SlideShare ist ein Scribd-Unternehmen logo
1 von 4
Downloaden Sie, um offline zu lesen
26 | The Journal of the CFA Society of the UK | www.cfauk.org
Feature | Professional Investor
Web users’ interactions and commentaries in the social media
space can reflect current opinions, views and experiences, and
therefore contain helpful information for market research.
Consumers who profess stronger positive affinity with a certain
brand are likely to have a higher customer lifetime value, a
predictor of the net present value of profits from a customer
over the entire future relationship with him/her. The evidence
so far is too little to demonstrate consistent results. However,
this new avenue demands further investigation with advanced
statistical analysis and larger scale application.
INTRODUCTION
In the early days when internet search algorithms were being
developed, who could have imagined that search data could be
used to predict the future? Yet here we are in 2012 witnessing
it. Many organisations have found that data extracted from
specific searches can predict – or at least model the future.
The Bank of England (BoE) is just one organisation that is
convinced that appropriately interpreted search data can act as
an indicator of future economic trends. In June 2011, a team
of researchers from the BoE released a report illustrating how
Fernan Flores asks whether the analysis of tweets or other social media
postings could be a useful predictor of market movements, as it has been
demonstrated in the case of Google search data.
Tweet: “@cfauk – is it true that tweets
can help predict a stock price movement?”
The Journal of the CFA Society of the UK | www.cfauk.org | 27
Professional Investor | Feature
results extracted from Google search data could predict changes
in unemployment and even house prices.
Being not a fan of social media sites, I had never used
Twitter, a micro blogging site, until I read an article describing
it as the new Google. As a market research analyst and
consequently a fan of Google, I was intrigued and registered for
Twitter to see what the buzz was about.
Twitter has indeed a search function that allows anyone to
browse through tweets, postings or status updates, sometimes in
real time. In fact, the research results by seeking out key words
from tweets proved to be very useful when I undertook some
competitive intelligence work for a client to check about its
competitor’s customer service. This was quite a revelation.
Web users’ interactions and commentaries in the social media
space can reflect current opinions, views and experiences and
therefore contain helpful information for market research.
Could the analysis of tweets or other social media postings be
a useful predictor of market movements though, as it has been
demonstrated in the case of Google search data?
Derwent Capital, a company which was originally established
as a hedge fund that used consumer tweets in its trading strategy
but has now repositioned itself as a technology provider giving
traders and investor access to its proprietary platform, said in an
article published in August 2011 that based on its research and
testing of randomly selected unstructured data from Twitter that
its algorithm, which helps classify a tweet into a sentiment (e.g.
alert, vital, happy), helped predict movements in liquid stocks.
A similar strategy was replicated by the University of
Manchester and Indiana University in a research paper
(Bollen, Mao, and Zeng, 2010), showing that Twitter data
analysed for sentiment predicted around 87.6% of the
movements in the Dow Jones industrial average. The study
was based on an assumption used in behavioural finance,
which states that “financial decisions are significantly driven
by emotion and mood… therefore, [it is] reasonable to assume
that the public mood and sentiment can drive stock market
values as much as news.”
ANALYSIS
In order to explain unstructured tweets, many social media
monitoring and analytics (SMMA) firms like Derwent Capital
have developed algorithms that categorise tweets (or any social
media postings) as positive, neutral or negative. The tweets are
further classified so that words that express stronger emotions
are classified at the extreme ends of a Likert scale such as the
ones illustrated in Chart 1 above.
The hypothesis that social media can be a strong indicator of
financial performance is based on the principle that consumers
