Knowledge Graph Construction Method for Mobile IOT Management Platform Device Anomaly

Mobile IOT management platform equipment abnormal information contains a large amount of data knowledge, and it provides decision support for intelligent operation and maintenance of power communication equipment. However, such data are mostly stored as unstructured data and cannot be directly extracted for inferential decision making. In this paper, with the advantage of knowledge graph in data storage and knowledge extraction, we propose Multi-feature entity relationship joint extraction methods (MFERJE), which firstly generates word vectors by using ALBERT word embedding model, and then realizes word vectors by using BiLSTM structure and attention mechanism to achieve multi-feature entity recognition, and finally discriminate multi-type entity relationship based on graphical matrix to obtain the entities and relationships contained in abnormal data, using location, entity label and lexicality as feature inputs to greatly reduce the loss of semantic information features, achieve effective utilization of equipment abnormal data, and achieve equipment abnormal information based on the knowledge of equipment abnormal fault alarm visualization of equipment abnormal information based on equipment abnormal fault warning knowledge. The experimental results show that the F1 value of the model reaches 0.78, the accuracy is improved by 4% compared with other traditional methods, and the knowledge extraction effect is better, which can effectively improve the efficiency of equipment maintenance, realize the accurate positioning of equipment abnormalities and the rapid retrieval of corresponding maintenance strategies, and promote the intelligent operation and maintenance of mobile IOT management platform.