Automated Management System For Operation And Maintenance Tickets

Modern large-scale software systems are becoming increasingly complex,with a large number of components,services,and applications,making the system more sensitive to various types of failures.At the same time,the root cause of a failure may involve the interactions between multiple components and services,increasing the difficulty of troubleshooting.In order to accurately identify the root cause of a failure and propose corresponding solutions,engineers need to have advanced technical knowledge and rich experience.In recent years,with the rapid development of artificial intelligence,the application of machine learning technology to help ticket system quickly analyze the root causes of abnormal problems and timely recommend experts to solve the ticket has become one of the research hotspots.Machine learning technology cannot only process large amounts of data quickly and efficiently,but can also improve the accuracy and effectiveness of ticket automation.However,the traditional methods for expert recommendation in tickets are mainly based on historical ticket transfer relationships and ticket similarity to obtain the Top-N sequence of experts.These methods may lead to high similarity between the recommended experts in the sequence,so that the recommended experts are unable to solve the ticket,resulting in prolonged resolution time.In addition,the traditional classified-based machine learning method is difficult to find actual abnormal nodes through complex system node relationships when performing root cause analysis for tickets.In view of the above problems,this paper designs and develops a ticket automation management system.The main work contents are as follows:To address the similarity problem in the sequence of recommended experts,this paper introduces a neural network attention mechanism to construct an expert recurrent recommendation network.The network assumes that the currently recommended expert is unable to solve the ticket,and provides feedback based on the current recommendation results for the subsequent recommendation to avoid the next expert with high similarity to the current expert,reducing ineffective recommendations.Also,to further improve the accuracy of expert recommendations,this paper extracts features by converting historical ticket descriptions and solutions to improve the information representation of ticket abnormal descriptions.To address the problem of the difficultly in finding actual abnormal nodes in complex software systems,text processing model and graph model are used to extract the feature representation of system nodes to improve the accuracy of root cause analysis for tickets.By using the graph convolutional neural network’s ability to aggregate neighboring node information,the graph feature representations of system nodes are learned.Combining with the textual features of system nodes,the similarity between the description nodes of tickets and the corresponding graph nodes is calculated to match the abnormal nodes.Finally,based on the expert recommendation method and the root cause analysis method proposed in this paper,an ticket automation management system was designed and implemented.By adopting B/S architecture,the system can quickly provide abnormal root cause analysis results and recommend experts for the current ticket by calling the ticket root cause analysis model and the ticket expert recommendation model.