Sentiment Classification in the Financial Domain Using v-SVM and Multi-Objective Optimisation

Sun Fan; Ammar Belatreche; SA Coleman; TM McGinnity; Y Li

doi:10.1109/SSCI.2015.134

Sentiment Classification in the Financial Domain Using v-SVM and Multi-Objective Optimisation

Sun Fan, Ammar Belatreche, SA Coleman, TM McGinnity, Y Li

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Online financial textual information containing a large amount of investor sentiment is growing rapidly and an effective solution to automate the sentiment classification of such large amounts of text would be extremely beneficial. A novel approach to sentiment classification is the application of multi-objective optimization combined with v-SVM to improve the overall accuracy and hence we present a Multi-Objective Genetic Algorithm (MOGA) based approach to automatically adjust the free parameters of a v-SVM classifier to optimise sentiment classification performance. The approach is implemented and tested using two online financial textual datasets and experimental results show that the overall classification accuracy has improved (4%-7%) compared with other baseline approaches.

Original language	English
Title of host publication	Unknown Host Publication
Publisher	IEEE
Pages	910-916
Number of pages	7
ISBN (Print)	978-1-4799-7560-0
DOIs	https://doi.org/10.1109/SSCI.2015.134
Publication status	Published (in print/issue) - 6 Dec 2015
Event	2015 IEEE Symposium Series on Computational Intelligence - Cape Town Duration: 6 Dec 2015 → …

Conference

Conference	2015 IEEE Symposium Series on Computational Intelligence
Period	6/12/15 → …

Keywords

Sentiment
Finance

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1109/SSCI.2015.134

Cite this

@inproceedings{33d93bd5fda94e69888ef706785e4392,

title = "Sentiment Classification in the Financial Domain Using v-SVM and Multi-Objective Optimisation",

abstract = "Online financial textual information containing a large amount of investor sentiment is growing rapidly and an effective solution to automate the sentiment classification of such large amounts of text would be extremely beneficial. A novel approach to sentiment classification is the application of multi-objective optimization combined with v-SVM to improve the overall accuracy and hence we present a Multi-Objective Genetic Algorithm (MOGA) based approach to automatically adjust the free parameters of a v-SVM classifier to optimise sentiment classification performance. The approach is implemented and tested using two online financial textual datasets and experimental results show that the overall classification accuracy has improved (4%-7%) compared with other baseline approaches.",

keywords = "Sentiment, Finance",

author = "Sun Fan and Ammar Belatreche and SA Coleman and TM McGinnity and Y Li",

year = "2015",

month = dec,

day = "6",

doi = "10.1109/SSCI.2015.134",

language = "English",

isbn = "978-1-4799-7560-0",

pages = "910--916",

booktitle = "Unknown Host Publication",

publisher = "IEEE",

address = "United States",

note = "2015 IEEE Symposium Series on Computational Intelligence ; Conference date: 06-12-2015",

}

TY - GEN

T1 - Sentiment Classification in the Financial Domain Using v-SVM and Multi-Objective Optimisation

AU - Fan, Sun

AU - Belatreche, Ammar

AU - Coleman, SA

AU - McGinnity, TM

AU - Li, Y

PY - 2015/12/6

Y1 - 2015/12/6

N2 - Online financial textual information containing a large amount of investor sentiment is growing rapidly and an effective solution to automate the sentiment classification of such large amounts of text would be extremely beneficial. A novel approach to sentiment classification is the application of multi-objective optimization combined with v-SVM to improve the overall accuracy and hence we present a Multi-Objective Genetic Algorithm (MOGA) based approach to automatically adjust the free parameters of a v-SVM classifier to optimise sentiment classification performance. The approach is implemented and tested using two online financial textual datasets and experimental results show that the overall classification accuracy has improved (4%-7%) compared with other baseline approaches.

AB - Online financial textual information containing a large amount of investor sentiment is growing rapidly and an effective solution to automate the sentiment classification of such large amounts of text would be extremely beneficial. A novel approach to sentiment classification is the application of multi-objective optimization combined with v-SVM to improve the overall accuracy and hence we present a Multi-Objective Genetic Algorithm (MOGA) based approach to automatically adjust the free parameters of a v-SVM classifier to optimise sentiment classification performance. The approach is implemented and tested using two online financial textual datasets and experimental results show that the overall classification accuracy has improved (4%-7%) compared with other baseline approaches.

KW - Sentiment

KW - Finance

U2 - 10.1109/SSCI.2015.134

DO - 10.1109/SSCI.2015.134

M3 - Conference contribution

SN - 978-1-4799-7560-0

SP - 910

EP - 916

BT - Unknown Host Publication

PB - IEEE

T2 - 2015 IEEE Symposium Series on Computational Intelligence

Y2 - 6 December 2015

ER -

Sentiment Classification in the Financial Domain Using v-SVM and Multi-Objective Optimisation

Abstract

Conference

Keywords

UN SDGs

Access to Document

Fingerprint

Cite this