Using a Combination of CNN and an RNN we were able to preidict the occurence of snow around Lake Michigan. The CNN was used to analyse image data and the RNN was used to analyse meteorological numeric data. This Notebook consists of the first method which we applied to try fitting the dataset along with the 1D and 2D images which were presented to us. In this method we created new features and used the 1D images to calculate intensities on the the days and different times. The 2D images were used as part of the Convolutional Neural Network (CNN) whereas the numerical data was used in a Recurrent Neural Network (RNN), both networks were concatenated to predict occurence of snowfall.
The other method we used in building an optimised Neural Network was to create a sliding window and analyse data along every few days.
Our Intuitions for building the models in different methodologies is elaborated in this notebook. (The dependent files are > 4 GB and hence I was unable to upload them on this forum)