Abstract:E-government service recommendation aims to find target service items for users based on their preferences, behaviors and other information, which is one of the key technologies for information overload in e-government cloud platforms. However, it has not received adequate attention in comparison to other recommendation scenarios like news and music recommendation. Graph Convolutional Networks (GCNs) have made significant strides in Collaborative Filtering (CF) recommendations. However, existing GCN-based CF methods are mainly based on matrix factorization (MF) and incorporate some optimization techniques (e.g., contrastive learning) to enhance performance, which are not enough to handle the complexities of diverse real-world recommendation scenarios. We empirically find that when existing GCN-based CF methods are directly applied to e-government service recommendation, they are limited by the MF framework and showing poor performance. This is attributed to the fact that the equal treatment of users and items by MF is not appropriate for scenarios where the number of users and items is unbalanced. In this work, we propose a new model, GCNSLIM, which combines GCN and sparse linear methods (SLIM) instead of combining GCN and MF to accommodate e-government service recommendation. In particular, GCNSLIM explicitly injects high-order collaborative signals obtained from multi-layer light graph convolutions into the item similarity matrix in the SLIM framework, effectively improving the recommendation accuracy. In addition, we propose two optimization measures, removing layer 0 embedding and adding nonlinear activation, to further adapt to the characteristics of e-government service recommendation scenarios. Furthermore, we propose a joint optimization mode to enhance the adaptability for more diverse recommendation scenarios. We conduct extensive experiments on a real e-government service dataset and two common public datasets and demonstrate the effectiveness of GCNSLIM in recommendation accuracy and operational performance. Our implementation is available at https://github.com/songkk5/GCNSLIM .

ITSM-GCN: Informative Training Sample Mining for Graph Convolutional Network-based Collaborative Filtering.

MixGCF

RGCF: Refined Graph Convolution Collaborative Filtering with concise and expressive embedding

HN-GCCF: High-order neighbor-enhanced graph convolutional collaborative filtering

IntentGC: a Scalable Graph Convolution Framework Fusing Heterogeneous Information for Recommendation

Diversifying Collaborative Filtering via Graph Spreading Network and Selective Sampling

Community-Enhanced Contrastive Learning for Graph Collaborative Filtering

Hypergraph Contrastive Collaborative Filtering

IA-GCN: Interactive Graph Convolutional Network for Recommendation

Cluster-based Graph Collaborative Filtering

Composition-Enhanced Graph Collaborative Filtering for Multi-behavior Recommendation

Graph-less Collaborative Filtering

GCNSLIM: Graph convolutional network with sparse linear methods for e-government service recommendation

SimpleX: A Simple and Strong Baseline for Collaborative Filtering

ENSG: Enhancing Negative Sampling in Graph Convolutional Networks for Recommendation Systems

Item Graph Convolution Collaborative Filtering for Inductive Recommendations

Dual channel group-aware graph convolutional networks for collaborative filtering

Reinforcement negative sampling recommendation based on collaborative knowledge graph

How Powerful is Graph Convolution for Recommendation?

SocialGCN: an Efficient Graph Convolutional Network Based Model for Social Recommendation

SimCGNN: Simple Contrastive Graph Neural Network for Session-based Recommendation