A feature selection method for mobile users outbound based on fisher score and approximate Markov blanket

Abstract

The invention relates to a method for selecting a mobile user's outbound feature based on Fisher points and an approximate Markov blanket, which belongs to the field of data mining. Firstly, the Fisher criterion is used to retain the features with strong classification ability, and the irrelevant and weakly relevant features are eliminated. Secondly, the two measurement methods of maximum information coefficient MIC and symmetric uncertainty SU are combined, and the correlation measurement standard MSCC is designed, and the irrelevant features are further eliminated by using the MSCC standard. Finally, combined with the MSCC metric, the redundant features in the Fisher candidate feature set are eliminated by using the approximate Markov-Blanket judgment condition, and finally the optimal feature subset with a smaller dimension is obtained. The invention can effectively select the outbound features of the mobile users and improve the classification accuracy of the model.

Description

A mobile user outbound feature selection based on Fisher score and approximate Markov blanket choose method technical field The invention belongs to the technical field of data mining, relate to a kind of movement based on Fisher score and approximate Markov blanket User outbound feature selection method. Background technique With the advent of the mobile Internet era, the scale of mobile Internet users continues to increase, and people's living and working styles are It was very different before, the high penetration rate of mobile devices brought the explosive growth of mobile data. Mobile data has data sampling The advantages of comprehensive and real-time performance are quite authoritative in the field of trend analysis and potential user mining. Digging provides good convenience. Feature selection is a key data preprocessing step in machine learning and data mining, which is to filter the most The process of reducing the feature dimension of the dataset ...

AI Technical Summary

Problems solved by technology

[0004] Mobile data generally contains high-dimensional features and is non-linear data. When the number of samples is limited, if a large number of features are used to design a classifier, the computational overhead is high, the classification performance is poor, and overfitting is prone to occur.

However, the above methods have the disadvantages of being easy to fall into local optimum and easy to delete useful features.

Images