Automatic learning model to predict transparency indicators for effective management of public resources
Introduction: This article is the product of the application of predictive analytical models to measures and indicators of corruption risk, as researched by the Pascual Bravo University Institution and the Francisco José de Caldas District University in 2022 for doctoral research on the risk model for state transparency.
Problem: From measurements of institutional capacities, it is possible to generate anticorruption measurements, such is the case of the National AntiCorruption Index (INAC for its Spanish acronym). However, there are improvements to be made in the indicators and the need to incorporate more and better measurements that support this scourge that has long been manifested in Colombia.
Objective: The objective of this research is to emphasize the need to take advantage of open data, to generate measurements of state institutional corruption and, therefore, metrics that support its transparency and integrity based on predictive analytical models to generate predictions about government indices.
Methodology: First, the importance of generating measurements for the management of corruption cases is pointed out. Then, the application of predictive analytical models to predict scores of the National AntiCorruption Index is evidenced, finding the best model to finally make a forecast based on the identification of the relevant variables.
Results: The implementation of higher levels of digital government (egovernment) can significantly contribute to the fight against corruption and the generation of better public policies that support controls and sanctions. It not only facilitates citizen access to state services, but also allows for more open and agile access to data. This constantly promotes transparency at all levels and at all times. The Huber regression that has been implemented, its smaller penalty function, and its linear rather than quadratic growth, make it more suitable for dealing with outliers. This improves the error meter estimates and provides a good estimate of the National AntiCorruption Index score.
Conclusion: It is essential to establish a framework that anticipates the behavior of INAC and directs public policy efforts towards transparency and the prevention of corruption. In addition, it is necessary to develop objective metrics, indicators, indices and risk models that promote and evaluate transparency in the fight against corruption. This implies generating early warnings, applying sanctions, implementing controls and designing improvement plans to promote recommendations based on data that can trigger actions and take advantage of free access to public information to support citizens and the country.
Originality: A predictive analytical model based on machine learning was trained to predict the future behavior of the National AntiCorruption Index, with the aim of supporting roadmaps for entities and creating improvement actions for national entities, in which it becomes necessary to explore open government data to create new indicators and improve current ones.
Limitations: The regression models on the historical data of free access for the INACs were selected, because in terms of measurement it is what is already consolidated and available for the generation of transparency policies, access to information and the fight against corruption. The challenge for future work is to have more historical data, and to create more indicators that support measurements with the creation of improvement actions per entity that is reflected in numerical measurements.