Enhancing Book Recommendation Systems through the Application of Sparse Features in Wide and Deep Learning Models
<p>Guo Jiajie</p> · International Journal of Frontiers in Engineering Technology · 2024
In processing large-scale user-item data, recommendation systems often face the challenge of information overload, making it critical to improve recommendation accuracy and efficiency. In recent years, deep learning-based recommendation algorithms have gradually become mainstream. However, due to the high sparsity of user-item interaction data, fully leveraging these sparse features to enhance model performance remains a pressing challenge in the field of recommendation systems. In this context, this paper investigates the application of sparse features in the Wide and Deep Learning model, based on the publicly available Book-Crossing dataset. To evaluate model performance, this study compares Random Forest, Gradient Boosting Decision Trees (GBDT), Multilayer Perceptron (MLP), and Wide and Deep Learning models using three metrics: Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Area Under the Curve (AUC). The experimental results demonstrate that the Wide and Deep Learning model outperforms the other models, particularly in terms of AUC and RMSE, confirming its advantages in handling sparse data and improving the performance of recommendation systems.