A Fast-training Approach Using ELM for Satisfaction Analysis of Call Centers
Jing Liu, Yingnan Zhang, Jin Hu, Xiang Xie, Shilei Huang · 2017
Analysis of the customers' satisfaction guarantees the improvement of service quality in call centers. In this paper, an intelligent satisfaction recognition system is introduced to analyze the customers' satisfaction through the customers' emotion recognition. The nature dialogues are collected from the Chinese call center. Support Vector Machine (SVM) and Extreme Learning Machine (ELM) are used for the mapping model respectively. According to the experiment, the best F score of SVM is 0.71. Compared to SVM, the best F of ELM is up to 0.723. The training time of SVM ranges from 1268s to 5002s while ELM's only ranges from 7.28s to 15.82s, with a decrease of 99%. ELM shortens the training time largely without damaging the performance. Because of the faster training speed, ELM is more beneficial to the model updating in real time. Therefore, ELM has a great edge on online learning.