who profess stronger positive affinity with a certain brand will
have a higher customer lifetime value, a predictor of the net
present value of profits from a customer over the entire future
relationship with him/her. If a brand or organisation has more
customers with stronger positive (or less negative) affinity, it
should have a positive financial outlook, which is reflected
through a strong stock performance.
To prove this relationship at a basic level, I plotted the
proportion of positive and negative sentiments against
the closing stock price of Apple (see Figures 1 and 2) and
Microsoft (see Figures 3 and 4). Because of the volatility of
the data, especially the sentiments, I used the data’s three-day
moving average standardised with z-scores in order to
compare the movements in the stock price and the sentiments
more evenly.
Chart 1: Likert Scale
“Financial decisions are significantly
driven by emotion and mood…therefore,
it is reasonable to assume that the public
mood and sentiment can drive stock
market values as much as news.”
Apple annoys me!
I will never buy an
iPhone again.
My iPhone is
getting
problematic.
My iPhone is
working ok.
I enjoy using
my iPhone.
I love my new
iPhone! I strongly
recommend that
everyone buys one too!
1 2 3 4 5
Positive sentimentsNegative sentiments
28 | The Journal of the CFA Society of the UK | www.cfauk.org
Apple (Jan - Dec 2011)
Microsoft (Jan - Dec* 2011)
Microsoft (Jan - Dec* 2011)
Apple (Dec 2011 - Jan 2012)
Feature | Professional Investor
As can be seen in Figures 1 and 2, the correlation coefficient
between the stock price and sentiments is very weak for Apple
and actually counter-intuitive as the positive sentiments trend
is negatively correlated with the stock price.
For Microsoft, a relationship seems to exist especially for positive
sentiments. As highlighted in Figures 3 and 4, there are days when
either the positive or negative sentiments clearly moved along with
the changes in stock price (as highlighted by the blue vertical lines).
While the accuracy of the technology developed by SMMA
firms in data mining has considerably improved over the years,
removing spam or filtering only relevant information remains a
challenge with the best technology achieving only an accuracy
level of between 75%-85% and the majority achieving an
accuracy level of between 50%-60%.
A deep-dive analysis of Apple verbatims reveals that a
considerable number of statements analysed refers to either
apple, the fruit, or apple juice. It is therefore not surprising
that the relationship between the sentiments and Apple’s stock
price hardly exists at all.
In contrast, the data mining technology has more accurately
analysed Microsoft given the uniqueness of the brand as
a term. The resulting correlation for Microsoft over a year,
however, remains weak. It could be possible that while
verbatims for Apple include irrelevant information, analysis
for Microsoft may have excluded tweets that refer to Microsoft
but have been omitted because consumers may have used their
own jargon when spelling the brand or have unintentionally
misspelled it (e.g. MS, Macrosoft, Mikrosoft, Microsof).
Figures 3 and 4:
Microsoft full year 2011 (positive and negative sentiments)
Apple (Jan - Dec 2011)
Apple (Jan - Dec 2011)
Microsoft (Jan - Dec* 2011)
Microsoft (Jan - Dec* 2011)
Apple (Dec 2011 - Jan 2012)
Source: Yahoo! Finance and Twitter
Figures 3 and 4. Z-scores of Microsoft’s closing stock price in NASDAQ versus z-scores of positive sentiments (Figure 3) and
negative sentiments (Figure 4) from Twitter. (Note that z-scores of negative sentiments are shown in reverse order as a decrease
in negative sentiment is expected to have a positive impact while an increase in negative sentiment is expected to have a negative
impact on a stock’s performance.)
Figures 1 and 2:
Apple full year 2011 (positive and negative sentiments)
Apple (Jan - Dec 2011)
Apple (Jan - Dec 2011)
Microsoft (Jan - Dec* 2011)
Microsoft (Jan - Dec* 2011)
Apple (Dec 2011 - Jan 2012)
Source: Yahoo! Finance and Twitter
Figures 1 and 2. Z-scores of Apple’s closing stock price in NASDAQ versus z-scores of positive sentiments (Figure 1) and negative
sentiments (Figure 2) from Twitter. (Note that z-scores of negative sentiments are shown in reverse order as a decrease in negative
sentiment is expected to have a positive impact while an increase in negative sentiment is expected to have a negative impact on a
