Feature selection method

A feature selection method for constructing a classification model for predicting a disease state based on omics data, wherein an electronic device is used to perform following steps, comprising:

• inputting a plurality of omics training data into a plurality of training models respectively to obtain a plurality of feature pools corresponding to the plurality of training models respectively, wherein obtaining the feature pools comprises:
• inputting the omics training data into each of the training models to obtain a plurality of omics features;
• classifying the omics features into a plurality of feature groups based on a plurality of feature types, so that each of the training models performs a feature extraction operation on the omics features included in each of the feature groups to obtain, for each of the training models, a plurality of first feature pools respectively corresponding to the feature groups, wherein the feature pools comprises the first feature pools;
• setting the feature pools corresponding to each of the feature groups as a feature set;
• sorting the omics features in each feature set based on a number of times that each of the omics features is selected by the feature pools to obtain a feature ranking of each feature set;
• extracting a corresponding number of a plurality of designated features from each feature set based on a weight corresponding to each of the feature groups and the feature ranking, wherein the weight corresponding to each of the feature groups is based on a ratio of the number of omics features included in each of the feature groups to the number of all omics features;
• and
• inputting the plurality of designated features into the classification model and using a learning algorithm to continuously adjust internal parameters of the classification model through iteration, wherein the classification model is a neural network.