Applying Profit-driven Metrics in Predictive Models: A Case Study of the Optimization of Public Funds in Peru

Fund allocation is a crucial concern in public management, as it is an important factor for economic performance in public investment. Governments spend substantial resources to improve these investments’ efficiency and effectiveness. The use of Machine Learning techniques has proven to be a relevant tool in the decision-making process. In this study we test an approach based on a profit-driven perspective, in order to assess predictive models in public resource allocation, valuing the net profit obtained by the allocation of funds for Peruvian researchers to have a specific performance measure to the fund allocation. A series of experiments were developed using data from 24 Peruvian universities. The use of a profit-driven metric allows to make better choices regarding predictive models and reaching better performance in public investment for Peruvian government. Use of Machine Learning techniques supports the correct identification and selection of researchers to optimally allocate limited resources in an emerging country and shows a novel use of predictive models in public management.