Research On Fault Text Multi-label Classification Of ICT System Based On Ensemble Learning

With the rapid development of the Internet and the advent of the information age,the data volume of the power grid is growing explosively,which makes the load pressure of the State Grid Information System increasing.State Grid Information System involves a wide range of complex levels.However,the fault of information system is inevitable,and its harm is incalculable.Therefore,it is a work of great significance to find a fast response and timely processing method for faults.In the era of rapid development of artificial intelligence technology,mining key information from massive data can make information system fault study technology more intelligent.State Grid Corporation ICT System has accumulated a large number of information system fault repair records in daily work.Most of these records are unstructured descriptive text data,which is difficult to classify and analyse by automated tools.Moreover,it has been observed that the occurrence of a ICT System fault may be the result of a number of multiple different causes.Aiming at features of ICT System fault reports,a multi-label text classification model based on the ensemble learning algorithm is proposed for fault-assisted decision.Firstly,starting from the text mining technology,we extract the cause categories and the fault problem description to convert a many-to-one causal relationship.Then we segment words,remove stop-word,vectorize for these texts.Finally,a classification method by combining Binary Relevance with the ensemble learning Gradient Boosting algorithm is adopted.The former is used for problem transformation,and the multi-label problem is transformed into a single-label problem.The latter is used for iterative training,and the accuracy is improved by gradient boosting.Experimental results show that this method is better than Binary Relevance based on Logistic Regression and ML-kNN for fault text classification.