Abstract:
"
One of the major pain points which telecommunication companies are facing is the
higher number of customer churn which is currently around 4% .Due to saturated
telecommunication market it has led to a fierce competition among the operators. Apart
from the sentimental value that a customer has for his/her current mobile number,
considering other operators that provide quality networks and packages becomes more
attractive. In that sense, data users have more freedom to select another convenient
operator as they do not have the weight of an existing number. Therefore, customer
churn is a common problem among the operators and distinguishing the possible churn
customers in prior is a challenge due to lack of information and accurate models to
predict.
In order to overcome this problem, there are two options; either retention incentive
schemes conducted throughout the entire customer base or focus on the new customer
accusation plan. Both the options are costly and will not have significant gain over
action taking for high probability of churning customers in near future. It is less costly
to focus on customer retention rather than customer acquisition. Recent studies show
that acquiring a new customer is 5 times costly than retaining an existing customer.
Churn prediction models have been a hot research topic in recent years. There are many
research materials on this topic locally as well as globally. However, there is a research
gap on predicting customer churn models with available customer transactions and
network data as the current methods mainly rely on surveys. An accurate churn
prediction model largely depends on complete and quality historical data. Currently
operators are unable to grasp all the necessary data about customers and only take
actions based on the assumptions made with available information and ground staff
knowledge from the sales and customer service teams.
Objective of this research is to fill the gap of the customer churn identification model
to help the telecommunication companies to proactively identify the customers who"
