Abstract: Obtaining information on graduate employability is crucial to every higher education institution. This is because such data would provide insight on the effectiveness of the institution curriculum in preparing human capital for the market needs. To date, the MARA Professional College (KPM) in Malaysia relies on graduates to manually provide data on their employment. Such an approach is not reliable as not all graduates provide the information to the institution. This study presents the application of data mining techniques in forecasting the KPM graduates employment type. In data mining, there exist three main tasks; classification, clustering, and association mining. The aim of this study is to forecast whether a particular graduate will be "employed", "unemployed" or "further study" 6 months after the completion of his study. The undertaken experiments include the utilization of five data mining techniques, namely, the Naive Bayes, Logistic regression, multilayer perceptron, K-nearest neighbor and decision tree. Furthermore, the experimental setup-up is based on three types of data proportion (training-testing) 70-30, 80-20 and 90-10. Based on the obtained result, it is learned that the Logistic regression is the best classifier for the in-hand dataset. In particular, the classifier is at its best when the 80-20 proportion is adopted. The produced classification model will benefit the management of the college as it provides insight to the quality of graduates that they produce and how their curriculum can be improved to cater the needs from the industry.