A Hybrid Approach using the Fuzzy Logic System and the Modified Genetic Algorithm for Prediction of Skin Cancer

In recent years the death rate from skin cancer (SC) disease tends to grow enormously. Various studies demonstrated that skin malignancy may rank third as a reason for death for each age gathering, after breast and lung cancer. It becomes necessary to diagnose this skin malignancy at an early stage. The objective of this study was to combine machine learning and soft computing techniques to achieve higher accuracy in the prediction of SC. To play out the exploration work we utilized two datasets, one from "111 Save Life hospital", Jamshedpur, India, and the other is the UCI repository skin cancer dataset. In this paper, a hybrid technique was utilized that combined the advantages of the fuzzy logic system (FLS) and the genetic algorithm (GA). Classifiers such as support vector machine (SVM) and Naive Bayes (NB) were implemented. The modified genetic algorithm (Modified_GA) is used to select the best features which will participate in the fuzzy rules generation process. The modified GA selects the best features along with the calculation of the accuracy of the system based on the selected features. A new rule reduction algorithm (RR_algorithm) is then utilized to reduce the certain number of rules to decrease the complexity of the rule base of the fuzzy system. For the SCC_dataset, the Modified_GA algorithm selects four features with an increased accuracy value of 89.6552%. The RR_algorithm reduces 20 rules from the rule base of the FLS with a constant accuracy value of 98% compared with the FLS with a larger number of rules. For the UCI_dataset, the Modified_GA algorithm selects 19 features with an increased accuracy value of 97.3684%. The selected features were further reduced with the help of sequentialfs() function. Now, the total selected features were 15 with an increased accuracy value of 97.3684% (obtained by the SVM classifier) and 98.6842% (obtained by the NB classifier). Experimental results on the two datasets show that the proposed strategies strikingly improves the forecast precision of skin malignancy.