🌫️ Air Quality Forecasting Using Machine Learning

Time-series forecasting of air quality using Facebook Prophet and real-world environmental data.

This project analyzes historical air quality sensor data and builds a machine learning model to forecast future air quality trends. The goal is to demonstrate time-series preprocessing, feature engineering, forecasting, and visualization using Python.

---

🧠 Skills & Technologies Demonstrated

Programming & Libraries

• Python
• NumPy
• Pandas
• Matplotlib
• Prophet (Facebook/Meta)

Core Concepts

• Data cleaning & preprocessing
• Handling missing values
• Time-series forecasting
• Exploratory Data Analysis (EDA)
• Data visualization
• Predictive modeling

---

📊 Dataset

Dataset used: Air Quality UCI Dataset

The dataset contains hourly air quality measurements including:

• CO(GT)
• NOx(GT)
• NO2(GT)
• C6H6(GT)
• Temperature
• Relative Humidity
• Absolute Humidity

These readings were collected from chemical sensors in an urban environment.

---

✨ Project Workflow

1️⃣ Data Loading & Exploration

• Loaded CSV dataset using Pandas
• Explored dataset shape, data types, and summary statistics
• Checked missing values and anomalies

---

2️⃣ Data Cleaning & Preprocessing

Real-world datasets contain noise and invalid values.

Key preprocessing steps:

• Replaced invalid values (-200) with NaN
• Handled missing values using mean imputation
• Converted Date and Time columns into datetime format
• Combined date & time into a single timestamp feature

This step prepared the data for time-series modeling.

---

3️⃣ Feature Preparation for Forecasting

Prophet requires specific column names:

• ds → timestamp
• y → target variable

The model predicts Relative Humidity (RH) as the air quality indicator.

---

4️⃣ Time-Series Forecasting with Prophet

Used Facebook Prophet to:

• Train model on historical data
• Capture trend and seasonality
• Generate future predictions (365 hours ahead)

This demonstrates real-world time-series forecasting.

---

5️⃣ Forecast Visualization

Generated visualizations including:

• Forecast trend graph
• Confidence intervals
• Trend and weekly seasonality components

These plots help understand long-term air quality patterns.

---

📈 Model Output

The model produces:

• Future predictions (yhat)
• Lower and upper confidence intervals
• Trend and seasonal patterns

This allows analysis of how air quality may change over time.

---