🏗️ ML-Based Class 5 Construction Cost Estimator

Predicting Early-Stage Construction Costs with Machine Learning

CSE6748 - Applied Analytics Practicum
Georgia Institute of Technology | Fall 2025

🚀 Quick Start • 📊 Features • 📖 Documentation • 👥 Team

🎯 Achievement: 21.97% MAPE - Exceeds Target by 3.03%

🎯 Overview

A production-ready web application that leverages Random Forest machine learning to predict early-stage construction costs with exceptional accuracy. Built for Construction Cost Database LLC, this tool provides Class 5 estimates (±25% accuracy) for infrastructure projects.

🏆 Key Achievement

Target Requirement: MAPE < 25% Actual Performance: MAPE = 21.97% Result: ✅ Exceeded target by 3.03%

🎓 Academic Context

Course: CSE6748 - Applied Analytics Practicum
Institution: Georgia Institute of Technology
Semester: Fall 2025
Client: Construction Cost Database LLC
Dataset: 17,025 historical projects (2010-2025)

✨ Key Features

📊 Analytics

Real-time exploratory data analysis with interactive visualizations

Features:

Cost distribution charts
Geographic heatmaps
Project type analysis
Statistical insights

💰 Cost Prediction

Instant ML-powered cost estimates with confidence intervals

Features:

13-feature input form
±25% confidence range
Similar project matching
Detailed breakdowns

📈 Model Insights

Comprehensive model performance tracking and comparison

Features:

4 algorithm comparison
Feature importance
Learning curves
Residual analysis

🎯 Model Performance

Current Production Model: Random Forest

Metric	Value	Status
Test MAPE	21.97%	✅ Target Met
R² Score	0.9463	🎯 Excellent
Test MAE	$271,543	📊 Strong
Test RMSE	$412,583	📈 Reliable
Dataset Size	17,025 projects	📦 Large Scale

Model Comparison Results

Model	CV MAPE	Test MAPE	R² Score	Status
🌲 Random Forest	23.30%	21.97% ✅	0.9463	🚀 Deployed
⚡ XGBoost	36.16%	36.37%	0.9258	📋 Alternative
💡 LightGBM	36.93%	37.62%	0.9232	📋 Alternative
📊 Gradient Boosting	42.48%	43.75%	0.9015	⚠️ Above Target

Why Random Forest?

✅ Best MAPE performance (21.97%)
✅ Highest R² score (0.9463)
✅ Excellent interpretability
✅ Robust to outliers
✅ Fast training and prediction

🛠️ Technology Stack

Backend

🐍 Python 3.11+         - Core programming language
🌶️ Flask 3.0.0          - Web framework
🤖 scikit-learn 1.3.2   - Machine learning
🐼 Pandas 2.1.4         - Data manipulation
🔢 NumPy 1.26.2         - Numerical computing
📊 Matplotlib 3.8.2     - Visualizations
🎨 Seaborn 0.13.0       - Statistical plots

Frontend

📄 HTML5 / CSS3         - Modern web standards
🎨 Bootstrap 5          - Responsive UI framework
📊 Chart.js             - Interactive charts
⚡ Vanilla JavaScript   - Dynamic interactions

Machine Learning Pipeline

🌲 Random Forest        - Primary algorithm
📏 StandardScaler       - Feature normalization
🏷️ OneHotEncoder        - Categorical encoding
🎯 K-Means Clustering   - Geographic regions
✅ 5-Fold CV            - Model validation

🚀 Quick Start

Prerequisites Checklist

Python 3.11 or higher installed
pip package manager available
500MB free disk space
Modern web browser

One-Command Setup

# Clone, setup, and run in one go
git clone https://github.com/kraryal/construction_new.git && \
cd construction_new && \
python -m venv venv && \
source venv/bin/activate && \
pip install -r requirements.txt && \
python app.py

Windows Users: Replace source venv/bin/activate with venv\Scripts\activate

Access the Application

🌐 Open browser: http://localhost:5000

That's it! 🎉

📦 Installation Guide

Step 1: Clone Repository

git clone https://github.com/kraryal/construction_new.git
cd construction_new

Step 2: Create Virtual Environment

Windows:

python -m venv venv
venv\Scripts\Activate.ps1

Mac/Linux:

python3 -m venv venv
source venv/bin/activate

Step 3: Install Dependencies

pip install --upgrade pip
pip install -r requirements.txt

Step 4: Verify Installation

python -c "import flask, pandas, sklearn; print('✅ Setup complete!')"

