| Popis: |
The stock market is a complicated system contained different information from diverse categories, strategies on predicting the trend of stock index has become a significant problem which attracts lots of concern. In the study, prediction on daily up and down of stock index was made based on machine learning, data of CSI300 index from 2010 to 2020 was collected from financial website as sample set and 30 factors was selected by feature selection as the input of model. Then predictive model was optimized by adding the sliding window, grid search and cross validation, through experiments on different classifiers of model including SVM, K nearest neighbor, Adaboost classifier, random forest, MLP classifier and comparison of results, the MLP classifier was the optimal classifier among all tested classifiers whose accumulative value was 1.922 and Sharpe ratio was 0.64. By using the optimal classifier of model to predict the trend of stock index, we can gain a high and stable return. |
