Modeling and visualize counted bicycles in Berlin with prophet and streamlit.io

Intro

A few days ago, AWS launched a new AppRunner service (https://aws.amazon.com/de/apprunner/).

I thought to myself: “Hey, let’s see how it works with a Mickey Mouse example”.

https://www.youtube.com/watch?v=NX8nqL9r9vc

Data

I found by accident inside the open data page from Germany the counting points in Berlin for bicycles interesting (https://bit.ly/2SzqLjO).

From my point of view an interesting collection of time series. To make it more interesting I added some historical weather data via https://www.visualcrossing.com/.

I used daily and hourly data preparations, but didn’t request weather data for every location in Berlin. For the weather, I used always, ‘Jannowitzbrücke-Süd’, because I worked there many years ago.

The data frame for modeling on a daily level looks like this (the holiday information matrix is added by prophet via the python-holiday lib):

See the notebook here: