Cancer Classification in Imbalance Data Using Double Filtering Approach

Cancer, also known as malignant neoplasm, is a complex and potentially fatal disease characterized by uncontrolled and abnormal cell growth in the body. The main problems with microarray cancer studies are the high curse of dimensionality and small sample size caused by redundant and irrelevant genes. To deal with the number of features that exceed the amount of data, this research purposed double filtering method Lasso-GA, Lasso is used to select the features based on feature correlation while Genetic Algorithm is used to optimize the most important features with accuracy traditional machine learning as its fitness function. The results show how effective the suggested method is; in breast cancer, the linear SVC model achieves excellent accuracy (0.93), precision (0.94), recall (0.94), and F1 score (0.94), while in lung cancer, the linear SVC, random forest, and logistic regression models perform well (accuracy: 0.95, precision: 0.92, recall: 1, F1 score: 0.95). Logistic regression is the most effective method for bladder cancer, with an accuracy of 0.82, precision of 0.77, recall of 1, and F1 score of 0.87.