Method and system for removing system recommendation deviation

Abstract

The invention discloses a method and a system for removing system recommendation deviation. The method and the system can solve the influence of deviation on a recommendation system under various conditions by using a universal deviation removing framework. According to the technical scheme, the method comprises the following steps: receiving input data related to recommendation deviation in a recommendation system scene as one or more feature groups of a data feature layer; performing vector processing on the received data according to categories of feature groups, converting feature values of the data into vector values so as to respectively convert the feature groups into corresponding vector groups, transmitting one or more vector groups to a multi-layer neural network, performing cross combination on word vectors in the vector groups in the multi-layer neural network to form a data set, and constructing a shared data layer by using the data sets; and dividing the data in the shared data layer into a plurality of different data groups, carrying out depolarization processing on the plurality of data groups by adopting corresponding depolarization strategies, and integrating and outputting the depolarized data of all the data groups.

Description

technical field [0001] The invention relates to a technology for removing system recommendation deviations, in particular to a method and system for removing system recommendation deviations by using a multi-objective model. Background technique [0002] As an important means to solve information overload and provide users with personalized content, the recommendation system has achieved remarkable results in various application scenarios. Traditional recommendation systems collect user behavior data as raw data, combine machine learning and other technologies, and finally return recommended content to users. However, in the process of collecting user behavior data, there are often various deviations that lead to a decline in the recommendation effect. [0003] Common deviations mainly include the following concentration situations: [0004] Selection bias: users are more inclined to rate items they like or dislike, resulting in missing rating data for items between the tw...

AI Technical Summary

Problems solved by technology

[0006] Exposure bias: Due to the influence of popular items or previous recommendation results, users are more likely to receive certain types of items, while other items are not exposed accordingly, resulting in items that users are interested in not being displayed to users, resulting in bias

[0007] Existing techniques for removing common deviations are often aimed at one or two specific deviations, but in real application scenarios, multiple deviations are often mixed together and cross-affected.

Images