🌤️ Weather Condition Prediction Using Machine Learning

This repository contains a machine learning project that focuses on predicting weather conditions using three different algorithms:

K-Nearest Neighbors (KNN)

Logistic Regression

Decision Tree

The models are trained and evaluated on three separate synthetic weather datasets, with a full workflow from data cleaning to model comparison.

📁 Project Structure

📦 Weather-condition-prediction

├── 📊 Logistic Regression.ipynb
├── 📊 KNN.ipynb
├── 📊 Decision Tree.ipynb
├── 📄 seattle-weather.csv
├── 📄 traffic_weather.csv
├── 📄 weather3.csv
└── 📑 Weather PPT (1).pdf

🔍 Datasets

Each dataset represents synthetic weather data with continuous and categorical features. The target variable in all three datasets is the weather condition.

├── seattle-weather.csv
├── traffic_weather.csv
├── weather3.csv

Data Cleaning Steps:

├── Removed missing values
├── Removed duplicates
├── Detected and eliminated outliers

🧠 Algorithms Used

✅ K-Nearest Neighbors (KNN)

Great performance on dataset 1

Simple yet effective for classification tasks

✅ Logistic Regression

Suitable for binary/multiclass classification

Interpretable model with solid baseline accuracy

✅ Decision Tree

Intuitive and visual model

Handles non-linear relationships well

📊 Model Performance (Test Accuracy)

Algorithm	Dataset 1	Dataset 2	Dataset 3
KNN	0.99	0.78	0.53
Logistic Regression	0.92	0.76	0.44
Decision Tree	0.98	0.75	0.46

📈 Analysis Summary

🔹 KNN outperformed other models across all three datasets.

🔹 Dataset 1 showed the highest predictive performance across all models.

🔹 This suggests dataset 1 is the most suitable for modeling weather conditions in this context.

🧾 Conclusion

Through our analysis, KNN was found to be the most accurate model, and Dataset 1 emerged as the most reliable dataset for this problem. The project demonstrates how classic ML algorithms can be leveraged for weather prediction when paired with proper data preprocessing.

📎 Presentation

For a summary of this project and visual representation of results, check out the 📄 Weather Project Presentation (PDF)