Proposal of Semi-supervised Visualization System Using Interactive Genetic Programming

Abstract

This paper proposes a semi-supervised visualization system using interactive Genetic Programming (GP). One of the recent trends in visualization research is an automatic generation of visualization. While such an automatic visualization can help users select appropriate graphs without the knowledge of visualization and reduce users' workload, a semi-supervised approach is also useful for representing users' preferences and purpose for generating a graph. The proposed method applies genetic operations to the source code of Vega-lite to generate new graphs. By employing interactive GP, users' preferences can be reflected in the new population. Experimental results show the proposed system can generate graphs that satisfy the test participants.