![]() | Only 14 pages are availabe for public view |
Abstract In Egyptian banking sector credit approval decision for personal loans is made using a pure judgment by credit officers due to the fact that machine learning is not widely used in practice. In this thesis, we have taken upon the challenge of delivering prediction approach for default customers of personal loans using machine learning. Our main objective is to predict default customers and analyze the trustworthiness of customers for getting a loan through available personal data and historical credit data. We used ABE dataset for training and testing, also we used 10 features from the application form and i-score report class that could be a helpful tool to credit staffs to take the right decision to decrease credit risk in order to avoid random techniques for customer selection. The data obtained were analyzed with different machine learning classification algorithms based on certain features in order to achieve higher accuracy. We compared between several methods before and after feature selection. We have observed that the most important features are (activity {u2013} income {u2013} loan amount) that can lead to high accuracy and decision tree has the best performance than any other machine learning algorithms with significant prediction accuracy of almost 94.85% |