Ciriculumvite Trading Competition

__Initial results__

1. Neural Nets with window size of 1
   Accuracy :: 0.326 +/- 0.364
As expected, not a good result. Observing only the last record to predict the next one is expected to be worse than 50%. 
Currently running with window size of 10. This process it taking long as I have used 31+1 parameters. To quicken the setup, removing some of the given indicators will be useful as most indicators are redundant in information.

2. SVM with window size of 1
   Accuracy :: 0.163 +/- 0.281
Same reason as above. 

More importantly in Stock prices (or any other instrument), it is sufficient to predict the change of the direction with high confidence and it is not necessary to predict the actual price.
  
3. Multivariate Regression Analysis Prediction
{{:research:difference_predicted_price.png|}}