[논문]속성선택방법을 이용한 전기자동차 소셜미디어 데이터의 감성분석 연구

프란시스 조셉 코스텔로; 이건창

doi:10.14400/jdc.2020.18.2.249

속성선택방법을 이용한 전기자동차 소셜미디어 데이터의 감성분석 연구
Exploring the Sentiment Analysis of Electric Vehicles Social Media Data by Using Feature Selection Methods 원문보기

디지털융복합연구 = Journal of digital convergence, v.18 no.2, 2020년, pp.249 - 259

프란시스 조셉 코스텔로 (성균관대학교 경영대학) , 이건창 (성균관대학교 글로벌경영학과)

초록
AI-Helper

본 연구는 전기자동차(EV)에 대한 소셜미디어 데이터를 기반으로 감성분석 (SA)과 속성선택 (FS)방법을 적용하여 전기자동차에 대한 일반 사람들의 의견을 보다 효과적이고 정확히 예측할 수 있는 새로운 방법론을 제안한다. 구체적인 방법은 다음과 같다. 첫째, 유튜브에 있는 전기자동차에 대한 일반 사람들의 의견을 추출하였다. 둘째, 분석의 효과성을 증대하기 위하여 카이 스퀘어, 정보획득량, 릴리프에프 등 세가지 속성선택 방법을 적용하였다. 그 결과 로지스틱 회귀분석 및 서포트 벡터 머신 분류 기법에서 가장 의미있는 결과를 얻을 수 있다는 것이 확인되었다.

Abstract ▼ AI-Helper

This study presents a recently obtained social media data set based upon the case study of Electric Vehicles (EV) and looks to implement a sentiment analysis (SA) in order to gain insights. This study uses two methods in order to fully analyze the public's sentiment on EVs. First, we implement a SA tool in which we used to extract the sentiment of comments. Next we labeled the data with these sentiments obtained and classified them. While performing classification we found the problem of dimensionality and also explored the use of feature selection (FS) models in order to reduce the data set's dimensionality. We found that the use of three FS models (Chi Squared, Information Gain and ReliefF) showed the most promising results when used alongside a logistic and support vector machines classification algorithm. the contributions of this paper are in providing an real-world example of social media text analytics which can be adopted in many other areas of research and business. Moving forward researchers can use the methodological approach in this paper to further refine and improve their own case uses in text analytics.

주제어

표/그림 (6)

그림 Fig. 1. Sentiment of electric vehicles in 2018 based on the YouTube data set. Blue represents the negative comments, orange the neutral, and green the positive.
표 Table 1. This Table presents the original number of features of the data set and the number of features after each FS method was implemented.
표 Table 2. Definition of the classifications models seen within this study [19].
표 Table 3. Result of classification task whereby no feature selection was used on the data set. These results act as the null hypothesis in which we use to benchmark the selected feature selection.
표 Table 4. In this table we show the results from implementing the single and ensemble classification models for both the CFS and RFF FS. t-Test is performed for both models against the original data set’s results seen in Table 3.
표 Table 5. In this table we show the results from implementing the single and ensemble classification models for both the IG and CHI FS. t-Test is performed for both models against the original data set’s results seen in Table 3.

AI 본문요약
AI-Helper

* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.

문제 정의

Reducing this ‘curse’ is the role of feature selection (FS) and also explored within this study. Overall, this study looks to explore these three important area of machine learning within a social media dataset (in this paper electric vehicles is the chosen case study) in order to provide useful implications to future researchers interested analyzing public sentiment of a target technology.
Furthermore, this paper has shown that FS models necessary within a high-dimensional social media data set. This research has presented FS and classification models that have good potential to perform in multi-class classifications tasks and therefore this study can be used as a reference for other practitioners in the field to look into the use of these methodologies in greater detail when undertaking social media data analytics.

