This study aims to develop a feature importance aware deep neural
network model for explainable recommender systems, which facilitates
improved predictive performance by considering the feature importance
within the training procedure and generates a recommendation list based
on the user’s preference. The proposed method minimizes the combination
of model losses and introduces a penalty term that specifically targets
and diminishes the detrimental impact of a particular feature on
the solution’s efficacy. This strategy ensures the determination of the
optimal solution by satisfying the explainable conditions imposed during
the training framework. Extensive experiments on publicly accessible
FilmTrust and MovieLens-100K datasets show notable recommendations
performance.
@InProceedings{10.1007/978-3-032-10489-2_10, author="Gupta, Pragya and Aishwaryaprajna and Guha, Debashree and Chakraborty, Debjani", editor="Mart{'i}nez, Luis and Camacho, David and Yin, Hujun and Dutta, Bapi and Yera, Raciel and Rodr{'i}guez Dom{'i}nguez, Rosa M. and Tall{'o}n-Ballesteros, Antonio", title="Feature-Importance Aware Deep Neural Network Model for Explainable Recommender Systems", booktitle="Intelligent Data Engineering and Automated Learning -- IDEAL 2025", year="2026", publisher="Springer Nature Switzerland", address="Cham", pages="109--120", }