Internet technology is vital in education as it is an enabler tool used by students for information exchange, communication and creation of knowledge. However, issues related to Internet use behaviour (IUB) and effects on students' performance are still being debated due to inconsistent results. Thus, the main aim of this paper is to propose a specific IUB classification model to discover comprehensive prediction models on students' performance by using educational data mining (EDM). Data related to the students' CGPA and their internet usage activities were collected from 469 undergraduate students. Primarily, in this study several techniques were ensembled to predict students' performance based on IUB. An EDM approach comprising of clustering, classification, correlation, and regression was used. The base classifiers including decision tree (j48) ensembled with EM clustering technique were exploited to develop a high accuracy of prediction results. At first, the usage of clustering technique was to cluster the dataset into smaller variable and applied classification technique to solve the complexity of the variables. Correlation come across to measure the degree of IUB and students' performance and later regression was applied to produce predictive results. Based on 11 IUB categories, the results indicate that online gaming has a negative significant effect on the students' performance. Additionally, the prediction model revealed a high correlation, which generates good predictions of 92%. The prediction model would enable higher learning institutions to detect as early as possible those students who are at risk. This will then help them in taking timely and proactive measures to improve their performance.