关键词: 水情信息, 分发, 实时推荐, ItemCF, LSTM, 二分类模型, 优化

Abstract: The demand for real-time recommendation of water information is growing stronger with the deepening of water conservancy informatization in China. Since the data of water is highly time-sensitive, recommendation system is required to provide real-time recommendation services. User-based collaborative filtering and item-based collaborative filtering (ItemCF) are two commonly used algorithms in the recommendation field. Both, however, are offline algorithms in nature and cannot meet the requirement of real-time distribution of water information. In this paper, a real-time recommendation algorithm for water regime information distribution based on Long-Short-Term Memory (LSTM) is proposed and optimized to ensure the accuracy of water information recommendation while ensuring the real-time recommendation.

Key words: water information distribution, real-time recommendation, ItemCF, LSTM, dichotomous model, optimization