Abstract
Despite the apparent benefits of machine learning (ML), for many practical applications, with architecture being no exception, the current bottleneck limiting its full potential is not the amount but the quality of available data.
We argue that in order to increase the quality of floor plan data sets, we need a priori approach to feature selection. In essence, to optimally allocate limited resources, we need to identify the most valuable features before collecting the data instead of the current practice of discarding irrelevant features after being acquired.
For this purpose, we evaluate 52 features describing a large variety of geometrical and contextual properties of 5558 Swiss residential floor plans and identify a set of five most valuable features. The value of each feature depends on the novel information it provides in combination with other features in the set. Consequently, the selected feature set aims to maximize the joint information value while minimizing redundancies. It is important to note that the proposed method and selected features are generalizable only for use in unsupervised ML applications but might not suit the specific needs of different supervised ML tasks.
We also discuss the possibilities of overcoming the limitation of our findings to the Swiss context. We show that the proposed feature selection method is robust enough to be fitted on small sample sizes (N = 20) and thus can be applied in other contexts to determine the most valuable feature before acquiring the complete data set.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The base 2 of the logarithm makes it possible to interpret the Shannon entropy as a measure of a lower bound on the number of bits […] needed on average to encode symbols drawn from a distribution.” [18].
- 2.
For feature selections consisting just of two features (N = 2), the cross-correlations as used in filter methods provides enough information to select set maximizing the joint information entropy. For N > 2, wrapper search will lead to set with higher joint IE.
- 3.
The search requires n(n + 1)2 calculations.
- 4.
Computational complexity follows a quadratic function, which means that even a small increase in the set of feature candidates can dramatically increase the computation time of the feature selection method.
- 5.
Due to scale sensitivity of K-means all numerical features were standardized to z-score before running the k-means clustering.
- 6.
The number of clusters represents the number of distinct categories in each higher-order categorical feature. Since the number of categories directly affects the IE and causes bias of the IE towards features with higher number of categories, we keep the number of categories constant throughout the data set.
- 7.
For details on data acquisition and methods applied to derive the floor plan features please see to the online reference by the data-set provider Archilyse AG: https://zenodo.org/record/7070952#.YymPDtJBy-Z.
- 8.
Some high-level room features such as area do not vary inside the room and thus are not accompanied by any low-level features.
- 9.
We select with replacement 1024 observations and repeat the process 20 times. These parameters were empirically calibrated to provide best accuracy by keeping the feature selection procedure under 30 min.
- 10.
The threshold 5% is arbitrarily chosen and depends on the purpose of the feature selection. Nevertheless, we use this threshold for comparison purposes through out this paper.
- 11.
For comparison, the Tabula rasa search takes 30 min, while the mRMR feature selection takes approximately 3 s on the same data set.
References
Cetinic, E., She, J.: Understanding and creating art with AI: review and outlook. ACM Trans. Multimedia Comput. Commun. Appl. 18, 66:1–66:22 (2022)
Tostevin, P.: The total value of global real estate (2021)
Narahara, T., Yamasaki, T.: Subjective functionality and comfort prediction for apartment floor plans and its application to intuitive searches (2022)
Guo, X., Peng, Y.: Floor plan classification based on transfer learning. In: 2018 IEEE 4th International Conference on Computer and Communications (ICCC), pp. 1720–1724 (2018)
