Using Partially-Ordered Sequential Rules to Generate More Accurate Sequence Prediction

Predicting the next element(s) of a sequence is a research problem with wide applications such as stock market prediction, consumer product recommendation, and web link recommendation. To address this problem, an effective approach is to mine sequential rules from a set of training sequences to then use these rules to make predictions for new sequences. In this paper, we improve on this approach by proposing to use a new kind of sequential rules named partially-ordered sequential rules instead of standard sequential rules. Experiments on large clickstream datasets for webpage recommendation show that using this new type of sequential rules can greatly increase prediction accuracy, while requiring a smaller training set.