Short-time traffic flow prediction is essential for intelligent traffic management. By accurately predicting traffic conditions in the near future, it helps to alleviate congestion, improve road efficiency, reduce accidents, and support timely traffic control. Short-time traffic flow exhibits uncertainty and randomness, and this paper proposes an SVR model for short-time traffic flow prediction on non-main and branch roads, using correlations between associated roads to improve accuracy. Association Rule Analysis: First, we use Pearson correlation to identify strongly correlated roads. This step helps in understanding the relationships between different roads and their traffic patterns. SVR Model Construction: Second, based on the identified correlations, we construct an SVR model using traffic data from the target road and its associated roads. The model parameters are optimized using grid search and cross-validation to ensure the best performance. Simulation and Evaluation: Third, we conduct simulation experiments using real traffic data from Xi’an city. The performance of our model is evaluated using metrics such as root mean square error (RMSE), mean absolute error (MAE), and mean relative error (MRE). Simulation experiments show that our model outperforms existing methods. Specifically, our model achieved an RMSE of 11.422, an MAE of 7.017, and an MRE of 0.139. In comparison, other models tested in our study, such as LSTM, Random Forest, and Gradient Boosting Decision Tree (GBDT), had higher error values. For instance, the LSTM model had an RMSE of 14.5, an MAE of 8.2, and an MRE of 0.165; the Random Forest model had an RMSE of 13.8, an MAE of 7.8, and an MRE of 0.152; and the GBDT model had an RMSE of 13.2, an MAE of 7.5, and an MRE of 0.148. These results demonstrate that our proposed SVR model, combined with association rules, is highly effective in predicting short-time traffic flow on non-main and branch roads, which are often overlooked in existing research.

To improve recommendation systems, it is essential to enhance both their practicality and accuracy, thereby supporting users in making informed shopping decisions. Incorporating two types of product relationships can effectively achieve these goals: first, the product relationships, like complements, and second, the social relationships among users. However, existing studies have paid little attention to user-side information or item-side information. This paper proposes a product recommendation model that utilizes cross-elasticity of demand and hypergraphs, referred to as Hg-CR. First, users and items build a hypergraph. The user–item interactions form the hyperedges. Also, users build a hypergraph between themselves based on their social relationships. Second, hypergraph attention networks (HANs) learn the relationships between nodes. They capture the key features of nodes and hyperedges with a high degree of adaptability. A community detection algorithm organizes users into groups for product recommendations by assessing their similarities. Within different communities, individuals seek complementary products based on the cross-elasticity theory of demand. Additionally, we provide recommendations for complementary products. Tests on real datasets show that the Hg-CR model is about 10% more accurate than the other baseline models.

In the era of information explosion, as a form of personalized recommendation, dialogue recommendation systems provide users with personalized recommendation services through natural language interaction. However, in the face of complex user preferences, the traditional dialogue recommendation system has the problem of a poor recommendation effect. To solve these problems, this paper proposes a user preference dialogue recommendation algorithm (KGCR) based on a knowledge graph, which aims to enhance the understanding of user preferences through the semantic information of the knowledge graph and improve the relevance and accuracy of recommendations. This paper proposes a personalized conversation recommendation algorithm framework for user preference modeling. The framework uses a bilinear model attention mechanism and self-attention hierarchical coding structure to model user preferences to rank and recommend candidate items. By introducing rich user-related information, the recommendation results are not only more in line with users’ individual preferences but also have better diversity, effectively reducing the negative impact of information cocoons and other phenomena. At the same time, the experimental results on the open dataset prove the effectiveness and accuracy of the proposed model in the personalized conversation recommendation task.

This paper presents an optimized twin-tower model for text retrieval that addresses limitations in traditional models through improved feature interaction and loss function design. We introduce an early interaction layer using cross-attention mechanisms and a ranking-optimized loss function. These innovations enable earlier feature interactions between queries and documents, enhance semantic relationship understanding, and optimize relative similarity rankings while reducing overfitting risk. Our experiments on NQ, TQA, and WQ datasets show substantial Top-K accuracy improvements over benchmark models like BM25, DPR, ANCE, and ColBERT. For example, our model achieves a 20.3% relative improvement in Top-20 accuracy on NQ compared to BM25, with only 17 ms retrieval latency. Ablation studies confirm the effectiveness of our improvements. This research demonstrates that enhancing feature interaction and optimizing loss functions significantly improves twin-tower model performance, providing valuable methodological insights for efficient semantic retrieval while maintaining computational efficiency.