Machine Learning Workflow Recommendations Based on Hierarchical Topics

Machine learning workflow platforms are growing rapidly. However, for developers, developing a machine learning workflow from scratch involves tedious and error-prone tasks such as component orchestration and parameter configuration. In response to this issue, this paper proposes a hierarchical topic-based machine learning workflow recommendation approach. This approach designs a hierarchical model for machine learning workflows based on the data characteristics of historical workflow templates. For the hierarchical model, it extracts key topic keywords from each layer and calculates the similarity between different workflows based on the topics. As a result, it recommends developers based on the similarity between historical machine learning workflows and the one currently being developed.