Prediction of Breast Cancer Risk Factors Using Neural Network Analytics: an Empirical Study

Abstract

Breast cancer is the most prevalent cancer affecting women. Identifying breast cancer risk factors are crucially to be established. The main objectives of the present study were to identify the predictors of breast cancer risk factors and their relative importance using neural network analysis. The present study depended on neural network analysis of data posted on [1]. The dataset is about predictors of breast cancer. There were 9 covariates included and one dependent variable, output (no disease (1), or disease (2)). The dataset was composed of 116 cases. The category of no disease comprised 79 (68.1%) cases, whereas the disease category included 37 cases (31.9%). Architecture model was built with some characteristics such as training part: gross entropy error was 23.7884, the percent of incorrect predictions was 10.1%. Stopping rule used was 1 consecutive step(s) with no decrease in error. For testing part, gross entropy error was 14,327, the percent of incorrect predictions was 13.5%. The relative importance of breast cancer was in the following order: glucose, resistin, BMI, age, leptin, adiponectin, MCP-1, insulin, and HOMA. Taken together, neural network analysis is an efficient tool to predict breast cancer risk factors.