Improving Prediction Accuracy in Drift Detection using LR in Comparing with Modified Light Gradient Boost Model

The goal of the proposed work is improving prediction accuracy in drift detection using Logistic Regression compared with modified light gradient boost model. The collection of 40 samples were taken by varying test and training data set size. These samples are divided into Two groups (Group 1 - Logistic Regression, Group 2 - Modified Light Gradient Boost) each having 20 samples and the accuracy was calculated to quantify the improving prediction accuracy in drift detection using LR and modified light gradient boost model. The G power is taken as 80%. The results for the simulation is 69% accuracy of LR, and the LGBM provides results with an accuracy of 98%. For the given dataset LR performs significantly less than the MLGBM in the prediction.