Weather Forecast Model Ml For Tashkent

Abstract

This repository contains a comprehensive weather prediction system for Tashkent, Uzbekistan, utilizing 23+ years of historical hourly weather data. The project implements and compares multiple machine learning approaches:

• Linear Regression - Baseline model with time-series features
• LSTM (PyTorch) - Deep learning model for sequential data

After extensive testing, Linear Regression showed superior performance and is used as the primary model in the demo application.

Key Features

23+ Years of Data: Historical hourly weather data from 2002 to present

Multiple Predictions: Temperature, humidity, wind speed, pressure, and rain probability

Interactive Dashboard: Beautiful Streamlit web interface
Dual Versions:

• Tashkent-specific model (this version)
• All Uzbekistan cities (see Uzbekistan_weather_forecast folder)

Demo MP4

Installation

Clone the repository

git clone https://github.com/ShakhzodMirmuminov/Weather_predection_model.git

cd Weather_prediction_model

Install dependencies

pip install -r requirements.txt

Create a .env file based on .env.example:

   cp .env.example .env

Add your OpenWeatherMap API key to .env:

   OPENWEATHER_API_KEY=your_api_key_here

Collect historical data

Run the data collection script to fetch and update weather data:

python3 data_miner.py 

Launch the application

streamlit run streamlit_app.py

Output

Model Performance

Linear Regression Model (Primary) Test Set Performance (Last 20% of data):

RMSE: 1.245°C
MAE: 0.42°C
R² Score: 0.9847

LSTM Model (PyTorch)
Test Set Performance:

RMSE: 1.380°C
MAE: 1.077°C
R² Score: 0.9874

Contributing

Contributions are welcome! If you'd like to improve the rain prediction model or add new features.

Thank You