Improving depression prediction using a novel feature selection algorithm coupled with context-aware analysis
A multivariate filter, namely MIC (Maximal Information Coefficient)-based minimum Redundancy Maximum Relevance (mRMR), to obtain importance ranking of features.
A wrapper, namely sequential involvement and eliminating redundancy with support vector machine (SIER-SVM), to determine the optimal feature subset.
A function to construct feature vector for each session using context-aware analysis.
A function invoking the 'topicWise_feature_mapping_eachSession.m' to construct feature vector for all the sessions.
A function to merge all types of features, including topic presence, audio features, video features, and semantic features.
- the MIC score calculated in mRMR-MIC is returned by the 'mine' function, which is included in the minepy library (version 1.2.4), an open-source library for the Maximal Information-based Nonparametric Exploration.
- The SVM model is implemented by the LIBSVM software package.
Z Dai, H Zhou, Q Ba, Y Zhou, L Wang*, G Li*. Improving Depression Prediction Using a Novel Feature Selection Algorithm Coupled with Context-Aware Analysis. Journal of Affective Disorders. 2021, 295: 1040-1048.
The codes released in this repository are free for academic usage. For other purposes, please contact Zhijun Dai (daizhijun@hunau.edu.cn)