VIVA-Tech International Journal for Research and Innovation Volume 1, Issue 5 (2022)

ISSN(Online):2581-7280
VIVA Institute of Technology
10th National Conference on Role of Engineers in Nation Building – 2022 (NCRENB-2022)

“SENTIMENT ANALYSIS OF STOCK NEWS USING NLTK”

Hardik Raut1, Rohini Hodge2, Harshal Bhoir3, Reema Yedge4

(Electronics & Telecommunication, Viva Institute of Technology, India)

Abstract: The popular theory about stock prediction is Efficient Market Hypothesis. Due to its failure many
research has been done in area of stocks prediction. In this research non quantifiable data such as news about
financial activities of a company is taken to predict its future stock trend with the sentiment of the news. As news
have a strong impact on stock values, this study aims at relationship of company news with stock trend. For this,
we created a NLTK model which depicts polarities of news article as negative or positive. To evaluate various
aspects of the proposed model experiments were conducted. The NLTK model is accurate and this tool is capable
of determining the emotional values without neutral sections. By comparing the results of experiments with the
movement of stock market values in the same time periods, we can establish the relation between change occurred
in the stock values with sentiment analysis of economic news headlines. The model has prediction accuracy of
more than 80% and in comparison, with 50% of accuracy with news random labelling; thereby increasing the
model accuracy by 30%.
Keywords - NLTK, VADER LEXICON, BERT, Sentiment analysis , Stocks

1 Introduction
The goal is to develop and use a model to sentiment prediction by making connections between news articles and
marking them with negative or positive sentiments. There square measure several opportunities currently to
perform sentiment analyses, for instance external services that square measure nearly fully able to use it during a
given context wherever it's required like TextBlob. additionally, there square measure choices that enable us to
make our own models, train them supported our own knowledge. Sentiment analysis with BERT is one among
the foremost powerful tools that we are able to use, however we are able to conjointly produce a repeated Neural
Network (RNN) furthermore or use the VADER Lexicon with NLTK tool and SentimentIntensityAnalyzer[1].

The share/stock market is one among the foremost vital economic participants. many folks try and interpret and
outline the various share market movements in some ways. during this article, we have a tendency to use totally
different tool to the sentiments analysis, particularly absorption on the economic or financial news, however in
terms of news, absorption solely on the headline of economic news. In today’s communication and news
utilization, the headlines of varied articles play an excellent additional vital role than before. Now, we have a
tendency to use sentiment analysis on the headlines of a selected company or corporations to see the consequences
of the headlines to the stock market. The question arises what proportion result has the economic headline while
not the economic news whole context, if it's any measurable result in any respect. we've got found that it extremely
has. Thus, we have a tendency to outline the various impacts and their proper significance with an awfully specific
and distinctive new approach.

1
www.viva-technology.org/New/IJRI
VIVA-Tech International Journal for Research and Innovation Volume 1, Issue 5 (2022)
ISSN(Online):2581-7280
VIVA Institute of Technology
10th National Conference on Role of Engineers in Nation Building – 2022 (NCRENB-2022)

Data is a crucial pillar of research. Primarily the headlines of economic news square measure required, what we
have a tendency to use for sentiment analysis. Secondary, totally different stock market knowledge are required
supported corporations. There square measure several potentialities for knowledge assortment and analysis from
‘conventional’ dictionary-based performed by humans to ‘more serious’ neural network that verify the sign of the
headlines of every economic news and label with applicable emotional polarity. within the case of stock market
knowledge, various tools square measure on the market to get stock market knowledge which might be even
company-specific that is vital to us. In each case, we have a tendency to work with the foremost up-to-date
knowledge as doable, supported the data provided by the businesses. Both, the headlines of the economic news
and stock price knowledge square measure associated with the fundamental measure that specified by the news.
So, the result of the given sentiment analysis and therefore the vary of stock market knowledge are applicable.

The analysis is often separated to consecutive sections, first collecting headlines of financial economic news
supported corporations and collect stock market knowledge as per the timestamps of the given economic news
headlines. Then prepare this knowledge and apply sentiment analysis tool NLTK and VADER Lexicon. Manage
this knowledge and compare the stock market knowledge and emotional knowledge with image and
rationalization. gift however the headlines of economic news will have an effect on totally different stock market
changes and therefore the public[9].

2 Literature Review
Yu et al [2] demonstrated a text mining based mostly framework to work out the sentiment of articles and illustrate
its impact on energy demand. News sentiment is quantified as a time series and compared with fluctuations in
energy demand and costs.

Khedr and Yaseen[6] aims at having an efficient model to predict stock market further trends with tiny error
quantitative relation and improve the prediction accuracy. wherever this prediction model is predicted on
sentiment analysis and historical stock market costs, worked with K-NN and naïve mathematician rule to earn the
ultimate results. we are able to separate the model for 2 stages. the primary stage is to see the news polarity is
positive or negative mistreatment naïve mathematician rule, the second stage incorporates the output of the
primary stage as input with the processed historical numeric knowledge to predict the long run stock trend of K-
NN rule.

