Feature selection in sequential data What methods can be used for feature selection in sequential data. What are the methods better than Random forest. Are there any feature extraction methods available for sequential data
 A: A bit more info on the nature of the problem would be helpful, but, in general, data being "sequential" if often a tip that the most appropriate approach is to look at the data as a time-series. Note that "time-series" does not necessarily mean that the the events occur at particular time intervals, so far as you can reasonably assign numerical time values to the sequential events. 
Approaches like random forest tend to be appropriate when looking at panel data. For a great explanation of the differences between panel and time-series data check out this stack exchange article. 
As described on the well-written wikipedia article on time series... 

Time series analysis comprises methods for analyzing time series data in order to extract meaningful statistics and other characteristics of the data

There are quite a bit of approaches, depending on the nature of your problem, but good places to start include exploration of:
- Autocorrelation (to examine )
- Spectral analysis (to examine cyclic behavior)
- Decomposition (to try to break the signal up into the various components that sum to it)  
If you'd like to explore models, first step is to understand which of the broad classes is more appropriate: autoregressive (AR) models, integrated (I) models, or  moving average (MA) models. Combos of these can be looked at using autoregressive moving average (ARMA) and autoregressive integrated moving average (ARIMA) models. Your question sparked my curiousity and I found this well-written, well-organized resource: http://www.itl.nist.gov/div898/handbook/pmc/section4/pmc4.htm 
From there you can go down a beautiful rabbit-hole of approaches. 
Good luck with your analysis!
