Recommender systems play a vital role in web-based information systems, especially in the domain of e-commerce. Most of these systems provide their recommendation based on user’s preferences. However, based on different situations of the user, their preferences can differ. Providing recommendations based only on the user’s preferences and ignoring their situation can be risky and can lead to inaccurate recommendations resulting in poor acceptance and a non-purchase. The aim of this research is to evaluate the risk of a user not purchasing before providing a recommendation to an ecommerce user. The proposed model classifies sessions as risky or non-risky, which indicates the likelihood of the session containing a purchase. The risk calculation is based on two perspectives and seven factors. We evaluate the work experimentally and use three metrics, which are True Positive Rate, True Negative Rate, and AUC (area under curve) of ROC (receiver operating characteristic) curve. The proposed model outperforms two state of the art algorithms on two different real world datasets, as measured by the AUC. The highest AUC of ROC is 0.97 when the users’ sessions history is included, and the lowest AUC of ROC is 0.79 when the users’ sessions history is not included. Our experiments show that the proposed risk calculation is a good predictor of a user’s purchase intention.