Deep Learning Approach for Detecting Customer Churn in Telecommunication Industry

Deep Learning Approach for Detecting Customer Churn in Telecommunication Industry

R. P. H. Liyanage, B. T. G. S. Kumara, Banujan Kuhaneswaran, Senthan Prasanth
DOI: 10.4018/978-1-7998-9553-4.ch009
(Individual Chapters)
No Current Special Offers


In today's business world, customer turnover is a significant problem. Communications companies aren't exempt from these problems. Retaining consumers is more important than recruiting new ones when it comes to business. Getting new clients is about five times as expensive as keeping old ones in this field. As a result, anticipating client turnover is a huge challenge for almost all organizations. This study focused on analyzing information on around 7000 post-paid subscribers by considering 21 different attributes. Initially, the data was fed into machine learning techniques such k-nearest neighbors, artificial neural networks, etc. In addition, deep neural networks (DNN) have also considered more than one hidden layer. A total of 4284 of the 7234 post-paid customers are considered non-churners, while the remaining 2950 are churners. The long short-term memory networks (LSTM) considered under the DNN produce results far superior to the other techniques, with the highest accuracy rate of 82.46%. Finally, the LSTM method was used to create the final prediction model.
Chapter Preview


Clients are a firm’s most precious asset. It must retain a solid client base while being competitive as the firm invests in its customers. If not, clients will quit (churn) the existing firms or service providers. The company’s reputation and economic sustainability may be compromised as a result. “Customer Churn” means removing a consumer commitment or a business line (supporter, user, subscriber, etc.). For instance, when a client is a frequent client of “A” businesses, the service provider switches to “B” businesses (Amin et al., 2017; Olle & Cai, 2014). Customer happiness is crucial to the success of any organization. Customer churn or retention levels are key industry measurements used by banks, telecoms companies, airlines, Broadband providers, pay-TV companies, and insurance organizations. The churn rate is significant because it indicates how consumers respond to goods, prices, and competitors. To lower customer churn, the capacity to recognize churn content/behaviors would allow early response procedures as part of retention efforts.

Churn estimation is crucial nowadays in the telecommunications sector. Its popularity has skyrocketed in recent years. The telecommunications industry has several service providers because of the need for effective contact. People have a limited number of service provider alternatives, and practically everyone has a telecom membership. Customer churn is common in this market since it is simple for subscribers to migrate to another service provider, including unsatisfied customers, switching costs, outstanding service consumption, and poor customer service (Sayed, Abdel-Fattah, & Kholief, 2018; Sharma, Panigrahi, & Kumar, 2013). Besides traditional theories, there are indeed emerging causes for client retention, such as unhappiness with the early interaction with the service provider and the difficulties achieving consumers’ unrealistic expectations.

The industry adopts a variety of techniques to remain competitive. The business or organization may need to develop new valuation customer services retention initiatives to accomplish this. Paying members is much more valuable than acquiring a new customer (N. Lu, Lin, Lu, & Zhang, 2012). Obtaining new clients is more expensive, and therefore, it is desirable to keep present ones. The corporation strives to keep long-term consumers seeing as they are its cash cow. The company spends a large amount of money on customer churn estimates in maintaining customers. The following are the two major styles of churners: Churners are classified as either voluntary or involuntary. Voluntary churners (Active churners) voluntarily leave a supplier. The business faces a challenging challenge in identifying and predicting these possible churners. Passive churners (involuntary churners) are fired without their permission by a corporation or group. These churners are easily recognized and terminated for various reasons, including the subscription list’s long-term failure to pay charges and fraud activities (Amin et al., 2019; Amin et al., 2017; Shaaban, Helmy, Khedr, & Nasr, 2012). Accidental churners and intentional churners are the two types of voluntary churners. Here’s a brief rundown of the churners as mentioned above.

  • 1.

    Incidental Churn Customers

Customers who leave a service provider due to personal reasons such as a change in their financial status or a relocation.

  • 2.

    Deliberate Churn Customers

This happens because of technological advancements. A change in service quality or psychological factors, such as a customer’s preference for new technology.

We can design a customer churn prediction model and detect the most likely list of churners by conducting thorough research with the collected data from customers. This strategy will particularly aid us in identifying consumer attrition among those who are most likely to leave. These strategies may also assist the company in addressing the problem mentioned above by establishing a model for retaining future churners. As a result, it should go without saying that developing the above model will save time and money. In the telecommunications industry, several churn predictions models have already been developed. These models use a variety of algorithms, such as Decision trees, Regression models, Neural networks, Bayesian models, SVM, and so on (Ahmad, Jafar, & Aljoumaa, 2019). The general goal of this study is to create a new hybrid churn prediction model that employs a multi-layer methodology.

Key Terms in this Chapter

Artificial Neural Network (ANN): ANN seeks to mimic the network of neurons that make up the human brain, allowing the computer to learn and make decisions in a human-like manner. Traditional programming computers create ANNs that operate like interconnected brain cells.

Unsupervised Learning: Machine learning procedure that involves the construction of models without the usage of a training set. Models unearth hidden patterns and insights in the data that is provided.

Deep Neural Network (DNN): A DNN’s input and output layers are separated by multiple levels. Neurons, synapses, weights, biases, and functions are crucial components of neural networks, regardless of their size or configuration.

Deep Learning: Deep learning is a method for teaching computers to mimic human behaviour. Deep learning models can sometimes outperform humans in terms of accuracy.

Supervised Learning: Computers are educated on labelled training data and then used to predict output in supervised learning, a subset of machine learning.

Machine Learning: This is the case of algorithms that learn on their own through experience and data intake.

Complete Chapter List

Search this Book: