Dany Pradana, Endang Sugiharti

Implementation Data Mining with Naive Bayes Classifier Method and Laplace Smoothing to Predict Students Learning Results

Introduction

Implementation data mining with naive bayes classifier method and laplace smoothing to predict students learning results. Predict student learning outcomes using data mining, Naive Bayes Classifier, and Laplace Smoothing. Achieves high accuracy (up to 94.937%) with feature selection, enhancing educational insights.

Abstract

Abstract. The application of information technology in the field of education produces big data. It retains information that can be treated as useful. Having data mining, can be used to model highly useful student performance for educators performing corrective actions against weak students.  Purpose: The study was to identify the application and accuracy algorithm Naive Bayes Classifier to predict students' study results. Methods: The prediction system for student learning outcomes was built using the Naive Bayes Classifier and Laplace Smoothing methods using a combination of two Information Gain and Chi Square feature selections. The experiment was carried out 2 times using different dataset comparisons. Result: In the first experiment using a dataset of 80:20, the accuracy Naive Bayes Classifier method with Laplace Smoothing and without Laplace Smoothing showed the same results as 94.937%. On the second experiment to equate dataset 60:40 results of the Naive Bayes Classifier accurate method without Laplace Smoothing only 86.076%, then score a 91.772% accuracy using the Laplace Smoothing. The improvement is caused by a probability of zero that can be worked out with Laplace Smoothing. Novelty: The selection feature process is very important in the classification process. Thus, in this study, information gain and chi square double selections of such features as information gain and so promote accuracy.

Review

This paper addresses a highly relevant topic within educational data mining: the application of predictive analytics to student learning outcomes. The ability to accurately forecast student performance is of significant value, enabling educators to identify at-risk students proactively and implement timely corrective actions. The study’s primary objective is to evaluate the effectiveness and accuracy of the Naive Bayes Classifier, particularly when enhanced with Laplace Smoothing, for this crucial prediction task. This focus aligns well with the growing trend of leveraging information technology-generated big data in educational settings. Methodologically, the authors developed a prediction system using the Naive Bayes Classifier, incorporating Laplace Smoothing to address potential zero-probability issues. A key aspect of their approach is the application of a combined feature selection strategy, utilizing both Information Gain and Chi-Square methods, which they suggest contributes to improved accuracy. The experimental evaluation was conducted in two phases with differing dataset splits (80:20 and 60:40). The results illustrate that while the 80:20 split achieved a high accuracy of 94.937% consistently, the 60:40 split showcased a notable improvement with Laplace Smoothing, boosting accuracy from 86.076% to 91.772%. This finding clearly demonstrates the practical benefit of Laplace Smoothing in scenarios where data distribution might otherwise lead to reduced model performance. The novelty claimed by the authors resides in the "double selection" feature engineering process, combining Information Gain and Chi-Square to enhance classification accuracy. While the abstract effectively highlights the beneficial impact of Laplace Smoothing in certain experimental conditions, a more detailed explanation of *how* these two feature selection methods were integrated or combined, beyond merely stating their use, would further clarify this contribution. Nevertheless, the paper presents a well-defined problem, a clear methodological approach for evaluating the Naive Bayes Classifier with Laplace Smoothing, and demonstrates promising results that underscore the utility of these data mining techniques in predicting student learning outcomes.

Full Text

Comments

You need to be logged in to post a comment.