Evaluating Multi-Dimensional Visualizations for Understanding Fuzzy Clusters
Ying Zhao, Feng Luo, Minghui Chen et al. · IEEE Transactions on Visualization and Computer Graphics · 2018
Fuzzy clustering assigns a probability of membership for a datum to a cluster, which veritably reflects real-world clustering scenarios but significantly increases the complexity of understanding fuzzy clusters. Many studies have demonstrated that visualization techniques for multi-dimensional data are beneficial to understand fuzzy clusters. However, no empirical evidence exists on the effectiveness and efficiency of these visualization techniques in solving analytical tasks featured by fuzzy clusters. In this paper, we conduct a controlled experiment to evaluate the ability of fuzzy clusters analysis to use four multi-dimensional visualization techniques, namely, parallel coordinate plot, scatterplot matrix, principal component analysis, and Radviz. First, we define the analytical tasks and their representative questions specific to fuzzy clusters analysis. Then, we design objective questionnaires to compare the accuracy, time, and satisfaction in using the four techniques to solve the questions. We also design subjective questionnaires to collect the experience of the volunteers with the four techniques in terms of ease of use, informativeness, and helpfulness. With a complete experiment process and a detailed result analysis, we test against four hypotheses that are formulated on the basis of our experience, and provide instructive guidance for analysts in selecting appropriate and efficient visualization techniques to analyze fuzzy clusters.
Cited by 64 works. doi:10.1109/tvcg.2018.2865020
Authors
References
- Fuzzy sets
- UCI Machine Learning Repository
- Statistical Comparisons of Classifiers over Multiple Data Sets
- FCM: The fuzzy c-means clustering algorithm
- A new approach to clustering
- Visualizing Data
- The dynamic HomeFinder
- Artificial neural networks for feature extraction and multivariate data projection
- Hyperdimensional Data Analysis Using Parallel Coordinates
- Principal Component Analysis and Factor Analysis