Model training data construction method and electronic equipment

The invention provides a model training data construction method, which comprises the following steps: determining a mutation sample and a non-mutation sample according to time sequence data; according to mutation interval detection features, determining features of the mutation sample and features of the non-mutation sample; and performing hybrid sampling clustering on the features of the mutation samples and the features of the non-mutation samples to obtain model training data. According to the scheme, more relevant features are designed according to the characteristic that data on the left side and the right side of a mutation interval have large changes in the data range, a hybrid sampling clustering optimization algorithm is provided, and the problem of sample overlapping caused by a traditional oversampling algorithm is solved. The constructed model training data can solve the problems of insufficient minority class samples and redundancy of majority class samples at the same time, and the classificati