Machine learning supervised analysis for enhancing incident management process

Abstract

This paper presents the descriptions of the attributes which are composed of an enriched incident management process event log. The degree of accuracy of the models for prediction depends on the usefulness of the log attributes utilized in building such models are and by using a machine learning algorithm approaches to the attributes gives a better comprehension of the background (underlying) process. This paper studies the classification method that applies for deciding which best feature among the features in an enriched event log dataset for the incident management process. The classification method compared to other methods used in related work. The result will state which one algorithm is the highest and the best one to be selected. The selection of an attribute is vital to building or creating a model with completion time prediction capability by determining concept description features to be learned and ways of feature combination. In this paper, a classification method was applied. Depending on the feature-target feature association (correlation), every attribute is separately analyzed. The outcomes indicate that the technique used outperformed human expert’s decision making. We conducted predictions over different periods and achieved satisfactory performance in terms of accuracy, whereas, the best performance for classifying is achieved by the Bayes Net algorithm of 85.2760% accuracy.