A Novel Method for Clustering High-Dimensional Data Using K-Means

The k-means approach is a fundamental tool in unsupervised machine learning (ML) for grouping high-dimension data. Data mining, recognizing patterns, analyzing images, bioinformatics, and ML are just a few of the fields that use clustering, an efficient quantitative data processing method. Data can be divided into categories of items with unique properties by using clustering. Numerous activities, management, and service data utilized by managers for system functional assurance are generated by cellular radio systems. Although there are strong clustering techniques like Self Organizing Mapping (SOM), there is almost no research demonstrating how well these techniques function with OAM databases that have an abundance of features (>20) gathered from real-world system adoption. Furthermore, there are still a few unresolved issues regarding the relevance of the clustering techniques. For example, when employing SOM, the beneath data fails to explicitly provide data on clusters following SOM training; therefore, the k-means method for classifying SOM components must be used afterward. In this regard, an approach for clustering radio units depending on a mix of SOM and K-means techniques is described in this study. The technique is used to identify the short-term conduct (hourly observational intervals) and long-term conduct (15 days observational duration) of wireless units by extracting cell shapes. The investigation makes use of OAM statistics gathered from a real 4G/LTE system installed in a significant European town.