Effective visualization of large multidimensional datasets

摘要: A new method for assisting with the visualization of large multidimensional datasets is proposed. We classify more than one million elements as large. Multidimensional data are two or dimensions, each which at least binary. involves representation in a low dimensional environment, such computer screen printed media. Traditional techniques not well suited to solving this problem. Our based part on field cognitive psychology called preattentive processing. Preattentive processing study visual features that detected rapidly and little effort by human system. Examples include hue, orientation, form, intensity, motion. studied ways extending applying research results from address our requirements. used investigations build tools allow user very accurately perform exploratory analysis tasks. These tasks searching target elements, identifying boundaries between groups common estimating number have specific feature. Our experimental were positive, suggesting dynamic sequences frames can be explore amounts relatively short period time. Recent work both scientific database systems has started problems inherent managing datasets. One promising technique knowledge discovery, "the nontrivial extraction implicit, previously unknown, potentially useful information data". hypothesise discovery filter reduce amount sent tool. Data do belong user-chosen group interest discarded, dimensionality individual compressed, unknown trends relationships discovered explored. We illustrate how them real-world This includes simulated salmon migration results, computerized tomography medical slices, environmental track ocean atmospheric conditions.