Predict the weather conditions of a day in NYC or other large cities using various machine learning algorithms. Observe the accuracy of each model in all categories of weather conditions. Prediction is split into 5 categories:

• Sky - Clear vs Cloudy
• Rain
• Thunderstorm
• Snow
• Fog - Mist and Haze included

This split allows the models to be properly fitted to a single feature rather than overextending themselves too thin.

Prediction requires four input values from the user:

• Date to predict
• Current Temperature
• Current Pressure
• Current Wind Direction

This four gives the best accuracy across all models tested. Any less or more increases the error of the predictions.

Models were trained on data from: https://www.kaggle.com/selfishgene/historical-hourly-weather-data?select=weather_description.csv.

Model is currently trained on NYC but can be easily swapped with other cities from the dataset.

Accuracy of some of the models used:	Example of weather pridiction:

Clone this repo and cd into it:

git clone https://github.com/ShanaryS/weather-prediction-ML.git
cd weather-prediction-ML

Create and activate your virtual environment:

• Windows:

virtualenv env
.\env\Scripts\activate

• MacOS/Linux:

virtualenv --no-site-packages env
source env/bin/activate

Install the required packages:

pip install -r requirements.txt

• Train Models:

python run_training

• Predict Weather:

python run_prediction

MIT