LDA–GA–SVM: improved hepatocellular carcinoma prediction through dimensionality reduction and genetically optimized support vector machine

Hepatocellular carcinoma (HCC) is a common type of liver cancer worldwide. Patients with HCC have rare chances of survival. The chances of survival increase, if the cancer is diagnosed early. Hence, different machine learning-based methods have been developed by researchers for the accurate detection of HCC. However, high dimensionality (curse of dimensionality) and lower prediction accuracy are the problems in the automated detection of HCC. Dimensionality reduction-based methods have shown state-of-the-art performance on many disease detection problems, which motivates the development of machine learning models based on reduced features dimension. This paper proposes a new hybrid intelligent system that hybridizes three algorithms, i.e., linear discriminant analysis (LDA) for dimensionality reduction, support vector machine (SVM) for classification and genetic algorithm (GA) for SVM optimization. Consequently, the three models are hybridized and one black box model, namely LDA–GA–SVM, is constructed. Experimental results on publicly available HCC dataset show improvement in the HCC prediction accuracy. Apart from performance improvement, the proposed method also shows lower complexity from two aspects, i.e., reduced processing time in terms of hyperparameters optimization and training time. The proposed method achieved accuracy of 90.30%, sensitivity of 82.25%, specificity of 96.07% and Matthews Correlation Coefficient (MCC) of 0.804.