Framework for extraction of crime insights from diversified data sources using a multiple minimum support FP-Growth Algorithm

Abstract

Introduction: Mining of crime insights from datasets can be employed to help in discovering useful crime insights for improved crime detection and prevention strategies. However, most of existing frequent pattern mining approaches lack flexibility in terms of user-defined minimum supports and diversified sources of input data. Objectives: This study was carried out to propose a framework for extaction of crime insights from diversified sources of data using the FP-Growth algorithm with multiple minimum supports. Methods: The study was based on systematic review of literature to establish the study gap followed by a comprehensive experimentation for the validation of the proposed framework. With regard to the systematic review, a PRISMA framework was employed. To accomplish experimentation of the suggested framework, two different sources of data were used; crimes database which provided structured data, and news articles which provided unstructured data. Results: The study came up with a generic and flexible framework that extracts crime insights from diversified data sources composing of four stages as follows: data sources, pre-processing, processing, and pattern visualization.. On top of being effective in the extraction of patterns, this approach yields a better runtime and memory use than classical FP-growth Conclusions: Multiple sources of crime data should be considered for effective extraction of crime insights. For it to be effective, frequent pattern mining approach must consider using multiple minimum supports.