Abstract:

[Objective] The big data-driven ocean artificial intelligence service platform integrates ocean artificial intelligence-related algorithms, software tools, and data resources to build a scientific research information environment that supports ocean artificial intelligence research. It provides a "one-stop-shop" machine learning as a service for scientific research users. [Methods] Based on the actual needs of data, software, algorithms, and customized workflows in the field of ocean artificial intelligence research, this paper proposes the design of a big data-driven ocean artificial intelligence service platform framework. It describes the overall architecture and the key technical methods for building the platform, and provides a case study of ocean artificial intelligence model research supported by the platform. [Results] The big data-driven ocean AI service platform provides a development environment for ocean AI models, supports functions such as data preprocessing, feature engineering, model training, hyperparameter tuning, model evaluation, model deployment, and model inference. It can assist ocean research scientists in building ocean artificial intelligence research workflows and in developing artificial intelligence model products for ocean scenario applications. [Conclusions] The big data-driven ocean artificial intelligence service platform, as a new type of scientific research information platform, will promote the cross-integration of marine science and artificial intelligence, promote the construction of artificial intelligence oceanography disciplines in IOCAS, and drive the paradigm shift in marine big data and artificial intelligence research.

Key words: transformation of research paradigms, ocean big data, artificial intelligence for oceanography, machine learning as a service, workflow management, model development