An Enhanced Spectral Clustering Algorithm with S-Distance

Calculating and monitoring customer churn metrics is important for companies to retain customers earn more profit in business. In this study, a prediction framework developed by modified spectral clustering (SC). However, the similarity measure plays an imperative role predicting with better accuracy analyzing industrial data. The linear Euclidean distance traditional SC replaced non-linear S-distance (Sd). Sd deduced from concept of S-divergence (SD). Several characteristics are discussed work. Assays conducted endorse proposed algorithm on four synthetics, eight UCI, two databases one telecommunications database related churn. Three existing algorithms—k-means, density-based spatial applications noise conventional SC—are also implemented above-mentioned 15 databases. empirical outcomes show that beats three algorithms terms its Jaccard index, f-score, recall, precision accuracy. Finally, we test significance results Wilcoxon’s signed-rank test, rank-sum sign tests. relative study shows interesting, especially case clusters arbitrary shape.