Overview

Cost overruns remain a major challenge in infrastructure projects worldwide, often leading to financial waste and resource mismanagement. This study introduces a machine learning–based approach using Artificial Neural Networks (ANNs) to analyze and predict cost performance in large-scale Norwegian road projects.

Methodology

• Dataset: 52 large-scale Norwegian road projects (each with > NOK 750 million budget).

• Features analyzed: region, project size, duration, number of contractors, tunnels, bridges, etc.

• Techniques used:

  Feature selection to identify key cost drivers.

  Two ANN models - Multi-Layer Perceptron (MLP) and Radial Basis Function Neural Network with Generalization Performance (RBFNN-GP).

  Compared to traditional linear regression and fuzzy regression models.

  Data gaps handled using Correlation-Based Regression Imputation for improved accuracy.

Key Findings

• RBFNN-GP outperformed all models, achieving high prediction accuracy:

      **Mean Absolute Percentage Error (MAPE):** 8.65% (training) and 9.81% (testing).
  

• MLP performed well but slightly less accurate (MAPE ≈ 12–15%).

• Traditional regression models showed significantly weaker accuracy (MAPE > 36%).

• Most influential factors affecting cost overruns:

  1. Final (actual) project cost
  2. Number of contractors
  3. Construction duration
  4. Cost deviation from quality assurance (QA2) reports

Conclusions

The study demonstrates that advanced ANN models can substantially improve cost prediction accuracy in infrastructure planning.

• These models help identify the key drivers of cost inaccuracy.
• They enable early-stage cost risk forecasting, supporting better decision-making and resource allocation.
• The RBFNN-GP model offers strong generalization even with small datasets, making it practical for real-world applications.

Industry Impact

• For policymakers and engineers: Provides a data-driven tool to reduce financial risk in large public works.
• For project managers: Enables smarter cost estimation and budgeting strategies.
• For investors and stakeholders: Enhances transparency and confidence in project feasibility studies.