Abstract: The evaluation and prediction of the operation state of university scientific research platforms plays an important role in promoting the healthy and efficient development of scientific research work. However, the complexity of data indicators, complex logical relationships, and numerous influencing factors have greatly increased the difficulty. From the perspective of big data, a method based on GCA-DEA-MSVC is explored. First, the GCA method is used to mine and extract key feature indicators that are closely related to the evaluation results from the database and classify to build a feature indicator database. After that, the DEA method is used to fuse the feature index database data to improve the data quality and build a relative efficiency index database. Finally, the feature index library and the relative efficiency index library were re-fused, and an efficient classification and prediction model for the evaluation of the operating status of the scientific research platform is constructed based on the improved MSVC method. An experimental study is conducted using the evaluation data of the key laboratory of the Ministry of Education to verify the effectiveness of the proposed method.

Key words: scientific research platform; data envelopment analysis; multi-output support vector classification; evaluation and prediction