Long Short-Term Memory and Gated Recurrent Unit for Automated Deep Learning Prediction | ||||
| IJCI. International Journal of Computers and Information | ||||
| Article 7, Volume 11, Issue 1, January 2024, Page 77-86 PDF (553.83 K) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/ijci.2024.235027.1119 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Esam Elgohary1; mohamed Galal2; Mostafa Aref3; Mona Gharib  4
| ||||
| 1Institute of National Planning, CLIP project manager, Cairo, Egypt | ||||
| 2Predictive Analytics department, National Bank of Egypt, Cairo, Egypt | ||||
| 3Computer Science Department, Faculty of Computer Science and Information Systems, Ain Shams University, Cairo, Egypt | ||||
| 4Mathematic Department, Faculty of Science, Zagazig University | ||||
| Abstract | ||||
| Recommender systems are nowadays an effective strategy to overcome the exponential growth of online products and services. The recommender systems assist customers to overcome over-choice-related problems and improve their satisfaction. This research presents an automated deep learning-based service for a personalized recommender system in the retail industry. This service automates deep learning data modeling processes regardless of the business case. It predicts the next best offer to a given customer based on the provided customers' behavior. Both Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) algorithms are used in parallel to train over the clients' data until the service chooses the best model performance to deploy. Finally, two case studies are presented to show the service performance in two different business cases. The first case study achieves a micro–Area Under Curve (AUC) score of 0.84 on a supermarket dataset, while the second one achieves 0.95 micro-AUC on the restaurant dataset. | ||||
| Keywords | ||||
| Predictive models; Deep learning; Automated prediction; Recommender systems | ||||
|
Statistics Article View: 99 PDF Download: 112 |
||||
