Building Type Classification (Python SciKit)

Does Form follow Function? Case Study Hamburg

Architects and planners have always assumed that there is a relationship between form and function. This work is investigating if the function of a building can be predicted based on its geometrical attributes by using supervised machine learning techniques.

Overview Content Notebook

1. Introduction and objective
2. Metadata
3. Cleaning and reshaping the dataset
4. Exploring the variables
   • Target variable
   • Independent variable
   • Key findings after initial analysis
5. Preparing the data for classification
   • Dummy variables
   • Manual feature selection
6. K-nearest Neighbor Classifier
   • Initial test runs
   • Initial results
   • Gridsearch
7. Decision Tree Classifier
   • Initial test runs
   • Gridsearch and pipelining
8. Random Forest Classifier
   • Gridsearch and pipelining
9. Key findings
10. References