제안 방법

Future research could look at experimenting with various setting that could potentially influence a greater result. In doing so future research could find a more appropriate model that could enhance the results seen in this paper and further the field of social media data analytics.
In this study we will implement four single classifiers: Decision Tree (DT), Naïve Bayes (NB), Support Vector Machines (SVM), and Logistic (Log).
In this study, we proposed the use of a newly acquired data set based on Electric Vehicle (EV) YouTube comments in a way to study the viability of sentiment analysis (SA) for an informal social media data set. Compared to prior research i.
[8] we also used K-Fold Cross Validation alongside a comparison of single classification algorithms Naïve Bayes (NB), Decision Tree (DT), Support Vector Machines (SVM), Logistic Regression (LR) and ensemble classification models Random Forrest (RF), Bagging (BA), Random Subspace (RS), AdaBoost (ADA). Specifically, in this study 10-Fold Cross Validation was implemented in each classification task in order to maximize the use of all the training set data. K-Fold Cross Validation is where data is partitioned into test and training sets.
In stage one, the search term “electric car” and “electric vehicle” was entered into YouTube’s search bar, with a timescale set from June 2018 to December 2018. This provided videos on EVs, which were then selected based on the criteria of relevance, comment themes, and comment size (total number of comments). Our reasoning for using YouTube comments was for the following: (1) Compared with Twitter which is limited to 240 characters, YouTube has no limitation on comment length.

대상 데이터

Term-Frequency Inverted Document Frequency algorithm was also applied to the text as well as the snowball stemmer and N-gram tokenizer. The final data set included 19,000 instances and 1508 attributes.

이론/모형

RFF when used alongside the SVM and LOG showed the most promising results. Both models achieved an accuracy of over 70% and we can reject the null hypothesis for all metrics when used with the LOG algorithm. For IG and CHI, similar results were seen whereby SVM and LOG showed the most promising results, however, no model was found to be completely significant in terms of all the metrics analyzed.
Carpenter [3] this paper looks to study EV social media data. In this paper we implemented the use of SentimentR [12]. After we implemented SA, we used this data to then classify the data based on the obtained SA results whereby positive, negative and neutral scores were produced.
Therefore, sentiment lexicons have started to adapt to deal with the deliberate nature of misspelled words and emoticons seen in social media [12]. In this study we implemented the use of SentimentR for our analysis. This package was selected due to its ability to give a text polarity score based on the use of lexicons as well as its ability to identify valence in sentences.
Next the FS was implemented with four common FS methods, namely ReliefF [20], Correlation-based FS [21], Information Gain [22] and Chi-squared [22] implemented for analysis. Furthermore, based on research undertook by da Silva, Hruschka and Hruschka Jr.

성능/효과

First, our SA showed that positive sentiment towards EVs was greater than negative sentiment. In this result it is clear to see that slightly positive comments are far greater than negative comments.
Additionally, when it comes to predicting this sentiment through classification techniques, we found that ReliefF, Information Gain and Chi-squared feature selection used alongside a logistic and support vector machines algorithm has the best accuracy in predicting the social media sentiment EVs . These results were further examined through a t -Test which showed these models to be significantly more accurate against the data set whereby no feature selection was implemented. With these results we have shown that when performing social media text analytics these using a wide range of models should be considered and with the findings from this paper developing future analysis based on social media data sets can be fruitfull when the significant models in this paper are considered.
These results were further examined through a t -Test which showed these models to be significantly more accurate against the data set whereby no feature selection was implemented. With these results we have shown that when performing social media text analytics these using a wide range of models should be considered and with the findings from this paper developing future analysis based on social media data sets can be fruitfull when the significant models in this paper are considered.

후속연구

Recommendations for the future are as follows. First, this research only explores the use of a YouTube data set for social media text analysis. Future research should look into encompassing a wider scoped data set that includes YouTube, Twitter, Facebook and other various social media platform sites [7, 8, 2].
Future research should look into encompassing a wider scoped data set that includes YouTube, Twitter, Facebook and other various social media platform sites [7, 8, 2]. Secondly, future research could expand the scope of model evaluation for the purpose of finding the best FS model for high-dimensional TM and the best classification model to be used alongside this FS. Such models seen in deep learning like CNN are a potnetial avenue to explore.
Carpenter [3] examined EVs through online web texts to great effect, however, a study on EV social media texts has yet to be fully explored. Therefore, the aim of this study is to explore the potential of text analytics based on a social media data set and attempt to find viable solutions in which can act as a starting point for other researchers to further develop more precise models based on their own case needs. This research thus differs from previous research attempts which have not used social media texts regarding the EV technology.

