th>attribute 1</th><th>attribute 2</th><th>...</th><th>attribute N</th></tr> +<tr><td>row 1</td><td>value1</td><td>value2</td><td>...</td><td>valueN</td></tr> +<tr><td>row 2</td><td>value1</td><td>value2</td><td>...</td><td>valueN</td></tr> +<tr><td>...</td><td>...</td><td>...</td><td>...</td><td>...</td></tr> +<tr><td>row M</td><td>value1</td><td>value2</td><td>...</td><td>valueN</td></tr> +</table> + +An attribute can be numerical, for example a "temperature" attribute, or +categorical, for example a "color" attribute. For classification purposes, +one of the categorical attributes is designated as a *label*, which means +that its value defines the *class* of the rows. +A classification rule can be represented as follows: +<table> +<tr><th> </th><th>attribute 1</th><th>attribute 2</th><th>...</th><th>attribute N</th></tr> +<tr><td>weight</td><td>w1</td><td>w2</td><td>...</td><td>wN</td></tr> +<tr><td>operator</td><td>op1</td><td>op2</td><td>...</td><td>opN</td></tr> +<tr><td>value</td><td>value1</td><td>value2</td><td>...</td><td>valueN</td></tr> +</table> + +For a given *target* class and a weight *threshold*, the classification +rule can be read : + + + for each row of the dataset + if (rule.w1 < threshold || (rule.w1 >= threshold && row.value1 rule.op1 +rule.value1)) && + (rule.w2 < threshold || (rule.w2 >= threshold && row.value2 rule.op2 +rule.value2)) && + ... + (rule.wN < threshold || (rule.wN >= threshold && row.valueN rule.opN +rule.valueN)) then + row is part of the target class + + +*Important:* The label attribute is not evaluated by the rule. + +The threshold parameter allows some conditions of the rule to be skipped if +their weight is too small. The operators available depend on the attribute +types: +* for a numerical attributes, the available operators are '<' and '>=' +* for categorical attributes, the available operators are '!=' and '==' + +The "threshold" and "target" are user defined parameters, and because the +label is always a categorical attribute, the target is the (zero based) +index of the class label value in all the possible values of the label. For +example, if the label attribute can have the following values (blue, brown, +green), then a target of 1 means the "blue" class. + +For example, we have the following dataset (the label attribute is "Eyes +Color"): +<table> +<tr><th> </th><th>Age</th><th>Eyes Color</th><th>Hair Color</th></tr> +<tr><td>row 1</td><td>16</td><td>brown</td><td>dark</td></tr> +<tr><td>row 2</td><td>25</td><td>green</td><td>light</td></tr> +<tr><td>row 3</td><td>12</td><td>blue</td><td>light</td></tr> +and a classification rule: +<tr><td>weight</td><td>0</td><td>1</td></tr> +<tr><td>operator</td><td><</td><td>!=</td></tr> +<tr><td>value</td><td>20</td><td>light</td></tr> +and the following parameters: threshold = 1 and target = 0 (brown). +</table> + +This rule can be read as follows: + + for each row of the dataset + if (0 < 1 || (0 >= 1 && row.value1 < 20)) && + (1 < 1 || (1 >= 1 && row.value2 != light)) then + row is part of the "brown Eye Color" class + + +Please note how the rule skipped the label attribute (Eye Color), and how +the first condition is ignored because its weight is < threshold. + +<a name="ClassDiscovery-Runningtheexample:"></a> +# Running the example: +NOTE: Substitute in the appropriate version for the Mahout JOB jar + +1. cd <MAHOUT_HOME>/examples +1. ant job +1. {code}<HADOOP_HOME>/bin/hadoop dfs -put +<MAHOUT_HOME>/examples/src/test/resources/wdbc wdbc{code} +1. {code}<HADOOP_HOME>/bin/hadoop dfs -put +<MAHOUT_HOME>/examples/src/test/resources/wdbc.infos wdbc.infos{code} +1. {code}<HADOOP_HOME>/bin/hadoop jar +<MAHOUT_HOME>/examples/build/apache-mahout-examples-0.1-dev.job +org.apache.mahout.ga.watchmaker.cd.CDGA +<MAHOUT_HOME>/examples/src/test/resources/wdbc 1 0.9 1 0.033 0.1 0 100 10 + + CDGA needs 9 parameters: + * param 1 : path of the directory that contains the dataset and its infos +file + * param 2 : target class + * param 3 : threshold + * param 