An Approach to Supporting Incremental Visual Data Classification

Abstract

Abstract-Automatic data classification is a computationally intensive task that presents variable precision and is considerably sensitive to the classifier configuration and to data representation, particularly for evolving data sets. Some of these issues can best be handled by methods that support users' control over the classification steps. In this paper, we propose a visual data classification methodology that supports users in tasks related to categorization such as training set selection; model creation, application and verification; and classifier tuning. The approach is then well suited for incremental classification, present in many applications with evolving data sets. Data set visualization is accomplished by means of point placement strategies, and we exemplify the method through multidimensional projections and Neighbor Joining trees. The same methodology can be employed by a user who wishes to create his or her own ground truth (or perspective) from a previously unlabeled data set. We validate the methodology through its application to categorization scenarios of image and text data sets, involving the creation, application, verification, and adjustment of classification models.

The State‐of‐the‐Art in Predictive Visual Analytics

“…Users can also manually assign the class label as part of the incremental modeling approach, and the system will highlight the data points whose prediction will change based on this update. Paiva et al [PSPM15] used a Neighbor Joining tree and a similarity layout view for interpreting misclassified instances to support the labeling and training set selection as part of an incremental learning procedure. We find that this concept of interactively labeling data as part of a model learning process is also supported by other PVA works [BO13, HNH*12, ZLH*16].…”

The State of the Art in Integrating Machine Learning into Visual Analytics

“…In their visual classification methodology, Paiva et al . [PSPM15] demonstrate that effective classification models can be built when users’ interactive input, for instance, to select wrongly labelled instances, can be employed to update the classification model. Along the similar lines, Behrisch et al .…”

“…The ideas employed under this section show parallels to the Active Learning methodologies develop in the ML literature [Set09] where the algorithms have capabilities to query the user for intermediate guidance during the learning process. In their visual classification methodology, Paiva et al [PSPM15] demonstrate that effective classification models can be built when users' interactive input, for instance, to select wrongly labelled instances, can be employed to update the classification model. Along the similar lines, Behrisch et al [BKSS14] demonstrate how users' feedback on the relevance of features in classification tasks can be incorporated into decisionmaking processes.…”

A novel method to classify organizational data using CIA tree approach

“…These factors ensure security through indexing [18]. Through this we can optimize spatial factors [19] [20]. Comparitve analysis on different technologies makes a perfect system [21].…”