A multiple kernel learning based fusion for earthquake detection from multimedia twitter data

• Published in: Multimedia tools and applications Vol. 77; no. 10; pp. 12519 - 12532
• Main Authors: Alqhtani, Samar M., Luo, Suhuai, Regan, Brian
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-05-2018; Springer Nature B.V
• Subjects: Accuracy; Computer Communication Networks; Computer Science; Data integration; Data Structures and Information Theory; Earthquakes; Feature extraction; Machine learning; Multimedia; Multimedia Information Systems; Multisensor fusion; Social networks; Special Purpose and Application-Based Systems; Multimedia data; Data fusion; Event detection; Data mining; Kernel method
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-017-4901-9

An efficient way of extracting useful information from multiple sources of data is to use data fusion technology. This paper introduces a data fusion approach in multimedia data for earthquake detection in twitter by using kernel fusion. The fusion method applies to fuse two types of data. The first type is features extracted from text by using bag-of-words method which is based on the calculation of the term frequency-inverse document frequency. The second type is the visual features extracted from images by applying scale-invariant feature transform. A multiple kernel fusion is applied in order to fuse the information from these two sources. Our experiments have indicated that comparing to the approaches using single data source, the proposed approach of using multiple kernel learning algorithm as early fusion increased the accuracy for earthquake detection. Experimental results for the proposed method achieved a high accuracy of 0.94, comparing to accuracy of 0.89 with texts only, and accuracy of 0.83 with images only.