Abstract
BACKGROUND: The mortality rates of prostate and testicular cancer are higher mortality in the northeast region. OBJECTIVE: We aimed to compare the efficacy of machine learning libraries in predicting testicular and prostate cancer mortality. DESIGN AND SETTING: A comparative analysis of the pyMannKendall and Prophet machine-learning algorithms was conducted to develop predictive models using data from DATASUS (TabNet) to Caicó (Brazil) and Rio Grande do Norte (Brazil). METHODS: Data on prostate and testicular cancer mortality in men from 2000 to 2019 were collected. The prediction accuracy of the Prophet algorithm was evaluated using the mean squared error (MSE), the root mean squared error and analyzed using the pyMannKendall, and Prophet libraries. RESULTS: The research data were made publicly available on GitHub. The machine test confirmed the accuracy of the predictions, with the root MSE (RMSE) values closely matching the observed data for Caicó (RMSE = 2.46) and Rio Grande do Norte (RMSE = 22.85). The Prophet algorithm predicted an increase in prostate cancer mortality by 2030 in Caicó and Rio Grande do Norte. This prediction was corroborated by the pyMannKendall analysis, which indicated a 99% probability of a rising mortality trend in Caicó (P < 0.01; tau = 0.586; intercept = 2.59) and Rio Grande do Norte (P = 2.06; tau = 0.84, and intercept = 119.63). For testicular cancer, no significant mortality trend was identified by Prophet or pyMann-Kendall. CONCLUSIONS: Libraries are reliable tools for predicting mortality, providing support for strategic health planning, and implementing preventive measures to ensure men's health. Addressing the gender gap in DATASUS is essential.