A Convolution Neural Network-Based Representative Spatio-Temporal Documents Classification for Big Text Data

With the proliferation of mobile devices, amount social media users and online news articles are rapidly increasing, text information is accumulating as big data. As spatio-temporal becomes more important, research on extracting spatiotemporal from data utilizing it for event analysis being actively conducted. However, if that does not describe core subject a document extracted, rather difficult to guarantee accuracy analysis. Therefore, important extract describes topic document. In this study, describing defined ‘representative information’, documents containing representative documents’. We proposed character-level Convolution Neuron Network (CNN)-based classifier classify documents. To train CNN model, 7400 training were constructed The experimental results show model outperforms traditional machine learning classifiers existing CNN-based classifiers.