Prediction of the success of the selected wart treatmentmethod using various machine learning algorithms

Warts,which are virus-based,are the most common dermatoses in society. Several treatment methods have been developed in the wart treatment. Recently, treatment methods of cryotherapy and immunotherapy have been got off the ground to the patients with common and plantar warts. On the other hand, there is no proof of which treatment method will be successful, yet. In this study, it was predicted if these two methods to be applied in the treatment of warts will success or not before staring the treatment using commonly-used machine learning algorithms. Algorithms were run by applying the two online and freely available UCI data sets to the inputs. These data were combined from two data sets from 180 patients with common and / or plantar warts. The cryotherapy was applied to the half of patients and the immunotherapy was applied to the other half. As a result, the success of the selected wart treatment method was predicted with the criteria of sensitivity, specificity, and accuracy. These values were obtained for the corresponding algorithms used, respectively: 68.43, 67.61 and 67.78for Naive Bayes, 26.32%, 94.37%and80.00% for Logistic Regression, 52.63%, 94.37%and85.56% for Decision Tree, 15.79%, 97.19% and 80.00% for 7-Nearest Neighbors, 47.37%, 95.78% and 85.46% for Support Vector Machines, 36.85%, 90.15%and 78.89% for Extreme Learning Machines, and 78.95%, 98.60% and 94.45% for Multi-Layer Perceptron. Multi-Layer Perceptron, which is resulted in the highest general accuracy among these methods, can predict whether the selected wart treatment will succeed or not more accurately than the few studies presented in the literature. Keywords: Wart, cryotherapy, immunotherapy, machine learning, naive Bayes, decision tree, logistic regression, k-nearest neighbors, support vector machines, extreme learning machines, multi-layer perceptron.