Deep Learning-Based Long and Short-Term Memory Integration for Enhanced Library Recommendation Systems

Library recommendation systems play an essential aspect in helping users navigate large collections on an individual basis. However, traditional recommendation techniques rarely take into consideration such aspects as temporal dynamics in users' preferences. This paper presents research that aims at enhancing library recommendation systems through the incorporation of deep learning mechanisms with long and short-term memory. We apply recurrent neural networks (RNNs) to model temporal features of the user interactions and make a comparison with Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). Moreover, attention mechanisms, are applied to identify important features within both long and short-term memory frameworks. In a set of experiments on real-world library usage data, we show that our approach outperforms the given baseline methods in terms of recommendation accuracy and relevance contributing to the enhancement of recommendation systems in libraries.