Even if we haven't analyzed in detail the internal dynamics of LSTM cells, we want to present a simple example of a time-series forecast using this kind of model. For this task, we have chosen a dataset of average Earth temperature anomalies (collected every month) provided by the Global Component of Climate at a Glance (GCAG) and available through DataHub (https://datahub.io).
It is possible to download the CSV files directly from https://datahub.io/core/global-temp; however, I suggest installing the Python package through the pip -U install datapackage command and using the API (as shown in the example) to get all the available datasets.
The first step is downloading and preparing the dataset:
from datapackage import Package
package = Package('https://datahub.io/core/global-temp/datapackage.json')
for resource in package...
