CNN Hyperparameter Optimization for Customer Emotion Classification Based on Voice Recording Analysis
Author :
Abstract
Call centers play a crucial role as communication channels for companies providing customer services, and efficient management of incoming calls is essential for customer satisfaction and company performance. The use of data analytics and artificial intelligence techniques for evaluating and improving call center performance is increasingly prevalent. This article presents a detailed analysis of a call center dataset belonging to Gurmen Textiles, a company specializing in textile products with a strong focus on customer service. The company utilizes its call center to address customer inquiries and aims to maximize customer satisfaction. In this study, a deep learning model was developed using the call center dataset of Gurmen Textiles to differentiate between irate and normal calls. Irate calls may indicate customer dissatisfaction or issues, while normal calls typically encompass routine customer requests. Accurate classification of irate and normal calls can assist the company in managing customer relationships and enhancing service quality. By leveraging data analytics and deep learning techniques, analyses of call center data can be conducted more efficiently, leading to positive impacts on customer satisfaction. Speech analysis, a commonly employed technique in such analyses, can provide valuable insights into customer emotional states or concerns by evaluating audio recordings in call center data. The proposed approach achieves compatible results in SAVEE public access dataset with %98 and significant result in Gurmen Textiles dataset with %97 accuracy of classification.
Keywords
Abstract