We need a generic algo which function as a windows app that will log into a specific MS-SQL2008 database and within +/-3 accurately 'predict' a "1" or "0" within a specific data set based on sub-sequences of that data within another data set..
Text Example (Also please see image example).
'prediction' data set 1 - 0,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,1
'sub-sequence' data set 2 - 0,1,0,-1,1,1,-1,0,-1,0,1,-1,-1,0,1
The 'prediction' data set will always go from 0 to 1 or from 1 to zero and everything in between that on the other data set will be our sub-sequence.
The calculation will need to be done in less than 5 minutes because the database updates with 'fresh' data every five minutes. When new data is entered into the database the algo will need to start the prediction again.
Real-time data is fed into the model, which start calculating probabilities after the 'sub-sequence' data arrival.
As more and more real-time 'sub-sequence' data is entered into the model - more and more probabilities 'should' eliminated - which should make the calculation easier and faster.
There will be 3 results when the prediction is done.
A prediction of '1'
A prediction of '0'
A prediction of '-1' or null.
The best way I've read to do this is to divide data into sets using sub-strings of length 4-13.