Виявлення аномальних цінових пропозицій у системі публічних закупівель з використанням алгоритмів кластеризації

Abstract

Focused on data measured in kilograms, the approach uses NLP (TF-IDF) and K-Means for semantic categorization of lots. Outlier detection is implemented via Principal Component Analysis (PCA) and the DBSCAN algorithm. Results demonstrate that this multidimensional model identifies price deviations more effectively than standard statistical methods. It facilitates automated monitoring of budget expenditures and supports corruption risk mitigation. The integration of semantic and numerical analysis ensures high accuracy in detecting suspicious procurement patterns and enhancing financial transparency.