This repository contains experiments on synthetic data generated by @sinasta. The project aims to generate a synthetic dataset of apartment floor plans using parametric tools and layout algorithms to enable automation within a defined range of variability.
Apartments are generated through parametric geometry subdivision using Python, Blender's Sverchok, and various algorithms such as Voronoi diagrams, KD trees, and treemapping algorithms. The generated data is analyzed from various geometric aspects using the Python library Topologic. Openings such as doors and windows are integrated into the basic geometry in a variable pattern within the framework of defined spatial and architectural rules. The data is then reconstructed into a three-dimensional model and stored in a common open-source BIM format. Energy performance simulation is done for each apartment using Energy+.
The evaluated synthetic data can be used as training data for a machine learning model that predicts the energy performance of the architectural objects using the topological graphs and their added dictionaries. The focus of this repository is on traditional machine learning algorithms (like RF) and naive algorithms.
The ongoing research for this thesis delves into the field of architectural design and energy efficiency, with a specific focus on the use of graph representation in architectural design and the application of machine learning techniques to classify buildings based on their energy performance. The study aims to contribute to the advancement of knowledge in the field of sustainable building design practices by utilizing a graph-based approach to architectural object representation and machine learning for building energy performance classification.