Leveraging Support Vector Machine for Predictive Analysis of Earned Value Performance Indicators in Iraq's Oil Projects

Abstract

Although earned value management (EVM) offers considerable advantages for schedule and cost control within oil projects, its implementation as a project control technique remains limited in Iraq. This study primarily aims to establish predictive models, utilizing support vector machine (SVM), to estimate earned value performance indicators namely schedule performance index (SPI), cost performance index (CPI), and to-complete cost performance index (TCPI) within the context of Iraqi oil projects. The dataset, encompassing 83 monthly reports spanning from 26th June 2015 to 25th August 2022, was sourced from the Karbala Refinery Project. This project, managed by the oil projects company (SCOP) under the Iraqi Ministry of Oil, represents one of the largest and most contemporary initiatives within the region. The results revealed significant findings, including an average accuracy (AA%) for CPI, SPI, and TCPI of 96.093%, 91.709%, and 66.024%, respectively. Correlation coefficients (R) were registered at 92.8%, 98.2%, and 93.3%, while the root mean squared error (RMSE) stood at 0.0969, 0.0604, and 0.2260 respectively. In conclusion, the SVM technique was employed in this study to derive predictive models, yielding superior accuracy for earned value indexes.