[Purpose/significance] In big data age, it is of great significance to construct an effective public opinion evaluation method based on the data from social media, and such method could contribute to network ecological governance and a healthy network development. [Method/process] Based on the information ecology theory, this paper constructs a social -network-based public opinion evaluation method using natural language processing techniques such as machine learning, sensitive judgment and keyword extraction. In the process of data processing, multiple methods were used to detect public sentiment so as to provide supporting evidence for evaluating the proposed indicator system. Such as, adaptive learning based on Adaboost, creating classifiers by using difference method and feature set, as well as statistical-based and rule-based sentiment analysis. On the practical level, the current paper selects several representative regions in the northeast, coastal, and western regions of China to evaluate regional ecological characteristics under the proposed evaluating method. [Result/conclusion] The construction of this evaluation method could provide
guiding for government, websites and netizens to work together and develop the social network space. It could also provide both theoretical and practical basis for future research in social network knowledge graph construction and network regulation strategy studies.