Performance evaluation of classification algorithms after data preprocessing in health data
Project no. 20199 · Status: Kapanmış · 2019
Summary
Classification algorithms are used in medical diagnosis, which is a critical decision-making process. Because the performance of the algorithms can be affected due to problems that may arise before classification, the data should be pre-processed before using the classification algorithms. In this study, possible problems such as missing data, noisy data, and class imbalance that may arise in health data in order to establish the correct disease diagnosis will be eliminated in data pre-processing. Then, it is aimed to apply the machine learning classification algorithms, to compare the classification performance results before data pre-processing with the classification performance after data pre-processing. Thus, the effects of these problems on decision-making processes can be reduced to provide more meaningful and more effective information from the data. Since there is no solution approach in the literature that evaluates all three problems (missing data, noisy data, and class imbalance) and passes to the classification stage, this project is unique in terms of the subject in the field. It is thought that analyzing with machine learning classification algorithms after applying data preprocessing to the possible problems that may be encountered in order to diagnose the correct disease in health data has a healing effect on the success of these algorithms. It is planned to fill the gap in this field by providing solutions to various problems of the data obtained in the medical diagnosis of this study.