참고문헌 (28)

X. Tian, Y. Geng, S. Zhong, J. Wilson, C. Gao, W. Chen & H. Hao. (2018). A bibliometric analysis on trends and characters of carbon emissions from transport sector. Transportation Research Part D: Transport and Environment, 59(December 2017) 1-10. https://doi.org/10.1016/j.trd.2017.12.009

상세보기
W. He, X. Tian, R. Tao, W. Zhang, G. Yan & V. Akula. (2017). Application of social media analytics: A case of analyzing online hotel reviews. Online Information Review, 41(7), 921-935. https://doi.org/10.1108/OIR-07-2016-0201

상세보기
T. Carpenter (2015). Measuring and Mitigating Electric Vehicle Adoption Barriers. PhD thesis, Waterloo, Ontario.
J. Kim, M. Han, Y. Lee & Y. Park. (2016). Futuristic data-driven scenario building: Incorporating text mining and fuzzy association rule mining into fuzzy cognitive map. Expert Systems with Applications, 57, 311-323. https://doi.org/10.1016/j.eswa.2016.03.043

상세보기
J. Li & H. Liu. (2017). Challenges of Feature Selection for Big Data Analytics. IEEE Computer Society, (March), 9-15. https://doi.org/10.1109/MIS.2017.38
M. N. Injadat, F. Salo & A. B. Nassif. (2016). Data mining techniques in social media: A survey. Neurocomputing, 214, 654-670. https://doi.org/10.1016/j.neucom.2016.06.045

상세보기
B. Li, K. C. C. Chan, C. Ou & S. Ruifeng. (2017). Discovering public sentiment in social media for predicting stock movement of publicly listed companies. Information Systems, 69, 81-92. https://doi.org/10.1016/j.is.2016.10.001

상세보기
N. F. F. da Silva, E. R. Hruschka & E. R. Hruschka. (2014). Tweet sentiment analysis with classifier ensembles. Decision Support Systems, 66, 170-179. https://doi.org/10.1016/j.dss.2014.07.003

상세보기
H. Yuan, R. Y. K. Lau & W. Xu. (2016). The determinants of crowdfunding success: A semantic text analytics approach. Decision Support Systems, 91. https://doi.org/10.1016/j.dss.2016.08.001

상세보기
C. Dhaoui, C. M. Webster & L. P. Tan. (2017). Social media sentiment analysis: lexicon versus machine learning. Journal of Consumer Marketing, 34(6), 480-488. https://doi.org/10.1108/JCM-03-2017-2141

상세보기
A. Ortigosa, J. M. Martin & R. M. Carrol. (2014). Sentiment analysis in Facebook and its application to e-learning. Computers in Human Behavior, 31(1), 527-541. https://doi.org/10.1016/j.chb.2013.05.024

상세보기
T. W. Rinker. (2018). sentimentr: Calculate Text Polarity Sentiment version 2.6.1. Retrieved from. http://github.com/trinker/sentimentr
C. T. Tran, M. Zhang, P. Andreae, B. Xue & L. T. Bui. (2018). Improving performance of classification on incomplete data using feature selection and clustering. Applied Soft Computing Journal, 73, 848-861. https://doi.org/10.1016/j.asoc.2018.09.026

상세보기
M. Tutkan, M. C. Ganiz & S. Akyokus. (2016). Helmholtz principle based supervised and unsupervised feature selection methods for text mining. Information Processing and Management, 52(5), 885-910. https://doi.org/10.1016/j.ipm.2016.03.007

