Analyze the Factors that Contribute to Accurate performance Predictions, Including Data Preprocessing, Feature Selection, and Model Hyperparameters

Abstract

A solid academic record boosts a university's standing and promotes student career chances, hence predicting academic performance has attracted attention in education. Using clusters obtained by Davies' Bouldin approach, a clustering data mining technique known as K-means is used in this study to identify critical characteristics impacting students' performance. Machine learning techniques find use in many fields, including medical diagnostics, image processing, cluster analysis, pattern identification, and natural language processing. Among the algorithms tested, SVM produced the most accurate predictions (96% accuracy rate) after parameter tweaking. The researchers in this study have looked at how the SVM, Decision Tree, naive Bayes, and KNN classifiers work. The results of adjusting the parameters significantly improved the four prediction models' accuracy. Feature selection algorithms and hyperparameter optimisation, two critical components for enhancing model performance, are also addressed. The findings highlight the need of carefully evaluating models, with Random Forest emerging as a dependable choice for accurate diabetes prediction.