An autoML algorithm to select linear regression model and its performance evaluation for linear datasets
Nagesha Shivappa · 2024
This paper presents an autoML algorithm to select linear regression model and its performance evaluation for any linear dataset. It computes and compares the performance of various multiple linear regression (MLR) models obtained by different training procedures. These models were found using training-testing data split, cross-validation and stratified cross-validation procedures and their performance evaluation is used for comparison. The selected multiple linear regression model’s performance is also compared with simple regression model to select the final model for deployment. The performance is measured using correlation coefficient, mean absolute error, root mean squared error, relative absolute error and root relative squared error. The algorithm is tested and evaluated using CPU performance dataset of UCI machine learning repository. Finally, the algorithm performs the testing of assumptions made by the linear regression for model development using the performance results and CPU performance dataset which proved that its relationship of target and features is linear. The algorithm selects the multiple linear regression model and performance produced by the stratified cross-validation training procedure for the given dataset.