English | Русский
AI_WeatherITSH - my school project that forecasts future weather by using neural networks and predicting temperature for given intervals. It also represents one of the most popular solution for time series forecasting problems (weather forecast in my case).
My model is made of Recurrent Neural Networks (RNN), to be more precise one of its variation - Long short-term memory (LSTM).
The main goal of project is to research and learn something new about Neural Networks and how they work.
It has bugs... and there's lot of things to improve.
- Make sure you have version of python
3.10.Xor higher. - Set up your environment (venv/conda etc... by your choice) and install all requirements.
- Compile and train the model by simply running
model.pyscript. (You can adjust model as you want, merely edit every needed variables)
Done... you've got a trained and fully capable of predicting model.
In this project, I used REST API from https://narodmon.ru site. To obtain the weather data, you have to set up some environment variables in narodmonitoring\.env file.
Before you do that, you must create an account there and get your api key, login and password.
Also, find any public sensors(pressure, temperature, humidity) then get their ids and insert into narodmonitoring\.env file.
Run __init__.py.
After when everything is done, run forecast.py script to get your forecast as matplot visualization.
Weather forecast from 10:00 to 22:00 in 24.07.2022 (Moscow)
Weather forecast from 0:00 to 12:00 in 28.01.2023 (Moscow) (not so accurate temperature, depends on time of training model)
