A Graph Oriented Based Recommender System for Financial Products and Services

A Graph Oriented Based Recommender System for Financial Products and Services

Author by Dr. Oluwaseun Ebiesuwa

Journal/Publisher: International Journal Of Service Science And Management

Volume/Edition: 3

Language: English

Pages: 1 - 16

Abstract

Recommender Systems are intelligent applications designed to
assist the user in a decision-making process whereby user wants
to choose one item amongst the potentially overwhelming set of
alternative products or services. This research is aimed at developing
an intelligent recommender system that provides high quality
recommendations in the financial domain. Hashed and anonymized
datasets (which are account statements) were acquired
from online sources and bank customers. The acquired data was
pre-processed using the Microsoft Excel 2016 and WEKA 3.8.3
data mining software. The K-nearest neighbor (KNN) algorithm
was used to classify the dataset and train the model. The trained
model was used to develop a recommender system using the
Java 2 platform Enterprise Edition (J2EE). For effective management
of the data and consideration of rapid increase in data
growth, a graph-oriented database approach was proposed and
utilized. The database management system used was the Neo4j.
From the evaluation of the algorithms implemented in the recommender
system taxonomy, the KNN algorithm recorded the best
performance building the model in 0.3seconds with an accuracy
of 89.8%. The fuzzy decision tree algorithm performed second
best building the model within 0.48 seconds with an accuracy of
62.8%. The decision table algorithm performed poorly building
the model in 3.9 seconds with an accuracy of 53%. However,
the baseline accuracy of the dataset used was evaluated to be
62.75% of accuracy in 0.4 seconds. It is therefore recommended,
as proposed in this study that the graph technology be used in
developing recommender systems especially for institutions with
massively growing data like the financial institutions. In addition,
bank products should be classified and targeted towards customers
in order to bolster their level of involvements and improve
financial inclusion. With a targeted product, customers will be
more willing to opt-in if products are suitable and within financial
reach. This will help financial institutions earn more and the customer’s
financial power will also be strengthened.


Other Co-Authors