Recommendation result optimization method and apparatus for recommendation model, and device and medium

Abstract

The present application is applicable to the technical field of model optimization, and particularly relates to a recommendation result optimization method and apparatus for a recommendation model, and a device and a medium. The method comprises: acquiring a classification boundary value used for classifying a prediction result generated by a recommendation model, and on the basis of the classification boundary value, determining a fuzzy interval that represents a classification error; performing screening on a training data set used during the training of the recommendation model, so as to obtain a target data set, the prediction result of which falls within the fuzzy interval, and performing prediction rule extraction on the target data set, so as to obtain at least one prediction rule; acquiring a prediction result of data to be recommended which corresponds to the recommendation model; and if it is detected that the prediction result of said data is within the fuzzy interval, performing recommendation prediction on said data on the basis of the at least one prediction rule, so as to obtain an optimized recommendation result corresponding to said data. Therefore, a prediction rule is extracted by means of analyzing corresponding data in a training data set, and the prediction rule is used to perform recommendation prediction on uncertain data, thereby improving the accuracy of a recommendation model.

Images