# Risk-indexed artificial neural network for predicting duration and cost of irrigation canal-lining projects using survey-based calibration and python validation

## TL;DR

This paper introduces a machine learning model to predict the time and cost of irrigation canal lining projects by combining risk analysis and expert input.

## Contribution

The novel contribution is an integrated risk-indexed artificial neural network framework for infrastructure project forecasting.

## Key findings

- The model achieved an R² of 0.92 during training and 0.82 during testing.
- Average prediction errors were 0.87 months for time and EGP 102,500 for cost.
- The model was implemented as a Python desktop application for practical use.

## Abstract

The purpose of this study is to develop and validate a risk-driven predictive model for estimating project duration and cost in irrigation canal lining projects, where uncertainties often lead to delays and budget overruns. Ninety-three factors were first reduced to twenty using AHP–RII (Cronbach’s \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\alpha = 0.954$$\end{document}). A multi-layer perceptron (128–64–32, ReLU, Adam, early stopping) was trained on 5000 simulated scenarios and validated on eight projects with leave-one-project-out cross-validation. The model had \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$R^2 = 0.92$$\end{document} (training), 0.82 (testing), and made errors within the limits of 0.87 months (time) and EGP 102,500 (cost) on average.The developed model was deployed as a Python-based desktop application, enabling engineers and planners to generate accurate time and cost forecasts during early project stages. This research introduces an integrated ANN-based framework that combines expert-driven risk assessment with machine learning, providing a practical decision-support tool for infrastructure projects.