Graph Regularized Meta-path Based Transductive Regression in Heterogeneous Information Network.

Area of Science:

• Network science
• Machine learning
• Data mining

Background:

• Real-world networks often exhibit heterogeneity, comprising diverse node and link types.
• Numerical prediction in these heterogeneous information networks is challenging due to limited information for unlabeled objects.
• Existing methods for homogeneous networks may not effectively capture the complexity of heterogeneous data.

Purpose of the Study:

• To develop a novel transductive regression model for heterogeneous information networks.
• To address the challenge of limited network-based information for precise estimations in complex networks.
• To combine principles of graph-based transductive classification and homogeneous network regression.

Main Methods:

• Introduced a graph regularized meta-path based transductive regression model (Grempt).
• Integrated concepts from graph-based transductive classification and homogeneous network regression.
• Employed meta-paths to capture relational information within the heterogeneous network.

Main Results:

• The proposed Grempt model demonstrates effectiveness in numerical prediction tasks.
• The method is computationally efficient in terms of time and space complexity.
• Numerical experiments validate the precision of the Grempt model.

Conclusions:

• Grempt provides an effective solution for numerical prediction in heterogeneous information networks.
• The model's efficiency and precision make it suitable for real-world applications.
• This work advances transductive regression techniques for complex network data.