Abstract:

[Application Background] The earth big data sharing service system is the data portal for the Chinese Academy of Sciences “Earth Big Data Science Project”, a strategic pilot science and technology project. It provides global users with a data-sharing system integrating data, computing, and services, and promotes the new model for earth science data sharing. [Objective] With continuous release of data resources, it will become more difficult for users to obtain data resources through only filtering, and searching, etc. How to use recommendation technology to help users obtain scientific data more efficiently is a problem for research. [Methods] Therefore, this paper designs an earth science data recommendation model, ESDRec, which uses a bidirectional long-short-term memory network and attention mechanism to model users’ interest preferences, and calculates the correlation degree of metadata feature attributes of scientific data. This work incorporates domain features of earth science data into ESDRec so that ESDRec can generate more accurate recommendation results. [Conclusions] By conducting comparative experiments on the real datasets of the platform, this paper verifies the effectiveness of the ESDRec model.

Key words: recommendation system, scientific data sharing, earth big data, deep learning, recurrent neural network