In Search for Best Meta-Actions to Boost Businesses Revenue

The ultimate goal of our research is to build recommender system driven by action rules and meta-actions for providing proper suggestions to improve revenue of a group of clients (companies) involved with similar businesses. Collected data present answers from 25,000 customers concerning their personal information, general information about the service and customers’ feedback to the service. This paper proposes a strategy to classify and organize meta-actions in such a way that they can be applied most efficiently to achieve desired goal. Meta-actions are the triggers that need to be executed for activating action rules. In previous work, the method of mining meta-actions from customers’ reviews in text format has been proposed and implemented. Performed experiments have proven its high effectiveness. However, it turns out that the discovered action rules need more than one meta-action to be triggered. The way and the order of executing triggers causes new problems due to the commonness, differential benefit and applicability among sets of meta-actions. Since the applicability of meta-actions should be judged by professionals in the field, our concentration is put on designing a strategy to hierarchically sort and arrange so called meta-nodes (used to represent action rules and their triggers in a tree structure) as well as to compute the effect of each meta-node. Furthermore, users will have more concrete options to consider by following the path in trees built from these meta-nodes.