Intelligent analysis of color preferences: search for associative rules vs. cluster analysis
The aim of the article is to present the experience of experimental implementation on the basis of modern software platforms and technologies of two different methods of data mining: (1) the method of associative rules, and (2) the method of clustering. The authors analyze potential and limitations of using both methods in socio-psychological research of color preferences. The material for the conducted experiment was the data of a socio-psychological study in the course of which the subjects (N = 50) were shown a color palette containing 27 different shades and were asked to select from it the colors which, in their opinion, best fit the interior of each of seven different room types: living room, entrance room, bedroom, bathroom, toilet, kitchen, and hallway. By means of the Apriori algorithm we obtained associative rules corresponding to the relationship between color preferences and room types. We discuss the potential of hierarchical clustering method application for obtaining conclusions, which cannot be achieved by calculating percentages. The choice of rules for combining clusters, which give the most objective and informative assessment of responses, is performed. It was proposed to calculate the distance between color choices of respondents using CIEDE2000 color difference metric. By means of concrete examples it is shown that both methods of intellectual analysis open wide possibilities for visualization of the revealed psychological mechanisms and regularities. Experiments were carried out, in the course of which it was established that the chosen methods allow to carry out an effective assessment of social and psychological research data. We can conclude that with a significant increase in the number of respondents and the range of possible variants of answers, the task of finding associations is effectively solved by parallel methods.