TY - GEN
T1 - Tracking multiple social media for stock market event prediction
AU - Jin, Fang
AU - Wang, Wei
AU - Chakraborty, Prithwish
AU - Self, Nathan
AU - Chen, Feng
AU - Ramakrishnan, Naren
N1 - Publisher Copyright:
© Springer International Publishing AG 2017.
PY - 2017
Y1 - 2017
N2 - The problem of modeling the continuously changing trends in finance markets and generating real-time, meaningful predictions about significant changes in those markets has drawn considerable interest from economists and data scientists alike. In addition to traditional market indicators, growth of varied social media has enabled economists to leverage micro- and real-time indicators about factors possibly influencing the market, such as public emotion, anticipations and behaviors. We propose several specific market related features that can be mined from varied sources such as news, Google search volumes and Twitter. We further investigate the correlation between these features and financial market fluctuations. In this paper, we present a Delta Naive Bayes (DNB) approach to generate prediction about financial markets. We present a detailed prospective analysis of prediction accuracy generated from multiple, combined sources with those generated from a single source. We find that multi-source predictions consistently outperform single-source predictions, even though with some limitations.
AB - The problem of modeling the continuously changing trends in finance markets and generating real-time, meaningful predictions about significant changes in those markets has drawn considerable interest from economists and data scientists alike. In addition to traditional market indicators, growth of varied social media has enabled economists to leverage micro- and real-time indicators about factors possibly influencing the market, such as public emotion, anticipations and behaviors. We propose several specific market related features that can be mined from varied sources such as news, Google search volumes and Twitter. We further investigate the correlation between these features and financial market fluctuations. In this paper, we present a Delta Naive Bayes (DNB) approach to generate prediction about financial markets. We present a detailed prospective analysis of prediction accuracy generated from multiple, combined sources with those generated from a single source. We find that multi-source predictions consistently outperform single-source predictions, even though with some limitations.
KW - Features combination
KW - Google trends
KW - Market prediction
KW - Multiple social media
KW - News sentiment
KW - Twitter burst
UR - http://www.scopus.com/inward/record.url?scp=85025137026&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-62701-4_2
DO - 10.1007/978-3-319-62701-4_2
M3 - Conference contribution
AN - SCOPUS:85025137026
SN - 9783319627007
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 16
EP - 30
BT - Advances in Data Mining
A2 - Perner, Petra
PB - Springer-Verlag
T2 - 17th Industrial Conference on Advances in Data Mining, ICDM 2017
Y2 - 12 July 2017 through 13 July 2017
ER -