sanjibkuanr/pm25-pollution-predictor

PM2.5 Air Pollution Prediction Model 🌫️

This project predicts the level of air pollution (PM2.5 concentration) using historical environmental data collected from Beijing between 2010 and 2014. It uses a machine learning model trained on weather and pollution-related features.

📊 Dataset

• Source: UCI Machine Learning Repository
• Data File: PRSA_data_2010.1.1-2014.12.31.csv
• Features Used:
  • Temperature
  • Dew Point
  • Pressure
  • Wind direction (CBWD)
  • Cumulated wind speed (Iws)
  • Cumulated hours of snow (Is)
  • Cumulated hours of rain (Ir)

🧠 Model

• Type: Random Forest Regressor
• Framework: Scikit-learn
• Target Variable: PM2.5 concentration
• Evaluation: R² Score, Mean Squared Error (MSE)

📁 Files

• pm25_model.pkl: Trained ML model
• README.md: Project documentation
• pm25_predict.py: Python script for inference (optional)

🚀 Usage

You can use this model with the following steps

import pandas as pd import joblib from huggingface_hub import hf_hub_download

Download the model

repo_id = "sanjibkuanr/pm25-pollution-predictor" filename = "pm25_model.pkl" model_path = hf_hub_download(repo_id=repo_id, filename=filename) model = joblib.load(model_path)

Check model's expected feature names

expected_features = model.feature_names_in_ print("Model expects features:\n", expected_features)

Prepare only the required features for input

sample_input = pd.DataFrame([{ "dewp": -21, "temp": -12, "pres": 1020, "iws": 2.0, "is": 0, "ir": 0, "cbwd_NW": 1, "cbwd_SE": 0, "cbwd_cv": 0, "no": 100 }])

Select only columns the model expects

sample_input = sample_input[expected_features]

Predict

prediction = model.predict(sample_input) print("Predicted PM2.5 level:", prediction[0])

Developed by Sanjib Kuanr as part of a Machine Learning learning initiative. Feel free to connect with me on LinkedIn! You are free to use, modify, and distribute.