Advancing breast cancer prediction using blockchain-secured hybrid genetic algorithm

Abstract

Feature selection using evolutionary algorithms-a well-liked technique for choosing pertinent characteristics in huge datasets is explored. In machine learning, feature selection (FS) is a key phase that helps to boost model efficiency, decrease overfitting, and improve model accuracy. Breast cancer (BC) prediction algorithms can be improved to attain more accuracy while protecting patient privacy and maintaining data integrity by utilizing blockchain. The suggested model incorporates the system with the blockchain method for smart contracts, and it evaluates and compares various learning algorithms and classifiers for the identification of breast cancer. Many different machine learning algorithm techniques were tested on the Wisconsin Diagnosis Breast Cancer data set. The accuracy obtained in different classifier such as XGBoost, AdaBoost, Logistic Regression, Linear SVM, Random Forest, KNN, Gradient Boosting, Radial SVM and Decision tree are 99%, 99.06%, 99.35%, 99.47%, 98.88%, 96.84%, 98.71%, 96.31% and 96.43%. The accuracy of each algorithm in determining whether a tumor was benign or malignant was demonstrated to be higher than 96.31%. The combination of GA and Linear SVM achieved the highest accuracy of 99.47%. In general, though, XGBoost outperforms the competition whether you use GA or not. Because of this, supervised machine learning techniques will be extremely useful in cancer research, especially for purposes of early diagnosis and prognosis. To solve the hazards of data breaches, unauthorized access, and tampering are present with traditional centralized systems, we have developed a safe and impenetrable system that protects patient data throughout the prediction process by integrating blockchain with machine learning.