🔥 Alberta Wildfire Prediction Model

Wildfires in Alberta pose a serious threat every year. This project uses public wildfire data to analyze patterns and predict the final severity of wildfires based on:

• 📅 Month
• 📍 Geographic location
• 🔥 Cause of wildfire
• 🌤️ Weather conditions

We hope this model can help relevant authorities allocate resources more effectively and respond in a timely manner.

---

📁 File Structure & Explanation

📘 0. Background & Case Study Folder

Contains public case studies and documents to:

• Understand the context of Alberta wildfires
• Justify why data analysis and modeling is necessary

---

📊 1. Datasets

• historic_data.csv / historic_data.xlsx
  📌 Raw data from the Alberta Government’s open data portal
  ⏳ Covers wildfire records from 2006–2023

• wildfire_data_dictionary.pdf
  📌 Official data dictionary from the public dataset website

• Data_prepration_cleansing_SQLquery.sql
  📌 SQL scripts used for data cleansing and import to Power BI - I did the same process in .py file in modeling, if you don't do the Powerbi dashborad, could skip it.

• Wildfire_data_dictionary.xlsx
  📌 Authored by this project
  📄 Includes:

  • Cleaned and derived fields
  • A list of features used for modeling

---

📊 2. Data Analysis

This section includes exploratory visualizations and dashboard reports designed to answer key business questions such as:

• Which regions are most susceptible to wildfires?
• Do the causes of wildfires differ significantly across regions?
• How do weather conditions (e.g., temperature, humidity) influence wildfire size?
• ……

📁 2.data analysis/

• Wildfire_Dashboard and Report.pptx
  👉 A Power BI dashboard/report addressing the business questions through visual insights. Ideal for business presentation use.

---

🤖 3. Modeling & Prediction

This section includes all modeling-related notebooks, trained models, and prediction outputs. Both LightGBM and XGBoost algorithms are used and compared for classification performance.

📁 3.model/

🧪 Notebooks

• 1.wildfire_size_prediction_preparation.ipynb

• 2.wildfire_size_prediction_model.ipynb
  ➤ Data preprocessing, feature engineering, and model training

• 3.wildfire_size_forecast_lightgbm.ipynb

• 3.wildfire_size_forecast_xgboost.ipynb
  ➤ Batch prediction using trained models on new user input

📦 Saved Model Files

File Name	Description
xgboost_model.pkl	Trained XGBoost model
lgbm_model.pkl	Trained LightGBM model
scaler.pkl	StandardScaler used for normalization
imputer.pkl	Imputer used for filling missing values
label_encoders.pkl	Encoders for categorical variables

📄 Input & Output Samples

File Name	Description
user_input.xlsx	Sample user input file (month, location, weather, etc.)
prediction_results_xgboost.xlsx	Prediction results using XGBoost model
prediction_results_lightgbm.xlsx	Prediction results using LightGBM model

---

🔧 Python Libraries Used

Library	Description
pandas	Data loading, transformation, and manipulation
numpy	Numerical computations and array operations
matplotlib	Visualization of charts and graphs
seaborn	Statistical data visualization
scikit-learn	Preprocessing, model training, evaluation utilities
imblearn	SMOTE: oversampling for class imbalance
lightgbm	LightGBM classifier (fast gradient boosting)
xgboost	XGBoost classifier for tree-based modeling
joblib	Saving/loading models and encoders
openpyxl	Reading/writing Excel files
pycaret	AutoML for model comparison and tuning
flask	Building a lightweight prediction web interface

or Install all dependencies with:

pip install -r requirements.txt

---

🌐 4. Flask App (Prediction System)

To make the model easily accessible to non-technical users, we built a simple Flask web app that allows users to upload data and get wildfire severity predictions through a user-friendly interface.

📁 4.my_flask_app/

🧩 Key Components:

File / Folder	Description
output_model.py	Main Flask script handling file upload, model loading, and prediction logic
templates/	Folder for HTML templates (e.g., form input page)
xgboost_model.pkl	Trained model used for prediction
scaler.pkl	StandardScaler used during preprocessing
imputer.pkl	Imputer for handling missing values
label_encoders.pkl	Encoded categorical fields
lgbm_model.pkl	Optional: alternative model for evaluation or switching

▶️ Usage

Run the Flask app locally:

cd 4.my_flask_app/
python output_model.py

Then open your browser and go to http://127.0.0.1:5000/（or the address showed in your command prompt windows) to upload your input and get the result.

---

📄 5. Final Report & Demo

📁 5.final report/

This folder contains the final presentation materials, including the executive summary, visualizations, and a demo walkthrough of the full system.

• Final Report_Wildfire Prediction Project.pdf
  📄 Coverted from the slides. Final project report summarizing the case study, data analysis, modeling process, and key recommendations

• Prediction Model Using Introduction Video.pdf
  🎥 This PDF contains screenshots and a linked video demo (page 26), which walks through the local model execution and web app usage, including how to upload a file and receive predictions.

---

📌 This complete workflow demonstrates how the wildfire prediction system can be applied in real-world scenarios, from data to decision-making.

🧠 If you use this project, please remember to cite the original repository. Attribution matters! ✅

📬 Contact

For questions or collaboration inquiries, feel free to reach out:

• 📧 Email: gekx1017@gmail.com
• 🔗 LinkedIn: Kaixiang (Kai) Ge