A Comparative Study of Regression Machine Learning Algorithms: Tradeoff Between Accuracy and Computational Complexity

Abstract

The computational complexity of Machine Learning is a mathematical study of the possibilities for efficient learning by computers which is the determination of looking for the best methods to solve a problem. The accuracy of a regression model's predictions must be reported as an error. According to the researchers, the most problematic issue is the lack of a properly defined machine learning assessment. In this research, Various types of machine learning regression algorithms, namely, Linear Regression, Support Vector Regression, Random Forest Regression, and Multilayer Perceptron Neural Network have been used to process and analyze the collected data in terms of comparison of their accuracy and the computational complexity. The applied dataset was collected using IoT sensors seeking an appropriate algorithm that is the fittest to the collected data to design a model system that represents the goal of specific future applications. The result shows that the Random Forest regression has the highest computational complexity and highest accuracy depending on the calculated error metrics (Mean Square Error, Mean Absolute Error, and R Squared score) which are (0.0002, 0.005, and 0.995) respectively. Based on that, Random Forest Regression will be adapted and implemented with the structure of a planned design system.