Problem structuring for the analysis of architectural design data
One of the many problems in architectural design is the multivariate nature of the design problem. Typically this problem has been resolved by ranking the various design elements and working through a series of design modifications considering each in turn. Unfortunately architecture is concerned with complex situations in which many variables are simultaneously related and the "fragmentary" approach gives little insight into the basic relationships obscured within the design data. To achieve that insight the complex of variables must be studied as a whole. This thesis describes a way of examining activity data sheets or other briefing data using a number of techniques based on multivariate statistical methods. The various techniques have been incorporated into a computer program called MAGIC - Multivariate Analysis by Graphical Interactive Computing. The program output is specially designed to produce diagrams to enable the designer to manipulate and investigate the design data easily and conveniently. The thesis reviews the problem of architectural design and its place in design methods theory, and the relationship of MAGIC to other layout planning programs. The program structure is outlined and detailed descriptions of the analytical techniques presented, together with those graphical techniques developed to present the results. Finally the application of MAGIC is shown in two practical examples.