상세보기
K. Seddig, P. Jochem & W. Fichtner. (2017). Integrating renewable energy sources by electric vehicle fleets under uncertainty. Energy, 141, 2145-2153. https://doi.org/10.1016/j.energy.2017.11.140

상세보기
M. Neaimeh, S. D. Salisbury, G. A. Hill, P. T. Blythe, D. R. Scoffield & J. E. Francfort. (2017). Analysing the usage and evidencing the importance of fast chargers for the adoption of battery electric vehicles. Energy Policy, 108, 474-486. https://doi.org/10.1016/j.enpol.2017.06.033

상세보기
D. Connolly. (2017). Economic viability of electric roads compared to oil and batteries for all forms of road transport. EnergyStrategy Reviews. https://doi.org/10.1016/j.esr.2017.09.005
L. H. Bjornsson & S. Karlsson. (2017). Electrification of the two-car household: PHEV or BEV? Transportation Research Part C: Emerging Technologies, 85(October), 363-376. https://doi.org/10.1016/j.trc.2017.09.021

상세보기
I. H. Witten, E. Frank & M. A. Hall. (2011). Data Mining: Practical Machine Learning Tools and Techniques (3rd ed.). Burlington, MA: Morgan Kaufmann Publishers Inc. https://doi.org/10.1016/B978-0-12-374856-0.00001-8
M. Robnik-Sikonja & I. Kononenko. (2003). Theoretical and empirical analysis of ReliefF and RReliefF. Machine Learning, 53(1), 23-69. https://doi.org/10.1023/A:1025667309714
M. A. Hall. (1999). Correlation-based feature selection for machine learning.
R. J. Quinlan. (1986). Induction of decision trees. Machine Learning, 1(1), 81-106. https://doi.org/10.1007/BF00116251

상세보기
G. Wang, J. Sun, J. Ma, K. Xu & J. Gu (2014). Sentiment classification: The contribution of ensemble learning. DecisionSupport Systems, 57, 77-93. https://doi.org/10.1016/j.dss.2013.08.002

상세보기
R. Togo, K. Magota, T. Shiga, K. Hirata, I. Tsujino, M. Haseyama & T. Ogawa (2018). Cardiac sarcoidosis classification with deep convolutional neural network-based features using polar maps. Computers in Biology and Medicine, 104(August 2018), 81-86. https://doi.org/10.1016/j.compbiomed.2018.11.008
A. Onan & S. Korukoglu (2017). A feature selection model based on genetic rank aggregation for text sentiment classification. Journal of Information Science, 43(1), 25-38. https://doi.org/10.1177/0165551515613226

상세보기
F. Wang, T. Xu, T. Tang, M. Zhou & H. Wang (2017). Bilevel Feature Extraction-Based Text Mining for Fault Diagnosis of Railway Systems. IEEE Transactions on Intelligent Transportation Systems, 18(1), 49-58. https://doi.org/10.1109/TITS.2016.2521866

상세보기
L. M. Abualigah, A. T.Khader, M. A. Al-Betar, & O. A. Alomari. (2017). Text feature selection with a robust weight schemeand dynamic dimension reduction to text document clustering. Expert Systemswith Applications, 84, 24-36. https://doi.org/10.1016/j.eswa.2017.05.002

상세보기
F. J. Costello & K. C. Lee. (2019). Exploring the Performance of Synthetic Minority Over-sampling Technique (SMOTE) to Predict Good Borrowers in P2P Lending. Journal of Digital Convergence, 17(9), 71-78. https://doi.org/10.14400/JDC.2019.17.9.071

원문보기 상세보기

저자의 다른 논문 :

표제어: PCR

동의어: Packet Collision Rate

용어 설명 출처 목록 (6)

용어 설명: PCR은 세균 특이성이 있는 primer를 이용하여 적은 수의 세균이 있을지라도 쉽게 검출할 수 있는 유용한 방법이며, 이를 이용하여 구강 내 치면세균막이나 타액에서 직접 세균을 검출할 수 있게 되었다[8].

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증