Nauata, N., Chang, K.-H., Cheng, C.-Y., Mori, G., Furukawa, Y.: House-GAN: relational generative adversarial networks for graph-constrained house layout generation. In: Vedaldi, A., Bischof, H., Brox, T., Frahm, J.-M. (eds.) ECCV 2020. LNCS, vol. 12346, pp. 162–177. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-58452-8_10
Saltz, J.S.: CRISP-DM for data science: strengths, weaknesses and potential next steps. In: 2021 IEEE International Conference on Big Data (Big Data), pp. 2337–2344 (2021)
Standfest, M.: Reducing bias for evidence-based decision making in design. In: Gengnagel, C., Baverel, O., Betti, G., Popescu, M., Thomsen, M.R., Wurm, J. (eds.) Towards Radical Regeneration, pp. 122–132. Springer, Cham (2022)
Tang, J., Alelyani, S., Liu, H.: Feature selection for classification: a review. In: Data Classification. Chapman and Hall/CRC (2014)
Guyon, I., Elisseeff, A.: An introduction to variable and feature selection. J. Mach. Learn. Res. 3, 1157–1182 (2003)
Zhao, Z., Anand, R., Wang, M.: Maximum relevance and minimum redundancy feature selection methods for a marketing machine learning platform (2019)
Standfest, M., et al.: Swiss dwellings: a large dataset of apartment models including aggregated geolocation-based simulation results covering viewshed, natural light, traffic noise, centrality and geometric analysis (2022). https://zenodo.org/record/7070952
Sheikhpour, R., Sarram, M.A., Gharaghani, S., Chahooki, M.A.Z.: A survey on semi-supervised feature selection methods. Pattern Recogn. 64, 141–158 (2017)
Alelyani, S.: On feature selection stability: a data perspective. Arizona State University (2013)
Solorio-Fernández, S., Carrasco-Ochoa, J.A., Martínez-Trinidad, J.F.: A review of unsupervised feature selection methods. Artif. Intell. Rev. 53(2), 907–948 (2019). https://doi.org/10.1007/s10462-019-09682-y
Devakumari, D., Thangavel, K.: Unsupervised adaptive floating search feature selection based on contribution entropy. In: 2010 International Conference on Communication and Computational Intelligence (INCOCCI), pp. 623–627 (2010)
Dy, J.G., Brodley, C.E.: Feature selection for unsupervised learning. J. Mach. Learn. Res. 5, 845–889 (2004)
Cover, T.M., Thomas, J.A.: Elements of Information Theory. Wiley (2012)
Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press (2016)
Shannon, C.E.: A mathematical theory of communication. Bell Syst. Tech. J. 27, 379 (1948)
Cover, T.M.: The best two independent measurements are not the two best. IEEE Trans. Syst. Man Cybern. SMC-4, 116–117 (1974)
Efron, B.: Computers and the theory of statistics: thinking the unthinkable. SIAM Rev. 21, 460–480 (1979)
Fix, E., Hodges, J.L.: Discriminatory analysis. nonparametric discrimination: consistency properties. Int. Stat. Rev./Revue Internationale de Statistique 57, 238–247 (1989)
Maron, M.E.: Automatic indexing: an experimental inquiry. J. ACM 8, 404–417 (1961)
Satopaa, V., Albrecht, J., Irwin, D., Raghavan, B.: Finding a “kneedle” in a haystack: detecting knee points in system behavior. In: 2011 31st International Conference on Distributed Computing Systems Workshops, pp. 166–171 (2011)
Rousseeuw, P.J.: Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. J. Comput. Appl. Math. 20, 53–65 (1987)
Brown, G., Pocock, A., Zhao, M.-J., Luján, M.: Conditional likelihood maximisation: a unifying framework for information theoretic feature selection. J. Mach. Learn. Res. 13, 27–66 (2012)
Ding, C., Peng, H.: Minimum redundancy feature selection from microarray gene expression data. J. Bioinform. Comput. Biol. 3, 185–205 (2005)
Hu, R., Huang, Z., Tang, Y., Van Kaick, O., Zhang, H., Huang, H.: Graph2Plan: learning floorplan generation from layout graphs. ACM Trans. Graph. 39 (2020)
Funding
This research was conducted in scope of the Neufert 4.0 project funded by the Federal Ministry for Housing, Urban Development and Building.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Bielik, M., Zhang, L., Schneider, S. (2023). Big Data, Good Data, and Residential Floor Plans. In: Turrin, M., Andriotis, C., Rafiee, A. (eds) Computer-Aided Architectural Design. INTERCONNECTIONS: Co-computing Beyond Boundaries. CAAD Futures 2023. Communications in Computer and Information Science, vol 1819. Springer, Cham. https://doi.org/10.1007/978-3-031-37189-9_40
Download citation
DOI: https://doi.org/10.1007/978-3-031-37189-9_40
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-37188-2
Online ISBN: 978-3-031-37189-9
eBook Packages: Computer ScienceComputer Science (R0)