ARE THEY LIKELY TO COMPLAIN ON PHISH OR SPAM? A PREDICTION MODEL

Abstract

Customers are the core of any business. With the extreme competition between leading telecommunication companies, customer satisfaction is always considered a priority. High rates of customer satisfaction increase both customer retention and attraction rates. As a result, telecommunication companies are always seeking new methods and strategies to achieve these objectives. In a typical call center, large volume of calls is received from customers complaining on phishing or spam attacks on daily basis. However, it is not possible to manually identify the purpose of the call. Hence, there is need of an automatic system that will classify calls as complaints on spam or phish or for other purposes. In this work we expand on previous efforts to focus more on consumers of phone spam or phish and optimize call centers outcome. The study focuses on both mediums of communication, phone call or message. A historical sample of complaining customers database was taken and a proper technical approach was used to analyze the calls. The proposed methodology uses different machine learning algorithms to build a robust call classifier. The performance of the baseline classifier achieves an accuracy of 63.4\% that is based on CatBoost. The predictive model will be able to identify whether an individual is likely complaining or to complain on a spam or phish attack. Moreover, we will be able to identify phish or spam consumers demographics. The results show that people of age 45 are more likely to complain. On the other hand, males are less likely to complain. The system can also be used as a call routing mechanism to increase call centers quality by pairing the caller to the best suited agent. Finally, we point to directions for future work.

Date of Award	2019
Original language	American English
Awarding Institution	HBKU College of Science and Engineering