I am in the process of compiling a large volume of data into csv format for importing into Matlab.
There will be approximately 30 columns of numeric data, ranging from about -100 to +500 and some columns will have up to 6 decimal places.
There will be 1 output column with whole number values of 1 - 5 inc. (Prediction output can use up to 2 decimal places if that helps accuracy, but my training data has no decimal places). I need high levels of accuracy at output value 1 (75% minimum) and output value 2 (65% minimum), output value 3 can then have whatever else is available as output value 4 is of only minor interest and output value 5 is no use to me whatsoever.
I will have a sample csv available sometime tomorrow with about 10 - 11,000 rows of data, and I expect there to be in the region of 300,000 rows once I have completed processing; I will then be updating once a month - adding 3 - 5,000 rows and re-training data.
The project will be awarded within 3 - 4 days if a suitable offer is made, and completion must be within 7 days of award.
I have now uploaded a csv with example training data. Columns A-Z are data and column AA is the required output.
Following discussions with one bidder, I have decided that the output column should only contain classes 1-3, not 1-5 (as in the first sample file). A new sample file with 30,000 rows of data has been uploaded.
10 freelancers are bidding on average £159 for this job
I can give you code which will do the predictions using [url removed, login to view] accuracy will depend on the strength of relation between the data and the target. the accuracy can change/vary dramatically if selected columns are used. Daha fazlası
Dear Sir, I have more than 7 years experience implement algorithms in matlab and c++. I believe that i can handle this project in easy way. Thanks, Pradeep