Academic achievement of a student in college always has a far-reaching impact on his further development. With the rise of the ubiquitous sensing technology, students' digital footprints in campus can be collected to gain insights into their daily behaviours and predict their academic achievements. In this paper, we propose a framework named AAPEDM (Academic Achievement Prediction via Educational Data Mining) to predict students' academic achievements based on the influencing factors we have discovered. Multisource heterogeneous data including Wi-Fi detection records, usage of smartcards, usage of campus network, is aggregated firstly. Then, instead of the self-reported features or traditional academic assessments like test scores, we extract features reflecting students' behavioural patterns. Specially, we define DOH (Degree of Hardworking) to improve the performance of the classifier. Finally, we analyze the features extracted and apply supervised learning methods to predict their academic achievements. Experiments are conducted on real-world data from 528 college students in one faculty, and the classification accuracy can be up to 88%. [For the full proceedings, see ED593090.]