J. Bean [3] used keywords tagging regarding airlines satisfaction on Twitter feeds to attain them for polarity and
sentiment. this may offer a fast plan of the sentiment prevailing regarding airlines and their client satisfaction
ratings. we've got used the sentiment detection algorithmic program based mostly on this analysis.

Wang et al. [7] introduced a public sentiment analysis throughout the irruption that is ready to provides perceptive
info in creating applicable response of public health. They analyse the Sina Weibo standard social media website
posts of China, ‘wherever the unsupervised BERT model is adopted to classify sentiment classes (positive, neutral,
and negative) and TF-IDF (term frequency-inverse document frequency) model is employed to summarize the
topics of posts’. Analysing social media posts with negative sentiment may contribute to understanding the
experiences and offers examples for different countries. The analysis offer insights on the changing of social
sentiment over time and therefore the topic themes connected to negative sentiment on the social media sites.
Sizable accuracy was achieved with TF-IDF topic extraction model and BERT classification model.

SmartSA is a lexicon-based sentiment analysis for social media. It integrates methods to capture discourse polarity
from 2 ways that, the interaction of terms with their matter neighbourhood and text genre like native and
international context. They conjointly introduce associate approach to breed a general lexicon, with genre-specific
jargons and sentiment. The results from numerous social media show that these methods of native and
international contexts considerably improve sentiment classification, and are complementary together.

Kalyani, Bharathi & Rao[4] in their analysis, used supervised machine learning for classification of headlines
and extra text mining techniques to look at news polarity. The news article with its polarity score and text reborn
2
www.viva-technology.org/New/IJRI
VIVA-Tech International Journal for Research and Innovation Volume 1, Issue 5 (2022)
ISSN(Online):2581-7280
VIVA Institute of Technology
10th National Conference on Role of Engineers in Nation Building – 2022 (NCRENB-2022)

to tf-idf vector area area unit fed to the classifier. 3 completely different classification algorithms (Support Vector
Machines “SVM”, Naïve Bayes and Random Forest) area unit enforced to analyze and enhance classification
accuracy. Results of all 3 algorithms area unit compared supported preciseness, recall, accuracy, and different
model analysis techniques. once evaluating the results of all classifiers, the SVM classifier performs satisfactorily
for unknown information. The Random Forest conjointly showed higher results when put next to the Naïve Bayes
algorithmic program. Finally, a graph of link between news articles and stock information is plotted.

Streaming knowledge encourage be an upscale supply of information analysis wherever data square measure
collected in Realtime[4]. The most important characteristics of such knowledge being its accessibility and
accessibility, facilitate in correct analysis and prediction. Robert et al.[8] show associate analysis that has been
created for creating money selections like stock market prediction, to predict the potential costs of a company’s
stock mistreatment twitter knowledge.

3 Methodology
As earlier mentioned, the most goal within the economic news headlines is to use the foremost upto- date
knowledge. All knowledge assortment and management are machine-driven. there's associate choice to the user
to mention the portal as a supply to manage the news. we have a tendency to used knowledge from ‘finviz.com’
for our analyses. Before grouping the info, it's doable to enter the stock market names of the businesses wherever
we might wish to collect the info of current economic events for analysis. It is possible to specify over one
company by listing as parameter. The task takes care of managing the suitable timestamps (news publication time)
and separating the news supported the businesses and build a csv backup file. This freshly compiled knowledge
is employed by the appliance for any analysis (as a part of sentiment analysis, comparisons, and different
potentialities.) it's vital to say that news timestamps play a task in collection extra stock market knowledge that
the analyses come about within the same fundamental measure. Thus, these economic news headlines outline the
interval for later stock market knowledge assortment separated for corporations.

4 Sentiment analysis
In the case of sentiment analysis, the headline of the economic news from every company is labelled to what
sentiment price it carries, and therefore the polarity price is additionally indicated. With the assistance of those
knowledge, we are able to build analyses and predict the result.

Figure 1. Part from the economic news headlines dataframe.

4.1. NLTK -- VADER lexicon

3
www.viva-technology.org/New/IJRI
VIVA-Tech International Journal for Research and Innovation Volume 1, Issue 5 (2022)
ISSN(Online):2581-7280
VIVA Institute of Technology
10th National Conference on Role of Engineers in Nation Building – 2022 (NCRENB-2022)

NLTK is abbreviation for Language Toolkit. This toolkit is one among the foremost powerful information
processing libraries that contains packages to create machines perceive human language associated reply to that
with an applicable response. Our main focus is to analyse the tendency of sentiments using sentiments Intensity
analyzer. The polarity price of the sentence’s scales between -1 and one rather like within the TextBlob. The
method for labelling info (positive, negative or neutral) and previous tool which was used are sort of similar. we
have a tendency to use VADER Lexicon during this section. VADER (Valence Aware wordbook and sEntiment
Reasoner) uses the tool specifically attuned to sentiments expressed in social media, and works well on texts from
different domains based on lexicon and rule-based sentiment analysis tool that's specifically.

