In this project, a new proposed classification algorithms, Deep Forest, was combined with text mining technique to predict the price trend in financial market. Specifically, the project was involved with some tech methods, incl. crawling news data into database, training corpus of given text bag, performing sentiment analysis, and machine learning model training & predicting. Obtained sentiment data and some technical indicators were assembled to predict the price trend. Out-of-sample test achieved an accuracy of 62.8%.
Slicing Sequence...
Training MGS Random Forests...
Adding/Training Layer, n_layer=1
Layer validation accuracy = 0.615384615385
Adding/Training Layer, n_layer=2
Layer validation accuracy = 0.553846153846
Slicing Sequence...
Total161times forecasting with a accuracy: 0.632173913043
gcForest accuracy : 0.628385093168