| Summary: | Handling customer complaints is a decision-making process that inherently involves a classification problem where each complaint should be classified exclusively to one of the complaint categories before a resolution is communicated to customers. Previous studies focus extensively on decision support systems (DSSs) to automate complaint handling, while few addresses the issue of classification imprecision when inaccurate or inconsistent information exists in customer complaint narratives. This research presents a novel DSS for handling customer complaints and develops an evidential reasoning (ER) rule-based classifier as the core component of the system to classify customer complaints with uncertain information. More specifically, textual and numeric features are firstly combined to generate evidence for formulating the relationship between customer complaint features and classification results. The ER rule is then applied to combine multiple pieces of evidence and classify customer complaints into different categories with probabilities. An empirical study is conducted in a telecommunication company. Results show that the proposed ER rule-based classification model provides high performance in comparison with other machine learning algorithms. The developed system offers telecommunication companies an informative and data-driven method for handling customer complaints in a systematic and automatic manner. © 2018 Elsevier B.V.
|
|---|
