Online survey for collective clustering of computer generated architectural floor plans

TL;DR

This study uses an online survey to analyze how architecture practitioners and students perceive and perform clustering of floor plans, aiming to improve educational methods and database retrieval systems.

Contribution

It introduces a hybrid evolutionary approach combined with an online survey to understand collective clustering behaviors in architectural floor plans.

Findings

Respondents' clustering criteria vary significantly.

Survey results inform development of clustering algorithms.

Insights support educational and database retrieval improvements.

Abstract

The aim of this study is to understand what are the collective actions of architecture practitioners when grouping floor plan designs. The understanding of how professionals and students solve this complex problem may help to develop specific programmes for the teaching of architecture. In addition, the findings of this study can help in the development of query mechanisms for database retrieval of floor plans and the implementation of clustering mechanisms to aggregate floor plans resulting from generative design methods. The study aims to capture how practitioners define similarity between floor plans from a pool of available designs. A hybrid evolutionary strategy is used, which takes into account the building's functional program to generate alternative floor plan designs. The first step of this methodology consisted in an online survey to gather information on how the respondents would perform a clustering task. Online surveys have been used in several applications and are a method of data collection that conveys several advantages. When properly developed and implemented, a survey portrays the characteristics of large groups of respondents on a specific topic and allows assessing its representation. Several types of surveys are available; e.g. questionnaire and interview formats, phone survey, and online surveys, which can be coupled with inference engines that act and direct the survey according to respondents' answers. In the present study, the survey was posed as an online exercise in which respondents had to perform a pre-defined task, which makes it similar to running an experiment in an online environment. The experiment aimed to understand the perception and criteria of the target population to perform the clustering task by comparing the results with the respondents' answers to a questionnaire presented at the end of the exercise.