weather prediction model

Project Description

This project implements a comprehensive machine learning solution for predicting rainfall in Australia using the weatherAUS dataset. The system analyzes various meteorological features to forecast whether it will rain tomorrow, providing valuable insights for weather prediction applications.

Dataset Overview

Source: weatherAUS.csv (Australian Weather Dataset)

Target Variable: RainTomorrow (Binary classification: Yes/No)

Features: 20+ meteorological variables including:

Temperature measurements (MinTemp, MaxTemp, Temp9am, Temp3pm)

Wind data (WindGustSpeed, WindSpeed9am, WindSpeed3pm, Wind directions)

Humidity levels (Humidity9am, Humidity3pm)

Atmospheric pressure (Pressure9am, Pressure3pm)

Cloud coverage and rainfall measurements

Location and temporal features (Month, Day)

Technical Implementation

Data Preprocessing

Feature Engineering: Extracted temporal features (month, day) from date information

Missing Data Handling: Removed columns with >40% missing values

Encoding: Converted categorical variables (RainToday, RainTomorrow) to binary format

Pipeline Architecture: Implemented robust preprocessing pipeline with:

Median imputation and StandardScaler for numerical features

Constant imputation and OneHotEncoder for categorical features

Machine Learning Models

Implemented and evaluated 5 different classification algorithms:

Logistic Regression - Accuracy: 99.85%

K-Nearest Neighbors (k=5) - Accuracy: 87.56%

Gaussian Naive Bayes - Accuracy: 79.63%

Decision Tree - Accuracy: 100%

Random Forest - Accuracy: 100%

Model Evaluation

Train-test split: 80/20 with stratification

Comprehensive metrics: Accuracy, Precision, Recall, F1-Score

Confusion matrices for detailed performance analysis

Classification reports for class-wise evaluation

Key Results

The ensemble methods (Random Forest and Decision Tree) achieved perfect accuracy on the test set, demonstrating excellent predictive capability. Logistic Regression also showed exceptional performance at 99.85% accuracy, providing a simpler yet highly effective alternative.

Technologies Used

Python 3.11

Libraries:

pandas, numpy (Data manipulation)

scikit-learn (Machine learning models and preprocessing)

Jupyter Notebook (Development environment)

Deliverables

Complete Jupyter Notebook with all code and analysis

Model accuracy summary report (CSV format)

Trained pipeline ready for deployment

Comprehensive evaluation metrics and visualizations

Project Applications

This weather prediction system can be utilized for:

Agricultural planning and crop management

Outdoor event planning

Transportation and logistics optimization

Emergency preparedness and disaster management

General weather forecasting services