Comparison of Binary Logistic Regression and Perceptron Neural Network Modelling in Terms of Prediction Using Breast Cancer Data

Abstract

Artificial network modelling are being used in areas of prediction and classification, areas where classical regression models and other related statistical techniques have traditionally been used. This study is aimed at comparing the Logistic Regression Model and the Perceptron Neural Network Modelling to detect the malignancy or benignancy of the tumorous cell of cancer patients more accurately. Data from the Breast Cancer Wisconsin (Diagnostic) was analysed using the Classical Regression Model and the Perceptron model. The study compared both Binary Logistic Model (BLM) and the Perceptron Neural Model (PNM) by their level of accuracy in predicting the breast cancer outcome. The Perceptron model had greater accuracy (98.3%) than the Binary Logistic Model (97.2%) This goes to show that the Perceptron Neural Network Model was a better predictive model as it had a higher accuracy in predicting the cancer model. The study also found that both the perceptron neural networks and logistic regression models can remarkably predict cancer very close to the actual values but the performance of the perceptron neural network model for prediction of cancer was higher and more precise. The study recommends that Perceptron neural network model as a better alternative to the Logistic Regression Model. Keywords: Logistic Regression, Perceptron, Classical, Neural Network