The Incident Analysis Research In Employee Benefit System Supporting

The objective of this design is to automate the IT incident troubleshooting process,instead of a non-standardized,experience-based,manual process.It bases on the records of historical incident descriptions and the corresponding finalized root causes,uses NLP and deep learning algorithms to build model,which is to assist the Product Supporter in locating the root cause of new incident reports.The innovation is to utilize the configuration difference between staging and production system environment in the feature engineering,to increase the accuracy and efficiency of this model.1.Build an efficient configuration comparison solution,which can automatically retrieve the latest client environment information.The framework is independent of business expansion and client scale change.2.Use NLP and deep learning,to process all historical incident records,including ETL,phase engineering,vectorization,and modeling.3.Vectorize the configuration difference of the environments,and combine with the incident description vector in the modeling feature engineering.4.Use several deep learning algorithms,including random forest,CNN,LSTM, GRU,RNN,RCNN,etc.,and optimize the parameters,to choose the best solution basing on the modeling accuracy.5.Build a light weighted UI webpage and interface to issue tracking system,the incident tracking system,to retrieve the latest incident report and use the model to determine the root cause.6.Test functionality of bridging the configuration difference in root causing,to prove this design could meet the objective by the test result.Eventually,this solution leads the change from 20 minutes experience-based manual process to 10 seconds automatical root,causing result delivery.It improves the efficiency significantly,cuts the labor cost,and creates a self-learning ecosystem to assist Production Support to deliver efficiency and accuracy.