Performance-Based Cost and Time Prediction Analysis of Tertiary Irrigation Projects Using Artificial Neural Network

Abstract

The LOS-04 project, a tertiary irrigation work package of the Rentang Irrigation Modernization Project (RIMP), has been ongoing for 26 months with a physical progress of 60.95%. This study aims to analyze project performance and predict the final cost and completion time by integrating Earned Value Management (EVM) with Artificial Neural Network (ANN). The EVM analysis revealed a Cost Performance Index (CPI) of 0.975 and a Schedule Performance Index (SPI) of 0.687, indicating that the project is experiencing cost overruns and significant schedule delays. To improve forecasting accuracy, an ANN model was developed using historical project data. The integration of EVM and ANN predicts a final project cost of Rp. 150.2 billion and a total duration of 52 months. These findings provide critical insights for project monitoring and evaluation, allowing stakeholders to formulate necessary corrective actions to mitigate further risks.

Conclusion

Based on the project performance analysis and the predictive modeling conducted using the 26-month dataset, the following conclusions are drawn: (1) Based on the analysis up to the 26th month, the project is experiencing significant deviations from the baseline plan. The physical realization only reached 60.95% against the target of 88.73%. This delay is quantified by the Schedule Performance Index (SPI) of 0.69, indicating poor time performance, and a Cost Performance Index (CPI) of 0.97, indicating mild cost inefficiency. (2) Despite the constraints of limited data characteristic of an ongoing project, the ANN model demonstrated high reliability. The validation process yielded a correlation coefficient (R) of approximately 0.99 with low RMSE values, confirming that the model effectively captures performance patterns without overfitting. (3) Both the ANN model and standard EVM method predict a critical delay, estimating the total duration to reach 52 months (a 16-month delay). However, regarding costs, the ANN model proves to be more sensitive to risk. While the standard EVM projected a final cost of Rp 148.6 Billion, the ANN estimated a higher requirement of Rp 150.2 Billion. This difference highlights the ANN's capability to provide a more realistic "early warning" by capturing non-linear cost fluctuations that linear EVM calculations might miss. (4) Based on these findings, corrective measures are urgently needed. It is recommended that stakeholders process a contract addendum for a 16-month extension and prepare a budget adjustment of approximately Rp 5.3 Billion to prevent further cost overruns.

References