Prediction of Breast Cancer using Deep Learning Algorithms and Gradient Boosting

Abstract

One of the main causes of death for women, particularly in poorer nations, is breast cancer. In order to lower death rates, timely diagnosis, detection, prediction, and effective treatment are now essential. Models for predicting and diagnosing breast cancer become more reliable and accurate as artificial intelligence, machine learning, and deep learning approaches are used more frequently. Examining the efficacy of various machine learning and contemporary deep learning models for breast cancer diagnosis and prediction is the goal of this study. This study contrasts cutting-edge approaches that make use of deep learning models with conventional machine learning classification methods. Deep learning models like Neural Decision Forest and Multilayer Perceptron were employed, along with well-known classification models like k-Nearest Neighbors (kNN), Gradient Boosting, Support Vector Machine (SVM), Neural Network, CN2 rule inducer, Naive Bayes, Stochastic Gradient Descent (SGD), and Tree. The Orange and Python tools were used to conduct the experiment, which assesses their diagnostic efficacy in detecting breast cancer. Transparency and accessibility in the study strategy are made possible by the evaluation's usage of UCI's publicly available Wisconsin Diagnostic Data Set. Result: In both malignant and benign instances, the mean radius ranges from 6.981 to 28.110, and the mean texture ranges from 9.71 to 39.28. SVM has the lowest accuracy and sensitivity at 88%, whereas gradient boosting and CN2 rule inducer classifiers do better. With an AUC value of 0.98%, the CN2 rule inducer classifier obtains the highest ROC curve score for both benign and malignant breast cancer datasets. With a higher AUC-ROC of 0.9959, accuracy of 96.49%, precision of 96.57%, recall of 96.49%, and F1-Score of 96.50%, MLP displays can differentiate between positive and negative classes. GB and the CN2 rule outperformed the other models among the most popular classifier models. Deep learning's MLP, however, yielded the best overall results.