TY - GEN
T1 - News and sentiment analysis of the European market with a hybrid expert weighting algorithm
AU - Creamer, Germán G.
AU - Ren, Yong
AU - Sakamoto, Yasuaki
AU - Nickerson, Jeffrey V.
PY - 2013
Y1 - 2013
N2 - This paper proposes a hybrid human machine system based on an expert weighting algorithm that combines the responses of both humans and machine learning algorithms. The general topic of the paper is the use of the crowd to interpret text, and the power of that interpretation to predict future events. This topic is addressed through an experiment, in which news sentiment is evaluated by crowds and experts in different configurations. Their classifications are used as training sets for machine learning algorithms, including one that weights both machine and human predictions. The testing is done based on Thomson Reuters news stories and the returns of the stocks mentioned right after the stories appear. The hybrid expert weighting algorithm forecasts asset returns similar to the different versions of the trained and crowd groups because it combines the best results of the machine learning algorithms with human answers. The forecast of the expert weighting algorithm does not always show the best performance in comparison with the other learning algorithms; however its performance is very similar to the best algorithm in most cases. From a cognitive perspective, the capacity of the expert weighting algorithm to select dynamically the best expert according to its previous performance is consistent with an evolving collective intelligence: the final decision is a combination of the best individual answers - some of these come from machines, and some from humans
AB - This paper proposes a hybrid human machine system based on an expert weighting algorithm that combines the responses of both humans and machine learning algorithms. The general topic of the paper is the use of the crowd to interpret text, and the power of that interpretation to predict future events. This topic is addressed through an experiment, in which news sentiment is evaluated by crowds and experts in different configurations. Their classifications are used as training sets for machine learning algorithms, including one that weights both machine and human predictions. The testing is done based on Thomson Reuters news stories and the returns of the stocks mentioned right after the stories appear. The hybrid expert weighting algorithm forecasts asset returns similar to the different versions of the trained and crowd groups because it combines the best results of the machine learning algorithms with human answers. The forecast of the expert weighting algorithm does not always show the best performance in comparison with the other learning algorithms; however its performance is very similar to the best algorithm in most cases. From a cognitive perspective, the capacity of the expert weighting algorithm to select dynamically the best expert according to its previous performance is consistent with an evolving collective intelligence: the final decision is a combination of the best individual answers - some of these come from machines, and some from humans
KW - Cognitive modeling
KW - Computational finance
KW - Crowdsourcing
KW - Machine learning
KW - Text analysis
UR - http://www.scopus.com/inward/record.url?scp=84893580951&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84893580951&partnerID=8YFLogxK
U2 - 10.1109/SocialCom.2013.61
DO - 10.1109/SocialCom.2013.61
M3 - Conference contribution
AN - SCOPUS:84893580951
SN - 9780769551371
T3 - Proceedings - SocialCom/PASSAT/BigData/EconCom/BioMedCom 2013
SP - 391
EP - 396
BT - Proceedings - SocialCom/PASSAT/BigData/EconCom/BioMedCom 2013
T2 - 2013 ASE/IEEE Int. Conf. on Social Computing, SocialCom 2013, the 2013 ASE/IEEE Int. Conf. on Big Data, BigData 2013, the 2013 Int. Conf. on Economic Computing, EconCom 2013, the 2013 PASSAT 2013, and the 2013 ASE/IEEE Int. Conf. on BioMedCom 2013
Y2 - 8 September 2013 through 14 September 2013
ER -