Information architecture (IA): using multidimensional scaling (MDS) and K -means clustering algorithm for analysis of card sorting data
Sione Paea, Ross Baird · Journal of Usability Studies archive · 2018
We present a method for visualizing and analyzing card sorting data aiming to develop an in-depth and effective information architecture and navigation structure. One of the well-known clustering techniques for analyzing large data sets is with the k-means algorithm. However, that algorithm has yet to be widely applied to analyzing card sorting data sets to measure the similarity between cards and result displays using multidimensional scaling. The multidimensional scaling, which employs particle dynamics to the error function minimization, is a good candidate to be a computational engine for interactive card sorting data. In this paper, we apply the combination of a similarity matrix, a k-means algorithm, and multidimensional scaling to cluster and calculate an information architecture from card sorting data sets. We chose card sorting to improve an information architecture. The proposed algorithm handled the overlaps between cards in the card sorting data quite well and displayed the results in a basic layout showing all clusters and card coordinates. For outliers, the algorithm allows grouping of single cards to their closest core clusters. The algorithm handled outliers well choosing cards with the strongest similarities from the similarity matrix. We tested the clustering algorithm on real-world data sets and compared to other techniques. The results generated clear knowledge on relevant usability issues in visualizing information architecture. The identified usability issues point to a need for a more in-depth search of design solutions that are tailored for the targeted group of people who are struggling with complicated visualizing techniques. This study is for people who need support to easily visualize information architecture from data sets.