An Effective Machine Learning Approach for Clustering Categorical Data with High Dimensions

Many modern real world databases include redundant quantities of categorical data that contribute in data processing and efficient decision-making with their advances in database technology. However, for the reasons that they are identical to measurements the clustering algorithms are only devised for numerical results. An immense amount of work is being performed on the clustering of categorical data using a specifically defined similarity measure over categorical data. Thereby, the dynamic issue with real-world domain, which does not clearly take the predictive form, is the inner function. The function is based on both unseen and transonic perspective. This then offers a detailed and inventive collaboration with categorical results. The paper describes a stratified, immune-based approach with a new similarity metric, in order to reduce distance function, for clustering CAIS categorical data. For successful exploration of clusters over categorical results, CAIS adopts an immunology focused approach. It also selects subsistent nomadic characteristics as a representative entity and organize them into clusters that quantify affinity. To minimize database throughput, CAIS is segmented into several attributes. The analytical findings show that the proposed solution yields greater mining performance on different categorical datasets and outperforms EM on categorical datasets.