Predictive Model Using Machine Leaning Approach for the Detection of Breast Cancer

One of the most common cancer among the women, which is diagnosed and increasing rapidly worldwide, is Breast cancer. Every year the percentage of women diagnoses by this invasive cancer is increasing. It is the major cause of death in women globally. It is critical for a healthy life to predict and diagnose cancer at an early stage. Early detection of breast cancer can considerably improve the prognosis and increase the likelihood of a patient's survival, since it allows for timely clinical treatment. As a result, fast analytics and feature extraction methods are required for high-accuracy cancer prediction, which can be accomplished utilizing Machine learning. In our research work which we present in this paper, we compare various machine learning (ML) algorithms including i) Random Forests ii) Logistic Regression, iii) Decision Tree and iv) Support Vector Machine. We evaluate and analyze the performance of these entire algorithms using area under the receiver operating characteristic (AUROC) curve, and confusion metrics and find the best machine learning model for prediction of breast cancer. The findings are calculated using the evaluation criteria of Precision, Recall, Accuracy, and Specificity. Confusion matrix based on evaluation parameters that put a greater emphasis on predicted cases. A performance evaluation is computed for various machine learning models. For simulation, we used the Wisconsin Dataset of Breast Cancer (WDBC) in our research. After simulation, the SVM model obtained 98.24% accuracy on testing test with an AUC of 0.993, while the logistic regression achieved 94.5% accuracy with an AUC of 0.998. With their mathematical models, these algorithms can be further tweaked to improve breast cancer prediction.