Page 421 - Contributed Paper Session (CPS) - Volume 2

CPS1915 Han G. et al.

Neural networks with different activation
functions applied in bank telemarketing
1
2,3
1
Han Gao ; Pei Shan Fam ; Heng Chin Low ; Xinwu Pan
1
1 School of Mathematical Sciences,Universiti Sains Malaysia
2 School of Computer Sciences, Harbin Institute of Technology, Harbin, China
3 Senior Front-End Developer, Alibaba Local Services Company, Hangzhou, China

Abstract
Telemarketing is broadly used in bank sector. It is a convenient and cost-
effective approach to selling products and services to the clients. An accurate
and reliable method for customer segmentation is essential in marketing.
Neural network is gaining more and more attention in classification area. The
objective of this research is to create a neural network model to single out
those who possibly will buy the term deposit, which will facilitate the decision
making in bank telemarketing. After the training process, the optimal
prediction accuracy (73.11%) for Model I occurred when the hidden layer with
50 neurons and the learning rate of 0.1. For Model II, it outperformed when
the hidden layer with 20 neurons and the learning rate of 1.2. Overall, the
proposed neural network model showed a promising prediction power for the
given data.

Keywords
Decision making; bank deposit; sigmoid function; tanh function

1. Introduction
Telemarketing, as a direct marketing, is an effective and convenient way to
sell products as well as services in several industries, such as medicine,
insurance and finance. With the rise of predictive dialer technology,
telemarketing began in the early 1990s (Hurst, 2008). It facilitates the way for
buyers and sellers to make a transaction. In banking sector, telemarketing
plays an irreplaceable role in marketing. Bank-telemarketing is a quite popular
way around the globe.
Effective decision-making matters a lot in marketing campaign. Decision
support systems (DSSs) are a set of useful tools to support the managerial
decision-making, which includes several sub-fields, such as personal DSS and
intelligent DSS (Moro et al., 2014). In Power’s (2008) research, he found out
that it can enhance the efficiency and accuracy in decision making when using
DSSs. Using artificial intelligence (AI) techniques to process the consumer data
is gaining more and more popularity (Alon et al., 2001; Huang et al., 2007;
Mazhar et al., 2007; Moro et al., 2014; Moro et al.,2015). Data mining (DM),
widely applied in AI and statistics, plays a vital role in DSSs, which allows the

410 | I S I W S C 2 0 1 9