stock’s performance.)
Apple (Jan - Dec 2011)
Microsoft (Jan - Dec* 2011)
Microsoft (Jan - Dec* 2011)
Apple (Dec 2011 - Jan 2012)
r = -0.26
r = 0.39
r = 0.14
r = -0.26
I conducted regression analysis and made various
combinations of analysis accounting for potential lag,
comparing the weighted average score of all sentiments (i.e.
rating extremely positive statements a 5, a relatively positive
statement a 4, a neutral statement a 3, a relatively negative
statement a 2 and an extremely negative statement a 1) and
comparing the net sentiment (i.e. the resulting proportion of
sentiments when negative is deducted from positive) but none
of the resulting analysis proved that the sentiments have a
strong relationship with a brand’s stock price.
With some effort, I manually cleaned hundreds of Apple
tweets (i.e. removing tweets that refer to apple, the fruit, or
apple juice) from December 2011 until January 2012. The
resulting comparison as shown in Figure 5 illustrates that
tweets that are more accurately filtered can potentially be
more effective in predicting a brand’s stock price, achieving a
correlation coefficient of 0.85.
CONCLUSION
While manually cleaned Twitter sentiments, at least for Apple
in this example, shows that consumer sentiment movements
movements can have a strong correlation to a company’s
stock price movements, the evidence so far is too little to
demonstrate consistent results. Clearly, this new avenue
consisting of exploitating Twitter or other social media websites
demands further investigation with advanced statistical analysis
and application on a larger scale to ascertain the relationship
between the two data sets.
With the rapid progress of technology in this field, especially
with search algorithms becoming more and more clever, it
is likely that the capability to demonstrate a correlation will
improve across time.
Can this work for non-consumer brands (e.g. BHP Billiton)?
Can sentiments on brands really have an impact on the stock
price of the company that owns them (e.g. PG tips, Bovril
and Persil owned by Unilever)? Can tweets from non-English
speaking countries and consumers, which are continuously
increasing in share as a proportion of total global tweets,
weaken or strengthen the relationship between sentiments and
stock price? These are just a few of the questions that we have
not even begun to address. Yet as technology develops, this will
spread into other compatible areas, geographies and cultures.
Given these issues, using tweets or any social media data
for trading strategy needs further exploration to strengthen
the case for it. But perhaps, based on Everett Rogers’ theory
of “Diffusion of Innovation” this may not be necessary for
innovators and early adopters – the consumer segments
which adopt technology ahead of the rest of the population.
Given the speed of technological innovation in data mining,
combined with advanced statistical analysis, I am confident
that using social media as a highly reliable predictor of stock
price movements can be achieved much sooner than expected.
However, when this point happens and when everyone else
starts to use insights from tweet sentiments for trading, then
the opportunity for arbitrage will have disappeared. ■
The Journal of the CFA Society of the UK | www.cfauk.org | 29
Professional Investor | Feature
Profile
Fernan Flores
Fernan Flores is a freelance market research
analyst and director at Zapienza, a Canary
Wharf-based market research consulting
firm that specialises in the technology and
finance sectors, which he established after
completing his MBA degree from the
Cambridge Judge Business School. Apart
from the technology and finance sectors, he
also does a considerable amount of work in
the not-for-profit sector and specialises in the
deployment of technology to solve healthcare
issues in developing markets. He has passed
the Level I exam of the CFA Program and
is a member of the CFA UK marketing and
communications committee.
Source: Yahoo! Finance and Twitter
Figure 5:
Apple 2 months December 2011 - January 2012
Microsoft (Jan - Dec* 2011)
Microsoft (Jan - Dec* 2011)
Apple (Dec 2011 - Jan 2012)
Source: Yahoo! Finance and Twitter
Figure 5. Z-scores of Apple’s closing stock price in
NASDAQ versus z-scores of positive sentiments using
data that are further filtered manually.
r = -0.85

