Cancer Disease Prediction Using an Integrated Ensemble Technique with ReliefF and FP

Cancer is a significant factor in the current increase in mortality rates. The utilization of machine learning is crucial in this context. An expedited diagnosis can aid in the prevention of Microarray data, which encompasses genetic information, serving as the fundamental prerequisite for machine learning. Microarray data is notable for its high dimensionality, characterized by a larger number of characteristics and a smaller sample size. While feature selection algorithms can minimize the complexity of the dataset, there are still numerous irrelevant qualities present. A hybrid model is formed by integrating the feature selection technique with optimized algorithms to achieve the most effective feature selection. This work utilizes ensemble machine learning approaches, such as Support Vector Machines (SVM), Multi-layer Perceptron classification (MLP), Logistic regression (LR), Decision Tree (DT), AdaBoost classifier (AdaBoost), XGBoost, LightGBM, and CatBoost, as classifiers. ReliefF is employed for feature selection, whereas Flower Pollination (FP) is utilized for optimization. The ensemble model is developed by considering the precision performance of the individual classifiers, including SVM, MLP, LR, DT, AdaBoost, XGBoost, LightGBM, and CatBoost. The results of each individual base learner were combined using bagging and boosting ensemble methods to determine the optimal model. An investigation was conducted on the outcomes of bagging and boosting ensemble methodologies. Work performance is currently assessed using many measures, such as accuracy, specificity, sensitivity, precision, and f1-score. Upon analyzing the results, we concluded that boosting is the most effective method for achieving superior outcomes compared to bagging ensemble techniques.