Dear candidate
As part of the recruitment process, we would like to ask you to accomplish the following assessment task which should take not more than a couple of evenings of your time. Please do your best and aim for an end-to-end MVP solution.
As result of the exercise, it is expected to receive a link to the github/gitlab/bitbucket repo with working executable code base and instruction on how to run it, and with the detailed report (as jupyter notebook, or markdown file) containing description of your model, assumptions you made and justification of your tech choices.
We would like to propose you to familiarise yourself with the housing market in Hamburg, hence to suggest you to build a service to forecast price for the flats available for rent in Hamburg.
In order to complete that exercise, we would like to formulate the following task for you:
- Given the data sample ml_eng_ay_data.csv.gz (see data dictionary for columns description), build a model to forecast total rent price for up to 6 months in the future. The model should perform better than the seasonal naïve model in terms of RMSE. Use as many available features as it would make sense (based on the features importance) for the model accuracy and stability.
- Deploy your model as a service (e.g. as a dockerized web server) so it is capable to run a batch-prediction.
- (Optional) Build a dockerized stateless service to re-train your model.
| column | type | comment |
|---|---|---|
| date | date | date when the ad was published |
| cnt_rooms | int | number of rooms in the flat |
| flat_area | float | living area of the flat (in square meters) |
| rent_base | float | base monthly rent for the flat (in euro) |
| rent_total | float | total (inlucing utilities) monthly rent for the flat (in euro) |
| flat_type | string | type of the property, e.g. appartment, roof_storey, etc. |
| flat_interior_quality | string | quily of the flat interior |
| flat_condition | string | flat condition, e.g. normal, good, etc. |
| flat_age | string | category of the flat's age (in years), e.g. <5, <10, ..., <50 etc. |
| flat_thermal_characteristic | float | energy consumption for the flat (in kWh per square meter per year) |
| has_elevator | boolean | indicates if the house has an elevator |
| has_balcony | boolean | indicates if the flat has a balcony |
| has_garden | boolean | indicates if a garden can be accessed from the flat |
| has_kitchen | boolean | indicates if the flat has built-in kitchen |
| has_guesttoilet | boolean | indicates if the flat has guest toilet |
| geo_city | string | city location of the flat |
| geo_city_part | string | city district location of the flat |
Note: we would recommend you to consider employing tools/services and languages we use:
- DSL/Programming language: Python (ver. >= 3.7.4)
- ML Frameworks/libraries: TensorFlow 2.0, GluonTS (ver. >= 0.3.4)
- Docker (ver. >= 19)
Good luck and hopefully see you soon in our headquarters for onsite interview.
Best regards
ABOUT YOU GmbH
Amtsgericht Hamburg HR B 120869
Geschäftsführer: Sebastian Betz, Tarek Müller, Hannes Wiese