A deep learning based hybrid recommendation model for internet users.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Science

Background:

• Traditional recommendation systems (RS) face limitations in accuracy, scalability, efficiency, and handling the cold-start problem.
• Personalized item suggestions are crucial, but existing methods often fall short.

Purpose of the Study:

• To present the HRS-IU-DL model, a novel hybrid recommendation system designed to enhance accuracy and relevance.
• To address key challenges in recommendation systems, including data sparsity and the cold-start problem.

Main Methods:

• The HRS-IU-DL model integrates user-based and item-based Collaborative Filtering (CF), Neural Collaborative Filtering (NCF), and Recurrent Neural Networks (RNN).
• Content-Based Filtering (CBF) with Term Frequency-Inverse Document Frequency (TF-IDF) is used for item attribute analysis.
• N-Sample techniques, Cosine Similarity, Singular Value Decomposition (SVD), and TF-IDF are employed for recommending similar items based on user-specified genres.

Main Results:

• The HRS-IU-DL model demonstrates superior performance compared to state-of-the-art approaches on the Movielens 100k dataset.
• Significant improvements were observed across key evaluation metrics, including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Precision, and Recall.

Conclusions:

• The proposed HRS-IU-DL model effectively overcomes limitations of traditional recommendation systems.
• This hybrid approach offers substantial advancements in personalized recommendation technology, addressing sparsity and cold-start issues effectively.