Improved Accuracy for Identifying At-Risk Students at Different Percentage of Course Length using Logistic Regression Compared with K-Nearest Neighbour Model

Abstract

As comparing to the K-Nearest Neighbor approach, the Logistic Regression algorithm is more successful in locating children who are at risk of dropping out of school sooner than the later algorithm does. This is because the former algorithm takes into account more factors. This is the key goal that we want to accomplish with our research project. In order to carry out this procedure, both the K-Nearest Neighbor approach and the Logistic Regression methodology are used. A dataset with a total of 3467 records was used in the testing and implementation of the approaches. Calculations were carried out to establish which of the two algorithms was more accurate, and the outcomes of these calculations were compared to the findings of the test groups. After being allowed to complete their tasks, the algorithms eventually provide accurate results. According to the data, using Logistic Regression (LR) in conjunction with the K-Nearest Neighbor technique (KNN) resulted in an accuracy of 76.32% on average, with p=0.997. This was determined by using the method of Logistic Regression (p0.05). The findings of the study were evaluated using an independent sample T-test, and the results showed that the findings did not have statistical significance. The K-Nearest Neighbor strategy will be replaced with the Logistic Regression (LR) method, which will be implemented in this work so that it may be employed. The LR method seems to have a higher level of accuracy than the K-Nearest Neighbor approach (KNN).