RE-GrievanceAssist: Enhancing Customer Experience through ML-Powered Complaint Management (2404.18963v1)

Abstract: In recent years, digital platform companies have faced increasing challenges in managing customer complaints, driven by widespread consumer adoption. This paper introduces an end-to-end pipeline, named RE-GrievanceAssist, designed specifically for real estate customer complaint management. The pipeline consists of three key components: i) response/no-response ML model using TF-IDF vectorization and XGBoost classifier ; ii) user type classifier using fasttext classifier; iii) issue/sub-issue classifier using TF-IDF vectorization and XGBoost classifier. Finally, it has been deployed as a batch job in Databricks, resulting in a remarkable 40% reduction in overall manual effort with monthly cost reduction of Rs 1,50,000 since August 2023.

• The paper's main contribution is the development of RE-GrievanceAssist, an ML-powered system that reduces manual effort in real estate complaint management by 40%.
• It employs TF-IDF vectorization with XGBoost and FastText classifiers to determine response needs, classify users, and categorize complaint issues efficiently.
• Deployed on the Databricks platform, the system achieves scalable processing and operational cost savings, enhancing overall customer experience.

The paper "RE-GrievanceAssist: Enhancing Customer Experience through ML-Powered Complaint Management" addresses the growing challenge faced by digital platforms in efficiently managing customer complaints, particularly in the real estate sector. To tackle these challenges, the authors introduce an innovative system called RE-GrievanceAssist, which provides an end-to-end solution specifically tailored for handling real estate complaints.

The RE-GrievanceAssist system is composed of three key machine learning components:

1. Response/No-Response Model: This component aims to determine whether a customer complaint requires a response. It utilizes TF-IDF (Term Frequency-Inverse Document Frequency) vectorization to process text data, which transforms the text into numerical values that reflect the importance of each term. An XGBoost classifier is employed for making the final decision on whether a response is necessary.
2. User Type Classifier: This part of the pipeline categorizes users based on the type of interaction or complaint they raise. The FastText classifier, known for its efficiency in handling text classification tasks, is utilized here to quickly and accurately classify users into predefined categories, enabling more personalized and relevant handling of complaints.
3. Issue/Sub-Issue Classifier: Similar to the response/no-response model, this classifier also uses TF-IDF vectorization, followed by an XGBoost classifier, to identify and categorize the specific issues and sub-issues raised in complaints. This classification is crucial for streamlining the resolution process by directing complaints to the appropriate departments or solutions.

The solution is deployed as a batch job on the Databricks platform, which provides scalable and efficient data processing capabilities, essential for handling the large volumes of complaints typical in real estate.

Significantly, the implementation of RE-GrievanceAssist has led to a substantial 40% reduction in manual effort required to manage complaints. This efficiency gain translates into a monthly cost saving of Rs 1,50,000 since August 2023. The paper showcases how integrating machine learning solutions in complaint management can not only enhance customer experience but also result in significant operational cost benefits.