Novice designers of visualizations need support to find easy-to-use color schemes for visualizations. This article introduces a method of conspicuity inference for a color selection support system that is applicable to a wide variety of visualizations. In previous research, data on color conspicuity by computer GUI intetface were collected by having subjects adjust the relative sizes of two figures on a ground in order to find conspicuity balance. This article outlines a new conspicuity inference system, which uses a neural network, to produce a color conspicuity value from the system RGB values of two figures and a potential ground. The output value is the ideal relative area of the two figures. This value is applied to visualization designs by weighting each conspicuity value with a ground coefficient, the relative size of every color in a design. Color selection and evaluation experiments show a significant relationship of conspicuity balance, as inferred by this system, for subjective ease-of-use evaluation scores for 200 novice designs. Topics for future research include developing the system's characteristically simple parameters within the limited scope and integrating conspicuity balance with other factors such as visibility, discriminability, and color harmony.