Figure 2 shows that the score of neutral value dominates all the cases of the sentiments result company wise.
Significant values were neutral as obtained from the analysis of economic news headlines. Neutral values This
level of neutral values has effect on analyses and comparisons of following changes in stock market. The results
from NLTK compared to TextBlob, shows the significant reduction in neutral values and so the results are
expected to be correct and realistic results with very less neutral values. In 51.50 % of the overall result's neutral
additionally to 50 % positive and 17 % negative. From the negative and positive groups, more positive are seen
to dominate, but the presence of neutral scores can produce uncertainty in the results.

The Figure 3 shows the outcome graph of scores separated by days interval. The results combined to provides us
a normalized score of the negative or positive the news for the company was. The polarity cannot be neutral in
total as the news of one polarity tries to move the neutral values some direction depending upon the polarity of
news. Thus, the subsequent figure is made, wherever on below of zero suggests that the negative section and top
of zero suggests that the positive section.

Figure 2. Results of sentiment analysis using NLTK and VADER as % positive, negative or neutral

4
www.viva-technology.org/New/IJRI
VIVA-Tech International Journal for Research and Innovation Volume 1, Issue 5 (2022)
ISSN(Online):2581-7280
VIVA Institute of Technology
10th National Conference on Role of Engineers in Nation Building – 2022 (NCRENB-2022)

Figure 3. Result of NLTK and VADER Lexicon Analysis days wise.

Conclusion
In this study, we classified the news headlines using sentiment analysis tools which is capable of analysing
emotions and classify different companies’ news headlines to examine the effect of sentiments on the price of
stocks in stock exchange. The emotions were segregated into the neutral, negative, positive classes. Neutral classes
in NLTK VADER lexicon were very less as compared to other tools used for sentiment analysis as seen in the
literature. Neutral values tend to produce variation in results when compared to the actual behaviour of stocks
prices. The emotional effect of the news on stock price is significant and so the results of sentiments analysis is
comparable to the exchange values for the given script. The positive values indicated that there might be increase
in the price of stocks, negative value indicated decrease in stock price, the neutral values have no effect on stock
price. The NLTK Vader Lexicon produces results, with very less model size. Since stock prices depends on
various factors other than news sentiment, direct correlation between headlines scores and stock price variation is
difficult to establish. Future work includes use of more sentiment analysis tool and natural language processing
to predict stock behaviour and match the prediction real prices of stock from the stock exchange.

References

[1] Anurag Nagar, Michael Hahsler, Using Text and Data Mining Techniques to extract Stock Market Sentiment from Live News Streams,
IPCSIT vol. XX (2012) IACSIT Press, Singapore

[2] W.B. Yu, B.R. Lea, and B. Guruswamy, A Theoretic Framework Integrating Text Mining and Energy Demand Forecasting, International
Journal of Electronic Business Management. 2011, 5(3): 211-224

[3] J. Bean, R by example: Mining Twitter for consumer attitudes towards airlines, In Boston Predictive Analytics Meetup Presentation, 2011

5
www.viva-technology.org/New/IJRI
VIVA-Tech International Journal for Research and Innovation Volume 1, Issue 5 (2022)
ISSN(Online):2581-7280
VIVA Institute of Technology
10th National Conference on Role of Engineers in Nation Building – 2022 (NCRENB-2022)

[4] Yauheniya Shynkevich, T.M. McGinnity, Sonya Coleman, Ammar Belatreche, Predicting Stock Price Movements Based on Different
Categories of News Articles, 2015 IEEE Symposium Series on Computational Intelligence

[5] P. Hofmarcher, S. Theussl, and K. Hornik, Do Media Sentiments Reflect Economic Indices? Chinese Business Review. 2011, 10(7): 487-
492

[6] Khedr, A. E., & Yaseen, N. (2017). Predicting stock market behavior using data mining technique and news sentiment analysis.
International Journal of Intelligent Systems and Applications, 9(7), 22. https://doi.org/10.5815/ijisa

[7] Wang, T., Lu, K., Chow, K. P., & Zhu, Q. (2020). COVID-19 sensing: Negative sentiment analysis on social media in China via bert
model. IEEE Access, 8, 138162–138169. https://doi.org/10.1109/Access.6287639

[8] Robert P. Schumaker, Yulei Zhang, Chun-Neng Huang, Sentiment Analysis of Financial News Articles

[9] Billah, M., Waheed, S., & Hanifa, A (2016, December 8–10). Stock market prediction using an improved training algorithm of neural
network. 2nd International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE), Rajshahi, Bangladesh (pp. 1–
4). IEEE. https://doi.org/10.1109/ICECTE.2016.7879611

6
www.viva-technology.org/New/IJRI