Data Visualization Classification Using Simple Convolutional Neural Network Model

Authors

  • Filip Bajić University Computing Centre (SRCE)
  • Josip Job Faculty of Electrical Engineering, Computer Science and Information Technology
  • Krešimir Nenadić Faculty of Electrical Engineering, Computer Science and Information Technology

DOI:

https://doi.org/10.32985/ijeces.11.1.5

Keywords:

data visualization, chart image classification, convolutional neural networks, computational modeling, chart recognition

Abstract

Data visualization is developed from the need to display a vast quantity of information more transparently. Data visualization often incorporates important information that is not listed anywhere in the document and enables the reader to discover significant data and save it in longer-term memory. On the other hand, Internet search engines have difficulty processing data visualization and connecting visualization and the request submitted by the user. With the use of data visualization, all blind individuals and individuals with impaired vision are left out. This article utilizes machine learning to classify data visualizations into 10 classes. Tested model is trained four times on the dataset which is preprocessed through four stages. Achieved accuracy of 89 % is comparable to other methods’ results. It is showed that image processing can impact results, i.e. increasing or decreasing level of details in image impacts on average classification accuracy significantly.

Downloads

Published

2020-04-15

How to Cite

[1]
F. Bajić, J. Job, and K. Nenadić, “Data Visualization Classification Using Simple Convolutional Neural Network Model”, IJECES, vol. 11, no. 1, pp. 43-51, Apr. 2020.

Issue

Section

Original Scientific Papers