Hi Srinath,

I will try this with some large datasets and  will find and read few
articles related to "moving window" approach.

Yes, this  was written in IPython notebook using Scikit-learn and
Matplotlib.

Thanks,
Upul


On Mon, Nov 30, 2015 at 8:28 AM, Srinath Perera <[email protected]> wrote:

> I am adding dev@ for record.
>
> Upul, this is beautiful :) !!
>
> I think we can do this with released components? Basically build the model
> in ML and use that in IoT product then.
>
> Also I think we should try this with more datasets + read about this ( I
> guess someone must have done this before), and add it to ML wizard.
> Currently we do not have a good way to predict autocorrelated data.
>
> Nirmal, we should use the same with United secuirty check time use case.
>
> I will be in Trace in the PM. Let's chat.
>
> Thanks
> Srinath
>
> p.s. what are you using to try this out? (ipynb thing).
>
> On Sun, Nov 29, 2015 at 10:17 PM, Upul Bandara <[email protected]> wrote:
>
>> Hi Srinath,
>>
>> As we discussed Friday morning, few time series models were trained using
>> simple linear regression with moving windows.
>> Three small (but real-world) datasets were used for training those
>> models.
>>
>> According to the generated prediction graphs, it looks like "moving
>> windows" are capable of (correctly) identifying trends in time series
>> graphs.
>>
>> Prediction graphs are shown in [1].
>>
>>
>> [1].
>> https://github.com/upul/scratch/blob/master/Time_Series_Using_Linear_Regression.ipynb
>>
>> Thanks,
>> Upul
>>
>> --
>> Upul Bandara,
>> Mob: +94 715 468 345.
>>
>
>
>
> --
> ============================
> Blog: http://srinathsview.blogspot.com twitter:@srinath_perera
> Site: http://people.apache.org/~hemapani/
> Photos: http://www.flickr.com/photos/hemapani/
> Phone: 0772360902
>



-- 
Upul Bandara,
Mob: +94 715 468 345.
_______________________________________________
Dev mailing list
[email protected]
http://wso2.org/cgi-bin/mailman/listinfo/dev

Reply via email to