0

I need a forecasting model that will use time series as well as seasonality. Example, to predict Feb 2013 data, I will use Jan 2013 data and Feb 2012 data.

I was trying to assess TripleExponentialSmoothing and SVM (as used in the Weka forecasting library). However, with limited statistical prowess, I am not sure what(and probably how) to leverage.

I need this as a part of an enterprise Java application. Is there a Java library that I could use?

halfer
  • 19,824
  • 17
  • 99
  • 186
sutanu dalui
  • 663
  • 7
  • 25
  • I see the question has been voted down. My apologies if this is unclear. If someone can atleast confirm whether I am thinking in the right direction that would be enough. – sutanu dalui Jan 30 '13 at 11:22
  • Triple Exponential Smoothing (if I remember correct, it's the version with seasonality) is actually very easy to code (at most 30 lines of `python`. Why not just implement it yourself? :) – greeness Jan 31 '13 at 21:12

2 Answers2

3

Take a look at Encog with Time Series Predication.

Once you have learned the concept of time series prediction with Encog, using weka and SVM will be the same thing.

USchneider
  • 153
  • 1
  • 6
1

Maybe WEKA is not the best tool for that stuff. You can use R from Java through JRI: http://stats.math.uni-augsburg.de/JRI/

Here is a small and quick tutorial about time-series analysis in R: http://www.simafore.com/blog/bid/105815/Time-series-analysis-using-R-for-cost-forecasting-models-in-8-steps

arutaku
  • 5,937
  • 1
  • 24
  • 38
  • Thanks for your response. I believe using R has a learning curve in itself. I had assessed RJava initially. If I am not mistaken, it is basically a JNI bridge for connecting to R instance from java. Thus I need to know R scripting for that, which unfortunately isn't my forte. – sutanu dalui Jan 30 '13 at 13:02