Responsive and intelligent service recommendation method based on deep learning in cloud service.

Area of Science:

• Cloud Computing
• Service Recommendation Systems
• Machine Learning

Background:

• The cloud service market faces service overload due to rapid expansion.
• Current service recommendations based on functional attributes neglect Quality of Service (QoS).
• QoS data sparsity and the cold-start problem hinder effective service recommendations.

Purpose of the Study:

• To enhance cloud service recommendations by incorporating non-functional attributes, specifically QoS.
• To develop a robust QoS prediction method to overcome data sparsity and cold-start issues.
• To improve user satisfaction with cloud service recommendations.

Main Methods:

• Organized QoS matrix data into service call records with user characteristics and QoS.
• Proposed a novel QoS prediction method utilizing a Gated Recurrent Unit Generative Adversarial Network (GRU-GAN).
• Employed time series data for QoS prediction and compared GRU-GAN against Collaborative Filtering (CF) and Matrix Factorization (MF).

Main Results:

• GRU-GAN demonstrated superior performance in QoS prediction compared to CF and MF.
• The proposed method effectively handles data sparsity, a significant challenge in QoS matrices.
• GRU-GAN provides more accurate QoS predictions, enhancing recommendation system effectiveness.

Conclusions:

• GRU-GAN offers a significant advancement in QoS prediction for cloud service recommendation systems.
• Addressing QoS data sparsity and cold-start problems leads to improved user experience and satisfaction.
• This research contributes to more intelligent and personalized cloud service discovery.