Article Preview
Top1. Introduction
When we work with a customer, we should know that customer satisfaction is the first important target (Hill & Alexander, 2006). In general, all the companies depend on customers as their backbone, especially the telecom companies on the significant increase in the number of telecom companies, that leads to customer migration from one company to another and create continuous warfare between the companies. So, all these companies try not only to retain their current customers but also to increase them; but how? And which set of customers that the company should take care of more than others? When dealing with customer behavior there is a set of customers who usually use the word-of-mouth and the same set of customers can negatively affect their neighbors, family, and friends. So, to determine which customers that may affect negatively in the telecom company reputation; will help the telecom company to proactively interact with them. But what is the way to determine this volatile customer? The proposed system defines a solution for all of these problems and more by observing the behavioral characteristics driven from the telecom customer conversations on the chatbot.
The problem entities are the identification of the volatile telecom customers that usually use the word of mouth; If they have a problem with the company, they may use the negative word of mouth and easily affect many other customers like their family friends and neighbors (Sato et al., 2018; Kim et al., 2016; Wilson et al., 2017). If the telecom company provides nothing to proactively interact with them and wins their consent and loyalty, they may cost the telecom company a heavy loss and puts it in an awkward position. So, this paper tries to help telecom companies to easily determine the high risk and volatile customers by observing behavioral characteristics derived from there textual data on the chatbot conversations and the language used in the conversation text. Analyzing customer textual data defines the big five personality traits that use to determine customer loyalty and customer satisfaction. References Castillo and Javier (2017) and Menidjel and Bilgihan (2017) show that customer loyalty can be determined from the Extraversion, Neuroticism and agreeableness values of the personality traits. Finally, and based on customer loyalty and use of word-of-mouth the proposed system divides customers into four clusters according to the dangerous percent. Hence, the proposed system uses the unsupervised machine learning k-means method in the clustering process whereas the k-means algorithm considers the best data mining clustering algorithm. Su et al. (2017) also says the proposed system uses the particle optimization technique to optimize the clustering results (Yazdani et al., 2017).
The most effective initial therapy for the volatile customer problem occurs if we can discover them in the early days of the problem. The late of discovering this problem or ignoring it costs the telecom company a heavy loss. So, detecting this problem automatically from the customer conversation textual data using machine learning, deep learning, and personality analysis is a vital process to give early warnings before it gets dangerous. Over the past years, there were some partially successful trials for finding a solution to increase the telecom customer loyalty by found a positive correlation between service recovery and customer satisfaction (Ibrahim et al., 2018) Reference Menidjel and Bilgihan (2017) try to find the relationship between big five personality traits, customer empowerment and customer satisfaction in the retail industry using the big five traits.