HYBRID CLUSTERING AND GENERALIZED ASSESSMENT FOR INTELLIGENT ANALYSIS OF MAHALLA ACTIVITY INDICATORS

Abstract

Monitoring mahalla activity requires the joint interpretation of indicators that differ in units, direction, completeness, and administrative significance. This paper specifies a hybrid workflow that combines deterministic initialization of k-means, internal cluster validation, and expert-informed generalized assessment. Equal, rank-based, and point-based weights are used to generate alternative component assessments for sensitivity analysis. The main contribution is a reproducible integration framework with clearly defined tie-handling procedures, checks for degenerate cases, and a selection rule for situations in which the Dunn and Davies–Bouldin indices provide different preferences. Historical cluster counts are considered as secondary evidence, while a transparent numerical example demonstrates the aggregation procedure. The proposed framework is intended for reproducible methodological evaluation, with a validation design provided to assess initialization, indicator selection, weighting sensitivity, and temporal consistency before operational deployment.