Hierarchically Structured Recommender System for Improving NPS of a Company

• Published in: Rough Sets and Current Trends in Computing pp. 347 - 357
• Main Authors: Kuang, Jieyan, Daniel, Albert, Johnston, Jill, Raś, Zbigniew W.
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2014
• Series: Lecture Notes in Computer Science
• ISBN: 331908643X; 9783319086439
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-08644-6_36

The paper presents a description of a hierarchically structured recommender system for improving the efficiency of a company’s growth engine. Our dataset (NPS dataset) contains answers to a set of queries (called questionnaire) sent to a randomly chosen groups of customers. It covers 34 companies called clients. The purpose of the questionnaire is to check customer satisfaction in using services of these companies which have repair shops all involved in a similar type of business (fixing heavy equipment). These shops are located in 29 states in the US and Canada. Some of the companies have their shops located in more than one state. They can compete with each other only if they target the same group of customers. The performance of a company is evaluated using the Net Promoter System (NPS). For that purpose, the data from the completed questionnaires are stored in NPS datasets. We have 34 such datasets, one for each company. Knowledge extracted from them, especially action rules and their triggers, can be used to build recommender systems giving hints to companies how to improve their NPS ratings. Larger the datasets, our believe in the knowledge extracted from them is higher. We introduce the concept of semantic similarity between companies. More semantically similar the companies are, the knowledge extracted from their joined NPS datasets has higher accuracy and coverage. Our hierarchically structured recommender system is a collection of recommender systems organized as a tree. Lower the nodes in the tree, more specialized the recommender systems are and the same the classifiers and action rules used to build their recommendation engines have higher precision and accuracy.