MACHINE LEARNING BASED PREDICTION ANALYSIS FOR THE SERVICE LEVEL AGREEMENT: A CASE STUDY OF DHANARAK ASSET DEVELOPMENT CO., LTD.

Abstract

            The purpose of this research is to analyze and compare the efficiency of the prediction models for the success of the service level agreement of Dhanarak Asset Development Co., Ltd by using building management system data. The researcher has used five techniques, including support vector machines, decision tree, k-nearest neighbor, perceptron and Adaboost algorithms to create 5 classification models and compare the performance of success prediction among those 5 different models. The results showed that Adaboot Algorithm gave the highest accuracy of 69%, support vector machine and decision tree equally yielded 67% accuracy, while the nearest neighbor and perceptron provided 66% and the 56% accuracies, respectively. Although the evaluating results of the five forecasting models gave a not very high accuracies, the forecasting models can actually be used in business by considering other factors, such as, risk analysis in preparing spare equipment for repairs, analysis of building deterioration, preparation of various resources in order to determine work plans to be successful according to service level agreements.