Designing Location Aware Active Atm Recommender for Banking Service
No Thumbnail Available
Date
2020-02-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Card banking customer in Ethiopia faces the problem of identifying nearby active ATM service.
Research shows that more than 50% of card banking customers are unsatisfied with the existing
card banking service at their respective banks in Ethiopia. The aim of this study is to explore and
design location aware active ATM recommender for banking service. To this end, the study first
did an experiment to construct a predictive model for determining active ATM. This is followed
by identifying nearby active ATM and presenting the optimal path from customer current position
to nearest active ATM.
For active ATM prediction J48 and PART algorithms are experimented. In this experiment 44,105
instances with 8 attributes employed and WEKA tool is also used to construct a prediction model.
For determining nearest active ATMs, Euclidean, City-block and Haversine algorithms with 13
records were employed. Also the experiment is conducted using GoogleMaps tool. For
determining the optimal path Dijkstra’s, A*, and BFS algorithms and 54,654 instances were used.
To carry out this experiment GoogleMaps tool has been used. For developing a prototype
application java program and MySQL tools has been used.
The experimental result for predicting active ATMs shows that J48 decision tree has better
accuracy i.e. 99.81 compared to PART rule induction algorithm. The experiment result for
determining nearest active ATMs shows that City-block has better performance than Euclidean,
and Haversine algorithms. The experiment result for determining the optimal path shows that BFS
has better performance than Dijkstra’s, and A* algorithms.
Finally, we design a prototype for recommending nearby active ATMs using J48 model, Cityblock
and BFS algorithms. Accordingly, the prototype recommends the nearest active ATMs with
the optimal path as per the user preference (ATM type and radius) and other related nearest active
ATMs which is different than the user preference. User acceptance test result shows that, the
prototype system save time for the customer. As a result, the customers are satisfied by the
prototype system. This study has limitation with respect to user need in how much effort the
customer wants to get ATM service and studying this issue enhance customer satisfaction.
Description
Keywords
Predict Active Atm, Location Aware, Determine Nearest Active Atm, Determine Shortest/Optimal Path, Recommender