Research Of Stock Up And Down Multi-Classification Based On Ada-CDT Model

Stocks are a high-risk and high-yield investment project in which investors use the price changes to get the difference.This paper proposes a new model to classify the stock's ups and downs.According to the results,investors will propose to buy,hold or sell at the opening of the second day.In this paper,a new adaptive cascade decision tree model(Ada-CDT)is proposed,which combines cascading structure and Adaboost.The innovations of this paper are as follows:One.Introduce different misclassification cost coefficients to improve the algorithm;Second,establish a new model that each level of the cascade structure consists of a decision tree,the weak classifier is linked by Adaboost rules,here is the improved algorithm after introducing the cost coefficient;Third,the stock price change rate label is studied,the classification label threshold is determined,and' the selection is used as the label classification value.And conduct correlation analysis with stocks of financial securities type,and select strong correlation stocks for empirical analysis.Apply the new model Ada-CDT to the stock classification,classify the stock according to the income,use the stock of CITIC Securities(code:600030)as the base stock,and then select 13 stocks with other strong correlations for empirical analysis.The ROC curve and the AUC value are used to evaluate the model and compare the four algorithms of gc-forest,C4.5,Random Forest and Adaboost.It is found that the constructed Ada-CDT model has a good classification effect on securities stocks compared to other algorithms.