Designing a Support System for Predicting the Survival of Patients with Melanoma Based on Data Mining Algorithms

Background: Melanoma is one of the most commonly diagnosed cancers and the second cause of cancer-related death among people. This disease is the rarest and most malignant type of skin cancer. In advanced conditions, it has the ability to spread to internal organs and can lead to death. In Iran, for several years, significant data about melanoma have been collected either manually or electronically, due to its prevalence and the high costs it leaves on the country's healthcare system, but despite these valuable data, the health system is still unaware of the high potential of data mining in predicting the survival of patients with melanoma. Therefore, the aim of this study was to design an intelligent system to predict the survival of these patients. Methods: This study was practical in terms of nature and descriptive-analytical and retrospective in terms of method. The research population consisted of patients with melanoma cancer from the database of the National Cancer Research Center affiliated to Shahid Beheshti University, located in Tajrish Martyrs Hospital, Tehran, Iran (between 2007 and 2012), who were followed up for 5 years (n = 4118). SPSS and Weka software were used to design the support system for melanoma cancer survival prediction. The final model for predicting melanoma cancer survival was selected based on the evaluation indices of data mining algorithms. Findings: Neural network algorithms, simple Bayes, Bayesian network (BN) and combination of decision tree with simple Bayes, logistic regression, J48, and ID3 were selected as the used models of the country's database. Based on the findings, the neural network performed better with a value of 0.97 in terms of accuracy and 91.03 in terms of features. Conclusion: The performance of the neural network in all evaluation indices was statistically higher than other selected algorithms. Therefore, this algorithm was chosen as a support system for predicting melanoma cancer survival.