Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🏗️ Predictive Project Risk Intelligence Platform

Overview

The Predictive Project Risk Intelligence Platform is an end-to-end machine learning solution designed to identify construction projects at high risk of schedule delays and cost overruns before construction begins.

The project simulates how an architectural or construction consultancy can use predictive analytics to support project planning, improve decision-making, and reduce financial risk.


Business Problem

Construction projects frequently exceed budgets and miss planned completion dates due to risks being identified too late.

Traditional project reporting is reactive, highlighting issues only after they have occurred.

This project demonstrates how machine learning can proactively identify high-risk projects during the planning stage.


Solution

This platform predicts:

  • Project delay probability
  • Cost overrun probability
  • Overall project risk score (0–100)

The solution combines predictive modelling, feature engineering, SQL business analysis, and an interactive Streamlit dashboard to support project managers and executives.


Tech Stack

  • Python
  • Pandas
  • NumPy
  • Scikit-learn
  • XGBoost
  • SQL
  • Streamlit
  • Plotly
  • Git
  • GitHub

Repository Structure

Predictive-Project-Risk-Intelligence-Platform

├── dashboard/
├── data/
│   ├── raw/
│   └── processed/
├── images/
├── models/
├── notebooks/
├── sql/
├── src/
├── README.md
├── requirements.txt
├── LICENSE
└── .gitignore

Dataset

A synthetic dataset of over 1,000 construction projects was created to simulate real-world project delivery data.

Each project includes:

  • Project type
  • Region
  • Budget
  • Planned duration
  • Contractor experience
  • Material inflation
  • Weather risk
  • Labour availability
  • Site complexity
  • Client changes
  • Permit delays
  • Supplier distance
  • Safety history

Target variables:

  • Delay
  • CostOverrun

Feature Engineering

Several business-focused features were engineered to improve predictive performance, including:

  • BudgetPerDay
  • InflationImpact
  • SupplierRisk
  • ContractorRisk
  • ClientRisk
  • ComplexityScore

Machine Learning

Two independent XGBoost classification models were developed.

Model 1

Predicts project delay.

Model 2

Predicts project cost overrun.

Model outputs include:

  • Prediction
  • Probability
  • Risk Score

Project Risk Score

Overall project risk is calculated using:

Risk Score = (0.6 × Delay Probability) + (0.4 × Cost Overrun Probability)

Risk Categories

Score Category
0–30 Low
31–60 Medium
61–80 High
81–100 Critical

Streamlit Dashboard

The dashboard provides:

  • Executive KPIs
  • Project filtering
  • Regional risk analysis
  • Contractor risk analysis
  • Risk driver analysis
  • Project Risk Explorer
  • Budget vs Risk visualisation
  • Delay vs Cost Overrun analysis

SQL Analysis

Example business questions answered:

  • Which regions have the highest average project risk?
  • Which contractors consistently underperform?
  • Which projects have the highest predicted risk?
  • What is the average project budget by project type?

Business Value

The platform enables organisations to:

  • Identify high-risk projects before construction begins
  • Improve budgeting decisions
  • Reduce schedule delays
  • Improve contractor selection
  • Support data-driven project governance

Future Improvements

Future versions may include:

  • SHAP explainability
  • Live database integration
  • Power BI dashboard
  • Geospatial project risk mapping
  • Real-time prediction API

Author

Saira Akbar

Machine Learning | Data Science | Project Risk Analytics

About

Machine learning platform predicting construction project delays and cost overruns using XGBoost, feature engineering, risk scoring, and Streamlit analytics.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages