Using AI techniques and historical data analysis for enhancing decision making

Other Title(s)

استخدام تقنيات الذكاء الاصطناعي و تحليل البيانات التاريخية لتعزيز اتخاذ القرار

Dissertant

Surur, Tariq Samir Isa

Thesis advisor

al-Harub, Ayish M.

University

Isra University

Faculty

Faculty of Information Technology

Department

Department Software Engineering

University Country

Jordan

Degree

Master

Degree Date

2022

English Abstract

Nowadays, the collection of due payments from the customers are the primary challenge in the companies.

Either the companies have a big or small business, collecting payments from the products or services sold to its customers is the main element of its success and continuance.

With the booming of the customer base and sales of complex products and services, the collection payments process becomes more difficult because it is not a pretty straightforward process.

This dilemma urges a solid collection strategy based on a dynamic and advanced AI solution to automate the collection management process without human intervention to mitigate the risk and collect the due amounts immediately.

This work introduces an approach using different AI algorithms (Random Forest, Support vector machine, and generalized linear model) to determine the factors leading to selecting the best collection scenario in order to enhance the system usability.

Furthermore, dependent and independent variables have to be identified in terms of selecting the factors of best collection scenarios, the Correlation Coefficient (CC) will be used in terms of identifying the relation of variables for this purpose, a big dataset of customer profiles are collected, this dataset is derived from the real business world (ESKADENIA Software Company).

As expected, the final model will face a challenge that is related to frequent customer behavior changes.

This work will consider the changes and re-learn the machine relay to these changes to avoid this challenge.

The dataset is separated into training and verification data by a 7:3 ratio.

As a result, the SVM model had the lowest accuracy (31.8%), while the GLM model had (96.52%) accuracy, while the Random Forest model had the highest accuracy (98.44%).

Main Subjects

Information Technology and Computer Science

No. of Pages

63

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-1414523