Hi, I've used Lucene a fair bit and one useful feature it has is the ability to boost fields to make them more relevant. E.g. matching Titles are more important than matching descriptions, so you can "boost" title fields to ensure they weigh in more in the final relevance calculation.
I expected there to be a similar concept of boosting or weightings with the bayes classifier too, but I can't work how to make that work. >From the examples, there's the target variable and the predictor variable and that's it in mahouts implementation. I guess I could fake boosting by duplicating phrases inside the predictor variable. E.g. When Classifying "Electronic Games", there is a "platform" feature, this feature is super important and weighs heavier than the "title" feature. Can anyone suggest what mahouts what approach is to weighing features in naive bayes classification? Thanks, V