Step 5: Prepare Data

# Ensure dataset is in correct location
ls data/base_data_for_model.csv

Step 6: Run Application

python app.py

Expected Output:

================================================================================
🏗️  ML-BASED CLASS 5 CONSTRUCTION COST ESTIMATOR
================================================================================

✅ Dataset loaded successfully: 17025 projects with 38 features
✅ Model loaded successfully
✅ System Ready!
🌐 Access at: http://localhost:5000
================================================================================

💡 Usage Examples

Web Interface

Navigate to Cost Estimator page
Fill project details form
Click "Calculate Cost Estimate"
View prediction with confidence interval

Input Example

Field	Value
Project Type	Pavement Markers
Budget Range	$3M-$6M
Complexity	Category 4
State	Michigan (MI)
County	Alcona County
Area Type	Rural
Inflation Factor	1.05
ACF	1.01

Output

Estimated Cost: $4,358,432.11
Confidence Range: $3,268,824 - $5,448,040
Similar Projects: 6,147 found

API Usage (Python)

import requests

# Endpoint
url = 'http://localhost:5000/estimate_cost'

# Project data
data = {
    'inflation_factor': 1.05,
    'official_budget_range': '$3M-$6M',
    'ciqs_complexity_category': 'Category 4',
    'cnt_division': 6,
    'cnt_item_code': 6,
    'county_name': 'Alcona County',
    'area_type': 'Rural',
    'acf': 1.01,
    'project_type': 'Pavement Markers',
    'project_category': 'Civil',
    'project_state': 'MI',
    'region': 'Region_3'
}

# Make request
response = requests.post(url, data=data)
result = response.json()

# Display results
if result['success']:
    print(f"💰 Estimated Cost: {result['estimated_cost_formatted']}")
    print(f"📊 Confidence Range: {result['confidence_interval']['lower_formatted']} - {result['confidence_interval']['upper_formatted']}")
    print(f"🔍 Similar Projects: {result['similar_projects']['count']}")

cURL Example

curl -X POST http://localhost:5000/estimate_cost \
  -d "inflation_factor=1.05" \
  -d "official_budget_range=\$3M-\$6M" \
  -d "ciqs_complexity_category=Category 4" \
  -d "cnt_division=6" \
  -d "cnt_item_code=6" \
  -d "county_name=Alcona County" \
  -d "area_type=Rural" \
  -d "acf=1.01" \
  -d "project_type=Pavement Markers" \
  -d "project_category=Civil" \
  -d "project_state=MI" \
  -d "region=Region_3"

📁 Project Structure

construction_new/
│
├── 📄 app.py                       # Main Flask application (500+ lines)
├── 📋 requirements.txt             # Python dependencies
├── 📖 README.md                    # Project documentation (this file)
├── 📚 INSTRUCTIONS.md              # Detailed setup guide
│
├── 📂 data/
│   └── 📊 base_data_for_model.csv # Training dataset (17,025 projects)
│
├── 📂 models/
│   ├── 🤖 construction_cost_model.pkl  # Trained Random Forest model
│   └── 📈 model_metrics.json           # Performance metrics
│
├── 📂 templates/                   # HTML templates
│   ├── 🏠 home.html               # Landing page with cards
│   ├── 📊 eda.html                # Exploratory data analysis
│   ├── 📈 model_comparison.html   # Algorithm comparison
│   ├── 🎯 dashboard.html          # Performance dashboard
│   ├── 💰 cost_estimator.html     # Prediction form (main feature)
│   ├── 🗂️ data_overview.html      # Dataset information
│   ├── 📚 documentation.html      # API docs & team info
│   ├── 🧭 base.html               # Base template with navigation
│   └── ❌ error.html              # Error handling page
│
└── 📂 static/
    ├── 🎨 css/
    │   └── styles.css             # Custom styles
    └── 🖼️ images/                  # Generated plots & assets

🔬 Model Details

Features Used (13 Features)

📊 Click to expand feature list

Economic Factors (2)

Inflation Factor - Range: 1.00 - 1.34 | Adjusts for year-over-year cost changes
Area Cost Factor (ACF) - Range: 0.80 - 1.19 | Geographic cost adjustment multiplier

