GDD-K-Means Text Clustering Algorithm Based on Grid Filtering Distance and Density of Outliers

In the era of big data, fully mining and utilizing the value of big data in line with the requirements of big data strategy plays a significant role in social development. Clustering algorithm can effectively partition unlabeled data sets through unsupervised learning process, and traditional K-Means algorithm is still the most widely used algorithm at present. By studying and learning various improved algorithms of traditional K-Means clustering algorithm, this paper has optimized the problems such as unsatisfactory clustering results caused by outliers and disadvantages of initial center point affecting initial partitioning. Good results have been obtained. Firstly, the grid filtering and LOF detection method of weighing distance and density are used to remove outliers. Then, the randomness of initial center selection is better eliminated by combining the "max-min principle" with the strategy of maximum weight, and the number of clusters is determined according to the BWP index. Experimental results have shown that compared with the currently popular clustering algorithms, the proposed GDD-K-Means clustering algorithm has achieved better results in different data sets, and the accuracy and F-number and other evaluation indexes are improved to a certain extent, and the calculation time complexity is effectively reduced.