An Analysis of Automated Visual Analysis Classification: Interactive Visualization Task Inference of Cancer Genomics Domain Experts

摘要: We show how mouse interaction log classification can help visualization toolsmiths understand their tools are used “in the wild” through an evaluation of MAGI - a cancer genomics tool. Our primary contribution is twelve visual analysis task classifiers, which compares predictions to inferences made by pairs and experts. uses common classifiers that accessible most evaluators: $k$ -nearest neighbors, linear support vector machines, random forests. By comparing classifier experts, we simple automated have up 73 percent accuracy separate meaningful logs from “junk” with 91 accuracy. second exploration trends using predictions, expands current knowledge about ecological tasks. third discussion inform iterative tool design. These contributions suggest viable method for (1) evaluating requirements client-side-focused tools, (2) allowing researchers study experts on larger scales than typically possible in-lab observation, (3) highlighting potential bias.