Machine Learning Approach and Model Performance Evaluation for Tele-Marketing Success Classification

Introduction

Today's world is going through a period in which the amount of data increases rapidly, and knowledge-based strategies and decisions become more important. This can be seen as a result of the development of information and communication technologies and the new systems where more large-scale data can be stored, and advanced analyses can be performed. The first step in accessing knowledge is storing the data. Today, data is collected from different sources. In order to make this stored data meaningful, it should be analyzed with the methods according to the structure of the data. Although the data analysis started with the use of statistical methods, today, data analysis is also carried out with Machine Learning algorithms as part of Artificial Intelligence (AI).

AI journey, which started with the Alan Turing's Turing test in the 1950s, continues today with driverless vehicles, more effective chat boxes and intelligent robots. AI is the ability to perform mental functions such as thinking, reasoning, and learning by computer or computer-controlled machines (Nabiyev, 2005). AI is a field that deals with the development of computer systems capable of processing symbols that can be used to solve problems that cannot be solved by algorithmic structures (Becerra-Fernandez et al., 2004). There are two basic steps of AI: Learning and practicing. In the learning step, various experiences are provided in order to determine the response of the computer systems in the face of an event, and in the second step, these systems are provided to produce new human-specific response without any command. Machine Learning methods are used in the learning step. Machine Learning is defined as the branch of computer algorithms developed to be used for transforming the stored data into a smart action by imitating the above-mentioned features of human (El Naqa & Murphy, 2015). Also, Deep Learning has recently become a prominent method of AI and Machine Learning. Deep Learning algorithms has been obtained by the development of Artificial Neural Networks (ANN) in order to make inference at a higher level and increase the predictive performance (Greenspan et al., 2016).

While all these developments are taking place, it is possible to talk about a world where competition conditions change. As a result of the globalizing world, the rivals have been moved from the national to the international level and the competition have become more difficult. At this point, to continue their existence and to determine the effective strategies and decisions, enterprises need to carry out new marketing and R&D activities and to use knowledge as an intellectual capital more effectively. In the banking sector, various campaigns are developed for customers and transactions with customers are wished to be kept active. Accordingly, one of the most important methods used to increase interactions and sales is marketing. With the development of telecommunication technologies, marketing has evolved towards telemarketing (Mustapha & Alsufyani, 2019). Telemarketing is a marketing method in which phone and call centers are used to acquire potential customers, sell products or provide services to existing customers (Kotler & Keller, 2016). However, the expectations, demands and needs of each customer may differ. For this reason, it should be determined which customers will be interested in the developed campaigns. With defining target customers, it is possible for the campaigns to reach higher success with less time and cost. Classification models to be created with machine learning algorithms and recorded data make it easier for tele-marketing managers to make more accurate and faster decisions at this point. Increasing marketing efficiency with data recorded in the database is one of the main issues that are still emphasized and require intensive research (Ghatasheh et al., 2020). Although many researchers carry out various studies with machine learning methods in line with this purpose, changing data set characteristics, new methods developed, different approaches in obtaining models show that this issue is open to research. For this reason, the problem of the success of telemarketing classification has been handled.