Machine learning model robustness evaluation method based on noise data

Abstract

The invention provides a machine learning model robustness evaluation method based on noise data. The method comprises the steps of original data set processing, noise data acquisition, model training, model prediction, accuracy reduction ratio calculation and model robustness evaluation. The original data set processing comprises the steps of collecting an original data set with a correct percentage label, and dividing an original training set and an original test set by adopting 10 times of 10-fold cross validation. The noise data acquisition comprises the following steps: on the basis of anoriginal training set, extracting t'= | D | * alpha data by adopting a stratified sampling method, and replacing a label of the data with an error label, and alpha is a noise data rate. Model training comprises the step of respectively inputting an original training set and a training set mixed with noise data to respectively construct an original model and a new model based on a common classification algorithm. Model prediction includes performing accuracy evaluation on an original model and a new model based on an original test set. Accuracy decline ratio calculation includes calculating arate of decline in accuracy of the new model relative to the original model. Model robustness evaluation comprises the steps of comparing the size of the rate of accuracy reduction in the transverse direction and the longitudinal direction, measuring the robustness of the model, and achieving the standard of judging the robustness of the model.

Description

technical field [0001] The invention belongs to the field of machine learning applications, and in particular relates to model robustness evaluation. We evaluated the robustness of the model by measuring the influence of noisy data on the model. Background technique [0002] Machine learning is an important branch of artificial intelligence research, which learns the representation of features by organizing and fitting parameters. Because of its high generalization ability and efficiency, it is more and more widely used in academia and industry. According to its learning form, it can be divided into: supervised learning, unsupervised learning and semi-supervised learning. Among them, supervised learning is an important aspect of machine learning. It is the process of learning data and its corresponding labels, training an intelligent algorithm, and mapping input data to labels. The data for a supervised learning problem consists of an input X and an output label Y, howeve...

AI Technical Summary

Problems solved by technology

If these labels are wrong, it seriously affects the ability to evaluate the algorithm on the development set, and correcting wrong labels will cost a lot

Images