PREDICTION OF CANCER INCIDENCE BY GENDER IN BELO HORIZONTE: INNOVATION IN THE USE OF ARTIFICIAL NEURAL NETWORKS FOR PUBLIC HEALTH PLANNING

Abstract

The study analyzed the incidence of cancer in Belo Horizonte from 2000 to 2020 with data provided by INCA, using the Multilayer Perceptron Neural Network (RNA_MLP) technique to predict cancer cases for the years 2021 to 2023. The city has a mostly female population (53.5%), which directly reflects the higher incidence of cancer among women, especially breast cancer. The analysis of the correlation between the population and the incidence of cancer revealed strong associations, with a correlation of 0.93 for male cases and 0.95 for female cases, in addition to 0.98 between cases in both genders, evidencing an almost synchronized growth pattern in cancer cases for both sexes. The performance of the RNA_MLP was evaluated based on quadratic errors, presenting a sum of squared errors of 1.901 in the training and 0.299 in the tests, indicating a good fit of the model. Relative errors were also lower in the tests, with 7.8% for the general model and 7.8% for both sexes. Forecasts indicated a continuous increase in cancer incidence between 2021 and 2023, with an upward trend for both males and females, reflecting not only the population increase, but also a possible improvement in the detection of the disease. Despite being a prediction based on estimates, the study highlighted the importance of the model to assist in the planning of public health policies and prevention strategies, considering the impact of cancer on the city's public health.