Employing artificial neural networks for constructing metadata-based model to automatically select an appropriate data visualization technique

摘要: Display Omitted Solution to automatically select appropriate visualization technique based on metadata is presented.A purpose built dataset extracted from existing knowledge in the field used train classifiers.A comparison of results obtained best ANN architecture performed with five other classifiers.The proposed system outperforms four classifiers terms accuracy and running time.The work brings new perspective visualization. Advances computing technology have been instrumental creating an assortment powerful information techniques. However, selection a suitable effective for specific data mining task not trivial. This selects given that user intends perform. The predicted artificial neural network (ANN)-based model which classifies input into one eight predefined classes. A discipline utilized network. covers techniques, including: histogram, line chart, pie scatter plot, parallel coordinates, map, treemap, linked graph. Various architectures using different numbers hidden units, layers, output formats evaluated find optimal architecture. performance networks measured using: confusion matrix, accuracy, precision, sensitivity classification. Optimal determined by convergence time number iterations. are compared classifiers, k-nearest neighbor, nave Bayes, decision tree, random forest, support vector machine. all execution time. trained also tested twenty real-world benchmark datasets, where approach provides two alternate visualizations, addition most one, particular dataset. qualitative state-of-the-art approaches presented. show assists selecting high accuracy.