TY - GEN
T1 - Holistic analysis of multi-source, multi-feature data
T2 - 5th International Conference on Big Data Analytics, BDA 2017
AU - Santra, Abhishek
AU - Bhowmick, Sanjukta
PY - 2017
Y1 - 2017
N2 - As a result of our increased ability to collect data from different sources, many real-world datasets are increasingly becoming multi-featured and these features can also be of different types. Examples of such multi-feature data include different modes of interactions among people (Facebook, Twitter, LinkedIn,..) or traffic accidents associated with diverse factors (speed, light conditions, weather,..). Efficiently modeling and analyzing these complex datasets to obtain actionable knowledge presents several challenges. Traditional approaches, such as using single layer networks (or monoplexes) may not be sufficient or appropriate for modeling and computation scalability. Recently, multiplexes have been proposed for the elegant handling of such data. In this position paper, we elaborate on different types of multiplexes (homogeneous, heterogeneous and hybrid) for modeling different types of data. The benefits of this modeling in terms of ease, understanding, and usage are highlighted. However, this model brings with it a new set of challenges for its analysis. The bulk of the paper discusses these challenges and the advantages of using this approach. With the right tools, both computation and storage can be reduced in addition to accommodating scalability.
AB - As a result of our increased ability to collect data from different sources, many real-world datasets are increasingly becoming multi-featured and these features can also be of different types. Examples of such multi-feature data include different modes of interactions among people (Facebook, Twitter, LinkedIn,..) or traffic accidents associated with diverse factors (speed, light conditions, weather,..). Efficiently modeling and analyzing these complex datasets to obtain actionable knowledge presents several challenges. Traditional approaches, such as using single layer networks (or monoplexes) may not be sufficient or appropriate for modeling and computation scalability. Recently, multiplexes have been proposed for the elegant handling of such data. In this position paper, we elaborate on different types of multiplexes (homogeneous, heterogeneous and hybrid) for modeling different types of data. The benefits of this modeling in terms of ease, understanding, and usage are highlighted. However, this model brings with it a new set of challenges for its analysis. The bulk of the paper discusses these challenges and the advantages of using this approach. With the right tools, both computation and storage can be reduced in addition to accommodating scalability.
KW - Aggregation functions
KW - Big data analytics
KW - Graph analysis and query processing
KW - Lossless composability
KW - Multi-source, disparate data
KW - Multiplex
UR - http://www.scopus.com/inward/record.url?scp=85038213112&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85038213112&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-72413-3_4
DO - 10.1007/978-3-319-72413-3_4
M3 - Conference contribution
AN - SCOPUS:85038213112
SN - 9783319724126
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 59
EP - 68
BT - Big Data Analytics - 5th International Conference, BDA 2017, Proceedings
A2 - Sureka, Ashish
A2 - Chakravarthy, Sharma
A2 - Reddy, P. Krishna
A2 - Bhalla, Subhash
PB - Springer Verlag
Y2 - 12 December 2017 through 15 December 2017
ER -