Comrades,
Could anyone point to research (academic, practitioners, etc.) on the application of machine learning techniques to time-series analysis, other than Neural Networks ?
I personally feel that when dealing with say time series of prices the situation is not the same as when figuring out what one would want to buy given his/her previous purchases.
There is a lot of shifting preferences through time (bubbles, etc), learning by market participants, non-stationarity, regime switching, etc. that may make applying ML techniques directly less effective.
Much appreciate any advice.
Machine Learning for time series analysis ?
- superconductor
- Posts: 0
- Joined: Thu Jan 01, 2004 12:00 am
Machine Learning for time series analysis ?
Try Recurrent Neural Networks. Those suit time series analysis much better. Google Scholar is a good place to start.
-
gelfand
- Posts: 0
- Joined: Thu Jan 01, 2004 12:00 am
Machine Learning for time series analysis ?
An old (1993) but good book is
Time Series Prediction: Forecasting The Future And Understanding The Past (Santa Fe Institute Series)
by Andreas S. Weigend (Author), Neil A. Gershenfeld (Editor)
Time Series Prediction: Forecasting The Future And Understanding The Past (Santa Fe Institute Series)
by Andreas S. Weigend (Author), Neil A. Gershenfeld (Editor)