Project Classification (4)

Project Type - Categorical | Specific construction work type (e.g., Pavement Markers)
Project Category - Categorical | General classification (e.g., Civil, Water & Sewer)
CIQS Complexity Category - Category 1-4 | Complexity rating from simple to complex
Official Budget Range - Categorical | Budget bracket (e.g., $3M-$6M, Less than 1M)

Geographic Location (4)

Project State - 50 US states | Location identifier
County Name - Varies by state | Specific county location
Area Type - Urban/Rural | Development density classification
Region - Region_0 to Region_3 | K-Means clustered geographic zones

Construction Details (3)

CNT Division Code - Range: 1 - 29 | Construction division taxonomy
CNT Item Code - Range: 1 - 61 | Specific item classification

Training Pipeline

graph LR
    A[Raw Data<br/>17,025 projects] --> B[Data Cleaning<br/>Fill missing values]
    B --> C[Feature Engineering<br/>K-Means clustering]
    C --> D[Train/Test Split<br/>80/20]
    D --> E[Preprocessing<br/>StandardScaler + OneHotEncoder]
    E --> F[Model Training<br/>Random Forest]
    F --> G[Validation<br/>5-Fold CV]
    G --> H[Production Model<br/>21.97% MAPE]

Preprocessing Steps

Missing Value Imputation
- Numerical: Median
- Categorical: Mode
Feature Engineering
- K-Means clustering for geographic regions
- Created 4 regional clusters from state coordinates
Feature Scaling
- StandardScaler for numerical features
- OneHotEncoder for categorical features
Train/Test Split
- 80% training (13,620 samples)
- 20% testing (3,405 samples)
- Random state: 42 (reproducible)

Model Configuration

RandomForestRegressor(
    n_estimators=100,     # Number of decision trees
    random_state=42,      # Reproducibility seed
    n_jobs=-1            # Use all CPU cores
)

📚 API Documentation

Endpoints

1. Cost Estimation

POST /estimate_cost

Request Body (Form Data):

{
  "inflation_factor": 1.05,
  "official_budget_range": "$3M-$6M",
  "ciqs_complexity_category": "Category 4",
  "cnt_division": 6,
  "cnt_item_code": 6,
  "county_name": "Alcona County",
  "area_type": "Rural",
  "acf": 1.01,
  "project_type": "Pavement Markers",
  "project_category": "Civil",
  "project_state": "MI",
  "region": "Region_3"
}

Success Response (200):

{
  "success": true,
  "estimated_cost": 4358432.11,
  "estimated_cost_formatted": "$4,358,432.11",
  "confidence_interval": {
    "lower": 3268824.08,
    "upper": 5448040.14,
    "lower_formatted": "$3,268,824.08",
    "upper_formatted": "$5,448,040.14"
  },
  "similar_projects": {
    "count": 6147,
    "avg_cost_formatted": "$1,310,815.37",
    "median_cost_formatted": "$856,470.56",
    "match_type": "exact"
  },
  "timestamp": "2025-11-25 14:30:22"
}

2. Dataset Statistics

GET /api/dataset_stats

Response:

{
  "total_projects": 17025,
  "avg_cost": 1142356.78,
  "median_cost": 856470.56,
  "min_cost": 10500.00,
  "max_cost": 15200000.00
}

3. Model Metrics

GET /api/model_metrics

Response:

{
  "test_mape": 21.97,
  "r2_score": 0.9463,
  "mae": 271543.90,
  "rmse": 412583.00,
  "n_features": 13
}

📸 Screenshots

🖼️ Click to view screenshots

Home Page

Cost Estimator Form (Empty)

Clean input form for entering construction project details.

Cost Estimator Form (Filled Example)

Example of the form with sample data entered for cost prediction.

Model Comparison

Comparison of different machine learning models' performance metrics.

Performance Dashboard

Interactive dashboard showing key project analytics and insights.

👥 Team

Project Contributors

Krishna Aryal
_{Data Engineering & Model Development}
📧 Email • 💻 GitHub

Kumar Sawan
_{Feature Engineering & Optimization}
📧 Email

Neema Kafwimi
_{Model Evaluation & Deployment}
📧 Email

🙏 Acknowledgments

Construction Cost Database LLC - Dataset provider and project client
Georgia Tech CSE6748 - Course faculty and teaching assistants
scikit-learn Community - Open-source ML library
Flask Team - Web framework development
Stack Overflow Community - Problem-solving support

📝 License & Citation

This project is part of an academic practicum for Georgia Institute of Technology.

