Stock data come from Yahoo_finance by Python.
News data come from tm.plugin by R.
1.Use Daubechies 4 wavelet to transform the Stock Data which comes from Yahoo_finance.
2.Difference the time series make it stationary.
3.Create ACF & Pacf pictures to find out p & q which is the parameter in ARIMA.
4.Predict the stationary time series by ARIMA(p,q). Because this ARIMA package can't do difference bigger than 2, thus I don't use ARIMA(p,d,q).
5.Revert difference which we do in step 2.
| ARIMA |
|---|
 |
Not good enough. I try to transform Price to the relation, like the relation between Open & Close attributes or today & yesterday. Stock can't be Predicted only based on history stock data, so we pull in new data. It's still not good but much better than before.
| SVM |
|---|
 |