Article Preview
Top1. Introduction
IT infrastructure maintenance includes all the tasks and necessary actions for their proper functioning. There are two main categories of IT maintenance: hardware and software one which are divided into several areas: application, virtualization, storage, server, network, security, systems…. Additionally, the problem can be started and treated from the simplest to the most complex. Moreover, it will take time to find the best solution depending on the user’s competence. These services can be provided on site or remotely in which they are conducted as corrective maintenance (Kent et al., 2017) or through preventive methods with fixed time (Mehmeti et al., 2018).Indeed, according to this context within IT infrastructure maintenance, several proprietary software has need appeared like: Nagios (Nagios IT management and monitoring product, 2007), Manage Engine (Manage Engine IT management and monitoring product, 2007) and Zenoss (Zenoss IT Infrastructure management and monitoring tool for hybrid IT environment, 2018). However, all these IT infrastructure solutions are preventive and limited when the problem is already occurred. Recently, the purpose of the maintenance task is to calculate the maintenance needs before the equipment fails, i.e. continuous monitoring (Callewaerta et al., 2017) with the aim to improve controls, processes and to prevent or detect fraud in an IT platform. Additionally, decision support systems have been developed to assist decision-makers in solving problems preventively and correctively thanks to supervised(Zenoss IT Infrastructure management and monitoring tool for hybrid IT environment, 2018) or unsupervised methods(Campos et al., 2007; Liao et al., 2013). Our study is part of corrective decision support systems (DSS) with semi-supervised approach to reduce the time of problem resolution. Unfortunately, the quality of the selected solution often depends on problem description quality and user knowledge level. In some situations, the problem description could be interpreted in different ways. In another situation, the pertinent terms do not appear in the situated problem. Nevertheless, these facts lead us unsatisfactory solutions. Therefore, we developed a text mining-based approach to explore, analyze all unstructured inputted data. Our aim is to provide automated support to the user and deliver new information to the system.
The rest of this paper is organized as follows: We first briefly review the different strategies for maintenance with some related works. Section three is devoted to the description of the proposed approaches by introducing our general architecture. In section four, we present the technics of terms extraction. Section five resumes the different strategies for text similarity measure with some corresponding related works. Section six describes our main idea about implemented extraction methods. Section seven experiments, evaluate, and discuss the obtained results during troubleshooting task. Finally, section eight summarizes the paper with some final remarks by pointing out possible directions for future works.