4 : number of crossover points for the multi-point crossover + * param 5 : mutation rate + * param 6 : mutation range + * param 7 : mutation precision + * param 8 : population size + * param 9 : number of generations before the program stops + + For more information about 4th parameter, please see [Multi-point Crossover|http://www.geatbx.com/docu/algindex-03.html#P616_36571] +. + For a detailed explanation about the 5th, 6th and 7th parameters, please +see [Real Valued Mutation|http://www.geatbx.com/docu/algindex-04.html#P659_42386] +. + + *TODO*: Fill in where to find the output and what it means. + + h1. The info file: + To run properly, CDGA needs some informations about the dataset. Each +dataset should be accompanied by an .infos file that contains the needed +informations. for each attribute a corresponding line in the info file +describes it, it can be one of the following: + * IGNORED + if the attribute is ignored + * LABEL, val1, val2,... + if the attribute is the label (class), and its possible values + * CATEGORICAL, val1, val2,... + if the attribute is categorial (nominal), and its possible values + * NUMERICAL, min, max + if the attribute is numerical, and its min and max values + + This file can be generated automaticaly using a special tool available with +CDGA. + + + +* the tool searches for an existing infos file (*must be filled by the +user*), in the same directory of the dataset with the same name and with +the ".infos" extension, that contain the type of the attributes: + ** 'N' numerical attribute + ** 'C' categorical attribute + ** 'L' label (this also a categorical attribute) + ** 'I' to ignore the attribute + each attribute is in a separate +* A Hadoop job is used to parse the dataset and collect the informations. +This means that *the dataset can be distributed over HDFS*. +* the results are written back in the same .info file, with the correct +format needed by CDGA. http://git-wip-us.apache.org/repos/asf/mahout/blob/c81fc8b7/website/old_site_migration/completed/classification/classifyingyourdata.md ---------------------------------------------------------------------- diff --git a/website/old_site_migration/completed/classification/classifyingyourdata.md b/website/old_site_migration/completed/classification/classifyingyourdata.md new file mode 100644 index 0000000..c2099c0 --- /dev/null +++ b/website/old_site_migration/completed/classification/classifyingyourdata.md @@ -0,0 +1,27 @@ +--- +layout: default +title: ClassifyingYourData +theme: + name: retro-mahout +--- + +# Classifying data from the command line + + +After you've done the [Quickstart](../basics/quickstart.html) and are familiar with the basics of Mahout, it is time to build a +classifier from your own data. The following pieces *may* be useful for in getting started: + +<a name="ClassifyingYourData-Input"></a> +# Input + +For starters, you will need your data in an appropriate Vector format: See [Creating Vectors](../basics/creating-vectors.html) as well as [Creating Vectors from Text](../basics/creating-vectors-from-text.html). + +<a name="ClassifyingYourData-RunningtheProcess"></a> +# Running the Process + +* Logistic regression [background](logistic-regression.html) +* [Naive Bayes background](naivebayes.html) and [commandline](bayesian-commandline.html) options. +* [Complementary naive bayes background](complementary-naive-bayes.html), [design](https://issues.apache.org/jira/browse/mahout-60.html), and [c-bayes-commandline](c-bayes-commandline.html) +* [Random Forests Classification](https://cwiki.apache.org/confluence/display/MAHOUT/Random+Forests) comes with a [Breiman example](breiman-example.html). There is some really great documentation +over at [Mark Needham's blog](http://www.markhneedham.com/blog/2012/10/27/kaggle-digit-recognizer-mahout-random-forest-attempt/). Also checkout the description on [Xiaomeng Shawn Wan +s](http://shawnwan.wordpress.com/2012/06/01/mahout-0-7-random-forest-examples/) blog. \ No newline at end of file
