Utilizing Machine Learning in Medical Diagnosis: Systematic Review and Empirical Analysis.

During the last few years, several real-life applications have attempted to utilize the proven high capabilities of artificial intelligence in general and machine learning in particular. Machine learning has been utilized in several domains, such as spam detection, image recognition, recommendation systems, self-driving cars, and medical diagnosis. This paper aims to survey the most related work of utilizing machine learning in the domain of medical diagnosis. Moreover, the paper proposes a comparative analysis for identifying and determining the best classification model and feature selection method in mind of handling medical datasets. Hence, four different medical datasets have been used to train twenty-three classification models and four well-known feature selection methods with respect to several evaluation metrics such as Accuracy, True Positive ratio, False Positive ratio, Precision, and Recall. The results reveal that RandomForest, J48, and SMO classifiers are the best classifiers when it comes to handling medical datasets respectively. Furthermore, the Gain Ratio method is the best choice for handling the step of feature selection.