Klasifikasi insiden pada aplikasi ServiceNow untuk menentukan keberhasilan Service Level Agreement

Abstract

The failure to achieve Service Level Agreement (SLA) has a negative impact on the satisfaction of the service that was promised. The SLA report that does not reach the target can have an impact on the promised service being not in accordance with what is paid by the business. On the other side, the service provider will have a bad reputation as a service provider company, and can even be fined for this. This study analyzed incidents using the CRISP-DM methodology and classification techniques with the Naive Bayes (NB), Logistic Regression (LR) and Support Vector Machine (SVM) algorithms which are most widely used in previous studies to predict SLA failures based on a review of studies. With help of the machine learning applications to do the data mining process, it is obtained that the SVM algorithm has the highest accuracy with a value of 75.05%, compared to LR 74.79% and NB 74.48%. The prediction prototypes are made with a programming language by adding incidents dataset, performing data reduction to balance data on target attribute, transforming data, and split training and testing dataset. This processes increasing the level of accuracy on the prototype to 83.32%.