Modelo preditivo temporal integrado a mineração de processos para o suporte das tomadas de decisões em gestão pública municipal

Abstract

This paper proposes a temporal predictive model integrated with Process Mining to enhance decision-making in municipal public administration. Public management faces challenges related to the efficiency and compliance of administrative processes, and Process Mining combined with predictive techniques aims to optimize these processes by providing predictions based on historical data. The general objective is to predict the time required for the completion of administrative processes in a municipal government in the interior of Brazil. The specific objectives include preparing the data through exploratory analysis to identify anomalies, missing values, and inconsistencies, implementing a predictive decision-making model based on Process Mining, and training and testing the chosen model with a training set. The methodology adopted is PM², a structured approach for Process Mining projects, which encompasses six phases: planning, extraction, data processing, mining and analysis, evaluation and improvement, and process support. The data were extracted from municipal government databases and processed to create event logs. Process Mining was performed using Disco software and the PM4Py library in the Python environment. The results show that the initial exploratory analysis revealed significant issues related to delays and inconsistencies in municipal processes. The implemented predictive model demonstrated effectiveness in predicting process execution times, allowing for the identification of bottlenecks and optimization of resource allocation. The accuracy rate of the predictive model was more precise for processes with lower variability in execution times, such as documentation and certificates, while processes with more variability due to their nature, such as human resources and payroll, showed higher variability and error rates. It is concluded that the implementation of the predictive model integrated with Process Mining proved to be an effective tool for improving municipal public administration, providing data-driven decision-making. Continuous application of these techniques can result in more efficient and compliant processes, benefiting the quality of services provided to the population.