Predictive behavior modeling is the use of mathematical and statistical techniques and/or data mining to predict future events or human behavior [1]. In [2], a set of new predictive modeling techniques is presented. These techniques can be used independently or in combination with traditional modeling techniques to predict medical outcomes, particularly in case of prostate cancer treatment. Beyond the prediction of medical outcomes, these techniques could also be considered in modeling and predicting human behavior. In [3], for example, the authors use dynamic Markov models, a mathematical method, to recognize human behavior.

More interesting are the works of Ajzen and Fishbein [4-7] on human behavior prediction, using for their prediction functions different parameters such as human attitude, perceived nom, and self-efficacy. The perceived nom, which is social pressure, has two aspects--injunctive and descriptive nom--both related to how the social network importantly impacts human behavior.

Nowadays, the Internet is an important platform where social activities are increasingly taking place and social communities are being built. The Internet is thus a source of information about individuals’ behaviors, health information, and so on. The Internet has therefore become an appropriate source to track and analyze human behavior in online communities. The authors of this paper seem to be pioneers in using ontology-based learning for human behavior prediction. They extend the set of predictive modeling techniques for event or human behavior prediction.

This paper is well written and structured; I cannot find any weakness. The authors judiciously explain their methodology, provide a well-described background tutorial, and judiciously point out the novelty of their predictive human behavior modeling technique. I recommend this paper to students and researchers working on health informatics, human behavior prediction, and predictive analytics.