Multi-dimensional text clustering method based on metric learning

Abstract

The invention discloses a multi-dimensional text clustering method based on metric learning. The multi-dimensional text clustering method mainly comprises the following steps: 1, selecting two dimensions from a data set and carrying out feature vectorization representation; 2, using a K-Means clustering method and combining metric matrix learning to respectively carry out initialization clusteringon the two dimensions; 3, judging whether the current clustering result reaches an end condition, if not, setting constraints to execute the step 4 on the upper limit constant, otherwise, ending thealgorithm, and outputting the clustering result to assist the downstream task; 4, selecting constraint pairs meeting conditions by utilizing the clustering result of the current dimension; 5, adding the constraint set generated in the step 4 into a dimension clustering process, and adjusting learning of a target function and a metric matrix to obtain clustering results of two dimensions; and 6, repeating the steps 3-5. According to the method, expression modes of data in different feature spaces are comprehensively considered based on a multi-dimensional text clustering algorithm of metric learning, multi-dimensional auxiliary clustering is carried out, and a good clustering effect is achieved.

Description

technical field [0001] The invention relates to a text clustering method, in particular to a multi-dimensional text clustering method based on metric learning, and belongs to the technical fields of machine learning and natural language processing. Background technique [0002] Multidimensional data is very common in practical applications in the era of big data. For example, the description of a webpage can be not only the words involved in itself, but also the related links of the webpage; in text-related tasks, text features, semantic information, and even user behaviors such as liking, reposting, and commenting on the text can be used. Also describe the text. Correspondingly, multidimensional clustering, as a basic task in machine learning, pattern recognition and data mining, has become an important extension of clustering. Traditional text clustering generally extracts features from text content, and then directly combines them with other dimension attribute features...

AI Technical Summary

Problems solved by technology

Such algorithms not only ignore the heterogeneity between different feature spaces, but also have no mechanism to explore the potential relationship between clustering results and multiple dimensions

Images