A kind of automatic more threshold values characteristic filter method and devices

摘要

A kind of automatic more threshold values characteristic filter method and devices, the characteristic filter method is in the iterative process of machine learning model training, the characteristic filter threshold value and feature correlation (S1) of epicycle iteration are calculated according to the result of last round of iteration, and according to the characteristic filter threshold value and feature correlation being calculated, characteristic filter (S2) is carried out to sample. This feature filter device includes computing module and characteristic filter module. This method and device are in each round iteration, different characteristic filter threshold values can be automatically generated, the precision for greatly improving filtering threshold values can greatly improve the accuracy of the training speed that machine learns automatically and machine learning model compared with existing fixed single threshold values.