Investigating Machine Learning Techniques Used for the Detection of Class Noise in Data: A Systematic Literature Review

Data provides valuable information and insights and assists in making strategic decisions. The quality of the data is distorted by noise, which negatively affects information, insights, and decisions made. It also influences a machine learning model’s complexity, accuracy, results, and ability to learn. Therefore, the correct model should be married with the appropriate dataset whilst taking cognisance of data noise. The study's objective is to identify machine learning techniques for class noise detection by considering various strengths and weaknesses to aid in selecting the correct model. A systematic literature review was used as a research approach and resulted in a taxonomy for noise treatment to classify machine learning class noise detection techniques whilst considering strengths and weaknesses of the techniques. Further research should focus on machine learning techniques to make selecting the appropriate data noise detection or managing approach easier.