Design of online learning behaviour feature mining method based on decision tree

In order to solve the problems of traditional feature mining methods, such as low precision of feature extraction and high time cost of mining, this paper proposes an online learning behaviour feature mining method based on decision tree. SVM is used to obtain online learning behaviour data and heterogeneous support vector, with online learning behaviour feature data extracted by transforming data form. Then, the behaviour feature data is preprocessed by the agglomerative hierarchical clustering method. Based on the analysis of the principle of decision tree, the root information gain maximisation data is obtained, and the online learning behaviour feature mining is realised by correcting the leaf node error. The experimental results show that the feature extraction accuracy of this method can reach 98%, and the mining time is always less than 2.5 s, which proves that it can meet the design expectations.