Unterschreiben Sie den Mietvertrag!

Berlin used to be an affordable place not so long ago. When I moved to Berlin, my uncle was paying around 580 euros for a one-bedroom apartment in the Mitte neighborhood. From there, you could walk through history, art, and culture without even taking a bus. Things have changed since then. Berliners not only need to compete between thousands of applications to find a place, but also they need to pay hefty rents to secure what they want.

I want to investigate this matter and answer some questions. More importantly, I decided to make a machine learning model predicting the rent prices in Berlin using different features. Since there are not many public datasets about berlin house prices available out there, I used a kaggle dataset. This dataset contains apartment rental offers in Germany from the dates 2018-09-22, 2019-05-10, and 2019-10-08 (Collected from immobilienscout24). I filtered out Berlin listings to work on it exclusively.

The berlin dataset is available on my Github.

• Gathered berlin listings from immobilienscout24, validated and cleaned data
• Performed an exploratory data analysis (plus AutoEDA) and answered different questions about the berlin real state market
• Built machine learning models to predict rent prices in berling using regression techniques

Libraries Used

• pandas
• numpy
• matplotlib
• seaborn
• scikit-learn
• missingno

AutoML by:

• pandas_profiling
• AutoViz

Dataset

The variables of this dataset are:

• regio1 - bundesland
• serviceCharge - aucilliary costs such as electricty or internet in €
• heatingType - type of heating
• telekomTvOffer - is payed TV included if so which offer
• telekomHybridUploadSpeed - how fast is the hybrid inter upload speed
• newlyConst - is the building newly constructed
• balcony - if a listing has a balcony
• picturecount - how many pictures were uploaded to the listing
• pricetrend - price trend as calculated by Immoscout
• telekomUploadSpeed - how fast is the internet upload speed
• totalRent - total rent (usually a sum of base rent, service charge and heating cost)
• yearConstructed - construction year
• scoutId - immoscout Id
• noParkSpaces - number of parking spaces
• firingTypes - main energy sources, separated by colon
• hasKitchen - has a kitchen
• geo_bln - bundesland (state), same as regio1
• cellar - has a cellar
• yearConstructedRange - binned construction year, 1 to 9
• baseRent - base rent without electricity and heating
• houseNumber - house number
• livingSpace - living space in sqm
• geo_krs - district, above ZIP code
• condition - condition of the flat
• interiorQual - interior quality
• petsAllowed - are pets allowed, can be yes, no or negotiable
• street - street name
• streetPlain - street name (plain, different formating)
• lift - is elevator available
• baseRentRange - binned base rent, 1 to 9
• typeOfFlat - type of flat
• geo_plz - ZIP code
• noRooms - number of rooms
• thermalChar - energy need in kWh/(m^2a), defines the energy efficiency class
• floor - which floor is the flat on
• numberOfFloors - number of floors in the building
• noRoomsRange - binned number of rooms, 1 to 5
• garden - has a garden
• livingSpaceRange - binned living space, 1 to 7
• regio2 - District or Kreis, same as geo krs
• regio3 - City/town
• description - free text description of the listing
• facilities - free text description about available facilities
• heatingCosts - monthly heating costs in €
• energyEfficiencyClass - energy efficiency class (based on binned thermalChar, deprecated since Feb 2020)
• lastRefurbish - year of last renovation
• electricityBasePrice - monthly base price for electricity in € (deprecated since Feb 2020)
• electricityKwhPrice - electricity price per kwh (deprecated since Feb 2020)
• date - time of scraping