Determinación de los factores de riesgo del cáncer de mama mediante aprendizaje automático y el índice SHAP

Abstract

Breast cancer is a malignant tumor that affects people all over the world, more frequently in the female sex, although it does not exclude the male sex. It is among the five deadliest types of cancer, having a greater influx in less developed countries where access to health programs is poorer. Finding the best machine learning (ML) algorithm for effective breast cancer prediction with the least chance of error. This allows us to set the objective of being able to analyze the different algorithms and select the least error that it has, enriching the algorithms with the Breast Cancer Wisconsin data set. It was concluded that the best algorithm is XGBoost with Shap; We use confusion matrices to see the performance of the algorithms, Roc Curve to know the global performance and its accuracy by dividing the correct predictions by the total predictions.