Mining Totally Ordered Sequential Rules to Provide Timely Recommendations

In this paper we show the importance of mining totally ordered sequential rules, and in particular we propose an extension of sequential rules where not only the antecedent precedes the consequent, but their itemsets are labelled with an explicit representation of their relative order. This allows us to provide more precise timely recommendations. Our technique has been applied to a real-world scenario regarding the provision of tailored suggestions for supermarket shopping activities.