A Study on Base Station Traffic Classification Based on K-means and Improved KNN Models

In this paper, time series data features are extracted for classification based on HTRN Dataset, and 30,000 data objects are randomly selected as the training test sample set for preliminary cluster analysis. Then, based on the improved KNN algorithm, the remaining samples are categorized into the categories with the closest distance. The original data are preprocessed and randomly sampled to obtain the test samples to divide the dataset; the tsfresh tool is used to extract the statistical features, entropy features and segmentation features of the time series as the corresponding feature vectors for clustering, and the feature vector F is obtained; at the same time, the importance of the individual features of the constituent feature vectors is analyzed based on the Random Forest method. The optimal number of clusters is obtained according to the contour coefficient and the elbow rule, and then the k-means method is used for feature-based cluster analysis; the improved KNN model is applied to classify the non-tested sample plots. The final results show that the above method can effectively improve the processing efficiency and accuracy, and has certain practical feasibility, in order to provide certain reference for research in other fields.