Deep Learning Based Weather Prediction Architecture Comparison

This repository contains a comparative look at LSTM and multi-layer feed-forward neural network architectures regarding the time series forecasting task of predicting future weather given past weather data.

Before running and viewing our reproduceable results, our Conda Python environment must be created to run the notebook in for best results.

Environment Setup

To construct a Conda Python environment to run this pipeline with, ensure Conda is installed and run

conda env create -f env.yml

To run this environment, run the command

conda activate weather_predict

Note that this environment must be connected to comparison_results.ipynb before running. For help achieving this, see this article.

Also note that this environment was only tested on OSX, so compatability issues may exist on other operating systems.