Abstract:The recommendation of appropriate development pathways, also known as ecological civilization patterns for achieving Sustainable Development Goals (namely, sustainable development patterns), are of utmost importance for promoting ecological, economic, social, and resource sustainability in a specific region. To achieve this, the recommendation process must carefully consider the region's natural, environmental, resource, and economic characteristics. However, current recommendation algorithms in the field of computer science fall short in adequately addressing the spatial heterogeneity related to environment and sparsity of regional historical interaction data, which limits their effectiveness in recommending sustainable development patterns. To overcome these challenges, this paper proposes a method called User Graph after Pruning and Intent Graph (UGPIG). Firstly, we utilize the high-density linking capability of the pruned User Graph to address the issue of spatial heterogeneity neglect in recommendation algorithms. Secondly, we construct an Intent Graph by incorporating the intent network, which captures the preferences for attributes including environmental elements of target regions. This approach effectively alleviates the problem of sparse historical interaction data in the region. Through extensive experiments, we demonstrate that UGPIG outperforms state-of-the-art recommendation algorithms like KGCN, KGAT, and KGIN in sustainable development pattern recommendations, with a maximum improvement of 9.61% in Top-3 recommendation performance.

What problem does this paper attempt to address?

The paper attempts to address the issue of overcoming the limitations of existing recommendation algorithms in terms of geospatial heterogeneity and the sparsity of historical interaction data when recommending sustainable development models. Specifically, the paper aims to improve the performance of recommendation systems by proposing a new method—User Graph Pruning and Intent Graph (UGPIG) method, thereby providing more accurate suggestions for sustainable development in specific regions. ### Main Issues: 1. **Geospatial Heterogeneity**: Existing recommendation algorithms often overlook the differences in natural, environmental, resource, and economic characteristics of different regions when dealing with geospatial features, leading to inaccurate recommendation results. 2. **Sparsity of Historical Interaction Data**: Historical interaction data for sustainable development models in specific regions are usually sparse, limiting the effectiveness of traditional recommendation algorithms. ### Solution: The paper proposes the UGPIG method, which mainly includes the following steps: 1. **Constructing User Graph (UG)**: By collecting geospatial feature data of the target region, a semantic network containing the region and its features is constructed. 2. **User Graph Pruning and Reconstruction (UGP)**: Pruning continuous feature nodes in the user graph and reintroducing these nodes through a discretization process to enhance the relevance of the knowledge graph. 3. **Constructing Intent Graph (IG)**: Using historical interaction data and manually defined intent vector sets to construct an intent graph, capturing users' preferences and intentions for sustainable development models. 4. **Information Aggregation and Feature Fusion**: Performing information aggregation on both UGP and IG to extract heterogeneous features, historical interactions, and intent features of the target region, then fusing the two types of features through an attention mechanism to generate the representation of the target region and calculate its preference scores for various sustainable development models. ### Main Contributions: 1. **Validating the Impact of User Graph Pruning and Intent Graph on Data Sparsity and Spatial Heterogeneity**: Designed and implemented the pruning steps of the knowledge graph to ensure the user graph is more refined and representative. Developed the intent graph to effectively capture historical interaction features and user intentions, improving recommendation performance. 2. **Proposing the UGPIG Method**: This method specifically targets the environmental geography field, successfully addressing the sparsity issue among recommended entities and incorporating spatial heterogeneity into the recommendation process. Through provincial case studies, the practical application of UGPIG is demonstrated, laying the theoretical and methodological foundation for future research on intelligent sustainable development pathways.

Unveiling Optimal SDG Pathways: An Innovative Approach Leveraging Graph Pruning and Intent Graph for Effective Recommendations

DPGN: Denoising Periodic Graph Network for Life Service Recommendation

Intent-aware Graph Neural Network for Point-of-Interest Embedding and Recommendation

Exploiting Structural and Temporal Influence for Dynamic Social-Aware Recommendation

STGIN: Spatial-Temporal Graph Interaction Network for Large-scale POI Recommendation

Social-Enhanced Explainable Recommendation With Knowledge Graph

Graph-enhanced Spatial-temporal Network for Next POI Recommendation

GUESR: A Global Unsupervised Data-Enhancement with Bucket-Cluster Sampling for Sequential Recommendation

Automatic Hypergraph Generation for Enhancing Recommendation with Sparse Optimization

Unveiling User Satisfaction and Creator Productivity Trade-Offs in Recommendation Platforms

Personalized Recommendation via Multi-dimensional Meta-paths Temporal Graph Probabilistic Spreading

DUSKG: A Fine-Grained Knowledge Graph for Effective Personalized Service Recommendation.

Graph-Sequential Alignment and Uniformity: Toward Enhanced Recommendation Systems

GPT4Rec: Graph Prompt Tuning for Streaming Recommendation

Exploring Behavior Patterns for Next-POI Recommendation via Graph Self-Supervised Learning

Building and exploiting spatial–temporal knowledge graph for next POI recommendation

A Hierarchical Optimizer for Recommendation System Based on Shortest Path Algorithm

Proactive Recommendation with Iterative Preference Guidance

A POI-Sensitive Knowledge Graph Based Service Recommendation Method.

Global and Personalized Graphs for Heterogeneous Sequential Recommendation by Learning Behavior Transitions and User Intentions

Eco-Aware Graph Neural Networks for Sustainable Recommendations