Forecasting using neural networks in Matlab -
i doing project on traffic forecasting particular road junction. part of data collection process, have collected count of number of vehicles on particular day (from 8 12 @ 15 minute time intervals each) passing through junction. following time-series after consolidation.
vehicle count
2619(8:00 8:15 am)
2925(8:15 8:30 am)
3090(8:30 8:45 am)
3075(8:45 9:00 am)
2786(9:00 9:15 am)
3403(9:15 9:30 am)
3460(9:30 9:45 am)
3122(9:45 10:00 am)
2977(10:00 10:15 am)
3000(10:15 10:30 am)
3081(10:30 10:45 am)
2842(10:45 11:00 am)
2691(11:00 11:15 am)
2816(11:15 11:30 am)
2878(11:30 11:45 am)
2965(11:45 12:00 am)
i trying develop simple short term forecasting model using neural network toolbox inbuilt in matlab should capable of forecasting next value in time series. instance, if supply first 15 values of above mentioned time series input, model should forecast 16th value in time series.
i have following doubts:
1.) since using single array i.e. vehicle count in problem, ok if use 'non-linear input output' option provided in nn toolbox instead of nar/narx?
2.)suppose wish supply first 15 values input set, best ratio can used divide data set training set,validation set , testing set given problem?
3.) how see forecasted 16th value in nn toolbox in matlab?
i beginner in field. highly appreciated.
if predicting series, series past values input, using nonlinear input-output series both inputs , targets equivalent using nar network.
the nar network give ability close loop, , perform multiple predictions past known data. otherwise 2 networks operate same.
Comments
Post a Comment