sebastianbarfort / sds

Social Data Science, course at University of Copenhagen
http://sebastianbarfort.github.io/sds/
12 stars 17 forks source link

Group 27 - Exam project description #63

Closed StineWesselhoff closed 8 years ago

StineWesselhoff commented 9 years ago

title: "Group 27: Exam Project Description" author: "Stine Wesselhoff, Esben Østergaard, Valdemar Stentoft-Hansen, Robert Ziegler" date: "23 November 2015"

Biking Behaviour and Weather in Copenhagen

The Idea

Copenhagen is one of the most bicycle-friendly cities worldwide. The weather can be harsh and often unpredictable. People in T-Shirts bike alongside people in full rain-gear. In Denmark, weather is more than just a small talk topic. It can be decisive whether you reach your destination without getting wet.

In this assignment we want to examine how the weather and season affects peoples biking behaviour. As far as we know this has not been done before. More specific we will try to come up with answers for the follow questions:

For our assignment we need weather data as well as the number of persons biking. Luckily, meteorology lives from recording and saving impressive amounts of data. We would like to use mainly the following two weatherbases: weatherbase.com, wunderground.com. These pages offer a lot of data and allow for scraping and downloading. We already started with scraping and first data manipulations. Unfortunately, at this point we are not able to gather data on specific precipitation. Nevertheless, a lot of historic hourly weather data is availabe and seems useful for our project Regarding the bike data, we use records of counters which are installed on bicycle tracks on Dronning Louises Bro and Raadhuspladsen in the inner city of Copenhagen. We contacted the operators of the bike counters and they provided us historic hourly data of the two bike counters.

Statistical Methods

First we will explorer the data by doing some data vizualisations and tables. We will use two statstical methods for two different purposes:

  1. Inference: In order to analyze the correlation and impact of different weather conditions, weekdays and time of year.
  2. Prediction: Using a CART analysis, we will try to predict predict bike usage for specific weather observations.

Then when looking at tomorrows weather forecast, you will be able to predict tomorrows number of bikes. This might be useful for e.g the suppliers of public transportation, under the assumption that people will substitute from bikes to public transportation.