Citation

If you use this work, please cite:

@misc{construction_cost_estimator_2025,
  title={ML-Based Class 5 Construction Cost Estimator},
  author={Aryal, Krishna and Sawan, Kumar and Kafwimi, Neema},
  year={2025},
  institution={Georgia Institute of Technology},
  course={CSE6748 - Applied Analytics Practicum}
}

⚠️ Important Disclaimers

Class 5 Estimates Only
This model provides conceptual estimates with ±25% accuracy. Not suitable for detailed bidding, final estimates, or contractual commitments.

Historical Data Limitation
Model trained on 2010-2025 data. May not capture unprecedented market conditions, novel construction methods, or future trends.

Professional Validation Required
Always validate estimates with construction professionals and adjust for project-specific factors not captured by the model.

🔮 Future Enhancements

Real-time cost index updates
User authentication & project history
Export to PDF/Excel
Mobile responsive improvements
Additional ML models (Neural Networks)
Integration with external cost databases
API rate limiting & authentication
Multi-language support

📞 Support & Contact

Need Help?

📧 Email: karyal@gatech.edu
📖 Documentation: Read the docs

Project Links

🌐 Live Demo: [Coming Soon]
📊 Dataset: PCS Historical Project Database
🎓 Course: CSE6748 - Applied Analytics Practicum

⭐ Star this repository if you find it helpful!

Built with ❤️ by the Georgia Tech Team

"Accurate Early-Stage Cost Estimation Powered by Machine Learning"

🏠 Home • 📊 Dashboard • 💰 Estimate • 📚 Docs

```

Name		Name	Last commit message	Last commit date
Latest commit History 40 Commits
__pycache__		__pycache__
data		data
models		models
report_images		report_images
static		static
templates		templates
venv		venv
Alternate Model_Clustering.ipynb		Alternate Model_Clustering.ipynb
README.md		README.md
app.py		app.py
data_prep_and_eda.ipynb		data_prep_and_eda.ipynb
lookup_counties.csv		lookup_counties.csv
make_plots.py		make_plots.py
organised_construction_notebook.ipynb		organised_construction_notebook.ipynb
requirements.txt		requirements.txt
sql_output_all.csv		sql_output_all.csv

kraryal/construction_new

Folders and files

Latest commit

History

Repository files navigation

🏗️ ML-Based Class 5 Construction Cost Estimator

Predicting Early-Stage Construction Costs with Machine Learning

🎯 Achievement: 21.97% MAPE - Exceeds Target by 3.03%

📋 Table of Contents

🎯 Overview

🏆 Key Achievement

🎓 Academic Context

✨ Key Features

📊 Analytics

💰 Cost Prediction

📈 Model Insights

🎯 Model Performance

Current Production Model: Random Forest

Model Comparison Results

🛠️ Technology Stack

Backend

Frontend

Machine Learning Pipeline

🚀 Quick Start

Prerequisites Checklist

One-Command Setup

Access the Application

📦 Installation Guide

Step 1: Clone Repository

Step 2: Create Virtual Environment

Step 3: Install Dependencies

Step 4: Verify Installation

Step 5: Prepare Data

Step 6: Run Application

💡 Usage Examples

Web Interface

API Usage (Python)

cURL Example

📁 Project Structure

🔬 Model Details

Features Used (13 Features)

Economic Factors (2)

Project Classification (4)

Geographic Location (4)

Construction Details (3)

Training Pipeline

Preprocessing Steps

Model Configuration

📚 API Documentation

Endpoints

1. Cost Estimation

2. Dataset Statistics

3. Model Metrics

📸 Screenshots

Home Page

Cost Estimator Form (Empty)

Cost Estimator Form (Filled Example)

Model Comparison

Performance Dashboard

👥 Team

Project Contributors

🙏 Acknowledgments

📝 License & Citation

Citation

⚠️ Important Disclaimers

🔮 Future Enhancements

📞 Support & Contact

Need Help?

Project Links

⭐ Star this repository if you find it helpful!

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages