A feature mining method for power production safety hidden danger based on wavelet multi-window spectrum extraction

Power production safety hidden danger data is difficult to accurately identify, which leads to the inability to predict and prevent potential safety hazards. Moreover, the format of relevant laws and safety regulations in the power industry is different, and the original text information cannot be directly used for processing and analysis. Therefore, it is urgent to develop a preprocessing method for power production safety-related text information fusion cleaning. This paper proposes a feature mining method for power production safety hidden danger based on wavelet multi-window spectrum extraction.The crawling technology is used to crawl non-structured hidden danger data in the power production safety system, and NLP segmentation system is used to segment hidden danger texts. A database model is constructed to establish a high-quality sample data set with minimal cost and minimal sample deviation. The rough concept lattice generation algorithm is used for frequent access feature partitioning of data streams. And then, by constructing the wavelet function, the hyperbolic frequency mother wavelet of the hidden danger feature library of power production is obtained. The multi-window spectrum features of the database information flow are extracted to achieve feature mining of power production safety hidden dangers. Simulation results show that this method can effectively reflect the information characteristics of power safety hazard data and provide a basis for subsequent knowledge graph construction.