Method and apparatus for predicting category of data object

Abstract

The present application relates to methods and apparatus for predicting the category of a data object. The method includes: extracting at least one object feature from the data object to be predicted; according to the object feature, obtaining a feature set from a feature tree constructed in advance based on existing data objects and corresponding data object categories in the database, the feature set includes Object feature pairs that are connected in the object features and single object features that are not related to other object features in the object features; according to the feature set, from the statistics in advance based on the existing data objects in the database and the corresponding data object categories and feature trees In the feature-category probability distribution, the probability distribution of each category corresponding to each object feature pair or object feature in the feature set is obtained; and according to the probability distribution of each category, the prediction category set of the data object to be predicted is determined. According to the solution of the present application, the accuracy rate of category prediction of data objects can be improved.

Description

technical field [0001] The present application relates to the field of data processing, and more specifically, to a method and device for predicting categories of data objects. Background technique [0002] With the continuous development of online data interaction, for some web servers, after obtaining the basic information of the data object, such as title, attribute description, etc., it is often necessary to hang the data object on the background category, so that it can be used as search data later The basis for object category navigation, data statistics of various dimensions, product library construction, etc. Therefore, it is necessary to predict the category of the data object to determine the associated category of the data object. [0003] In a category prediction scheme in the prior art, it is based on the category click dictionary, wherein the category click dictionary is based on the user's historical query words and the category clicks corresponding to the hi...

AI Technical Summary

Problems solved by technology

[0004] However, due to the sparse category clicks of users, it is impossible to cover a large amount of data; the input of some query words is also accompanied by the phenomenon of malicious users swiping query words (some users use certain query words to repeatedly query to improve the information associated with themselves. click-through rate), resulting in inaccurate click data for word categories, seriously affecting the accuracy of the categories predicted by these data

In addition, when predicting categories, an inaccurate category may be predicted due to the repetition of certain words in the title

Images