Performance Evaluation of Machine Learning Techniques on Resolution Time Prediction in Helpdesk Support System

Abstract

Estimating incident resolution times accurately is critical to maintaining an effective resource allocation for customer service. In order to meet this need, this paper explores machine learning techniques widely applied in the Resolution Time Prediction and identify the performance of chosen approaches via benchmarking dataset. The proposed method starts with data preprocessing, such as removing outliers and missing values and determining any irregularities in the resolution times distribution. Subsequently, we automatically choose the most relevant features using various statistical techniques. As the last stage of our prediction pipeline, we will apply different machine learning approaches the dataset to find the effectiveness of model and conclude the best technique based on the model accuracy and model fitting time. By applying this strategy, we hope to gain a better understanding of the factors affecting incident resolution times, which will eventually result in better resource allocation and planning for customer support operations.