A systematic literature review on features of deep learning in big data analytics

• Main Authors: Hordri, Nur Farhana, Samar, Alireza, Yuhaniz, Siti Sophiayati, Shamsuddin, Siti Mariyam
• Format: Article
• Published: International Center for Scientific Research and Studies 2017
• Subjects: T Technology; QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/66166/; http://home.ijasca.com/data/documents/Vol_9_1_ID-19_Pg32-49.pdf

The aims of this study are to identify the existing features of DL approaches for using in BDA and identify the key features that affect the effectiveness of DL approaches. Method: A Systematic Literature Review (SLR) was carried out and reported based on the preferred reporting items for systematic reviews. 4065 papers were retrieved by manual search in four databases which are Google Scholar, Taylor & Francis, Springer Link and Science Direct. 34 primary studies were finally included. Result: From these studies, 70% were journal articles, 25% were conference papers and 5% were contributions from the studies consisted of book chapters. Five features of DL were identified and analyzed. The features are (1) hierarchical layer, (2) high-level abstraction, (3) process high volume of data, (4) universal model and (5) does not over fit the training data. Conclusion: This review delivers the evidence that DL in BDA is an active research area. The review provides researchers with some guidelines for future research on this topic. It also provides broad information on DL in BDA which could be useful for practitioners.