I have a java project forecasting stock price next day.
Java code implement weka forecaster 'eat' a csv file named '[login to view URL]' (same as uploaded [login to view URL]) and write forecast result to '[login to view URL]'.
Java code also call [login to view URL] to run 'FC.R' to 'eat' the same Dataset file '[login to view URL]' and write forecast result to '[login to view URL]'.
Problem is my org FC.R script just do basic time-series forecast but what I need is:
1) Forecast should also consider the correlation with other columns.
For example, in uploaded file [login to view URL], Columns B ~ J should be put into forecast calculation.
2) Lag and Seasonal parameters of can be set.
3) In [login to view URL] green columns are data developed from history records except the last instance that is developed by stock open price of the date to be forecast.
So, in the uploaded file cell I23 & J23 are the assumed value during running the script.
In weka, I put into overlay instances and pass to weka forecaster. If r-script not support this mode, may skip this approach.
4) '[login to view URL]' structure is not fixed except first column always is date column.
Orange and Green columns set may size from 0 to unlimited.
Yellow columns set may size from 1 to unlimited.
5) 'Logarithm' at least contains Arima, ETS & RWF. Additional forecasters are appreciated, of course.
6) DataSet length is not fixed. In [login to view URL] contains only Rows 2~23, in fact dataset may contains more than thousands of rows.
7) The r-script will be written by java line by line.
Please inform which value should be changed in the new r-script vary with changing structure.
8) '[login to view URL]' consist of two Sheets. One for forecast value & another for accuracy value.
Above points are based on my knowledge from Weka, if r-script not possible work similarly, please leave your proposal.