Data analysis and forecasting.

Several years ago I had a freelancer write a Matlab app for me. Unfortunately, the data source became obsolete and I had to set about finding a new source for quality data. I have done that now, and have also upgraded my Matlab to R2017a, which is installed but hasn't been used yet. The data is in .csv format and the layout is: Columns A-L. For my information only Columns M - CQ. Input data Column CR. Output (headed "FinPos") I have 2 sets of data and am uploading samples (2,000 rows) of both. Sample A has over 140,000 rows of data in the full set Sample B has over 97,000 rows of data in the full set Both data sets have the same layout, number of columns etc, but sample B has a lot more "0" values in it. I would expect that an app that is efficient on Sample A, should be the same on sample B; I don't believe that merging the two sets would be advantageous, but will be guided by you. The output column has numbers ranging from 1 to 50. My primary interest is in the "1" output, and I would expect a minimum accuracy of 40% in forecasting this. I have a small interest in the "2" and "3" outputs, and would expect an accuracy of 25%+ for these. The other figures I am not concerned about and won't be relying on them in any way. So, to sum up,; I need an app that will analyse the data and results in the full data sets and then provide me with a forecast output when new given data. I am looking for this to be as simple as possible for me to run...........I train the network by loading the dataset, and then get a forecast by loading and running the new data. I have a pc that will be dedicated to this and will be happy to give remote access via teamviewer for the successful coder to install their work and instruct me on it's use. I will also supply the full data sets to the successful bidder. I am getting more data daily, so there will usually be some available for testing purposes without the output filled in, but which I will have actual results for. If it is of any help, the last freelancer also used "Weka" with Matlab (I don't have that installed now, but am happy for it to be installed if it helps), and he achieved results well in excess of my minimum requirements. I'm sure that for someone fully conversant with neurals and Matlab 2017a, this won't be a big challenge.