A combined principal component analysis and clustering approach for exploring enormous renovation design spaces

Abstract

This paper presents a new approach for supporting renovation designers in exploring the vast landscape of renovation scenarios. The approach integrates renovation scenario sampling (based on a semantic domain model of renovation, called NovaDM), scenario clustering (specifically k-means) in terms of Key Performance Indicator evaluations, and dimensionality reduction (specifically Principal Component Analysis) so that the sampled scenarios and discovered clusters can be visualized in a meaningful way. The approach is empirically evaluated on a large residential building case study in Denmark. Case study results show that the visualization captures and communicates critical high-level information about the "structure" of the enormous, complex underlying renovation design space, in terms of qualitative KPI profiles (e.g. fairly low privacy, very high cost, etc.), thus significantly enhancing renovation decision making.