Weitere ähnliche Inhalte

Was ist angesagt?

FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...
FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...
FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...ijscai
 
BigData Analytics for UK Retail Sector
BigData Analytics for UK Retail SectorBigData Analytics for UK Retail Sector
BigData Analytics for UK Retail SectorChun-Kai (Ken) Huang
 
Stock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysisStock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysisjournal ijrtem
 
Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?
Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?
Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?Molly Gibbons (she/her)
 
2018 Edelman Trust Barometer_ Italian Launch
2018 Edelman Trust Barometer_ Italian Launch2018 Edelman Trust Barometer_ Italian Launch
2018 Edelman Trust Barometer_ Italian LaunchEdelman Italia
 

Was ist angesagt? (11)

Parallel session iv d4
Parallel session iv d4Parallel session iv d4
Parallel session iv d4
 
NLP journal paper
NLP journal paperNLP journal paper
NLP journal paper
 
Document(2)
Document(2)Document(2)
Document(2)
 
FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...
FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...
FORECASTING MACROECONOMICAL INDICES WITH MACHINE LEARNING: IMPARTIAL ANALYSIS...
 
BigData Analytics for UK Retail Sector
BigData Analytics for UK Retail SectorBigData Analytics for UK Retail Sector
BigData Analytics for UK Retail Sector
 
Stock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysisStock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysis
 
Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?
Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?
Are Positive or Negative Tweets More "Retweetable" in Brazilian Politics?
 
The Prattle Primer
The Prattle PrimerThe Prattle Primer
The Prattle Primer
 
muthu.shree
muthu.shreemuthu.shree
muthu.shree
 
2018 Edelman Trust Barometer_ Italian Launch
2018 Edelman Trust Barometer_ Italian Launch2018 Edelman Trust Barometer_ Italian Launch
2018 Edelman Trust Barometer_ Italian Launch
 
Analytics anecdotes
Analytics anecdotesAnalytics anecdotes
Analytics anecdotes
 

Ähnlich wie Can tweets help predict a stock's price movements?

Twitter sentimentanalysis report
Twitter sentimentanalysis reportTwitter sentimentanalysis report
Twitter sentimentanalysis reportSavio Aberneithie
 
customer behavior analysis for social media
customer behavior analysis for social mediacustomer behavior analysis for social media
customer behavior analysis for social mediaINFOGAIN PUBLICATION
 
Stock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysisStock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysisIJRTEMJOURNAL
 
The impact of sentiment analysis from user on Facebook to enhanced the servic...
The impact of sentiment analysis from user on Facebook to enhanced the servic...The impact of sentiment analysis from user on Facebook to enhanced the servic...
The impact of sentiment analysis from user on Facebook to enhanced the servic...IJECEIAES
 
How to Contextualize Data for Meaningful Insights
How to Contextualize Data for Meaningful Insights  How to Contextualize Data for Meaningful Insights
How to Contextualize Data for Meaningful Insights Virginia Bautista
 
Will facebook ever_drive_ecommerce
Will facebook ever_drive_ecommerceWill facebook ever_drive_ecommerce
Will facebook ever_drive_ecommercetzvitybo
 
Will facebook ever_drive_ecommerce-julio2011
Will facebook ever_drive_ecommerce-julio2011Will facebook ever_drive_ecommerce-julio2011
Will facebook ever_drive_ecommerce-julio2011Digital Pymes
 
Quantifying online buzz and its impact on product launch
Quantifying online buzz and its impact on product launchQuantifying online buzz and its impact on product launch
Quantifying online buzz and its impact on product launchGinvanglian Tombing
 
THE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNING
THE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNINGTHE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNING
THE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNINGIRJET Journal
 
Ryan Boehme Thesis Paper Final 8-11-2016
Ryan Boehme Thesis Paper Final 8-11-2016Ryan Boehme Thesis Paper Final 8-11-2016
Ryan Boehme Thesis Paper Final 8-11-2016Ryan Boehme
 
Consumer Segmentation with Bayesian Statistics
Consumer Segmentation with Bayesian StatisticsConsumer Segmentation with Bayesian Statistics
Consumer Segmentation with Bayesian StatisticsEsteban Ribero
 
Data Matrix Of Cpi Data Distribution After Transformation...
Data Matrix Of Cpi Data Distribution After Transformation...Data Matrix Of Cpi Data Distribution After Transformation...
Data Matrix Of Cpi Data Distribution After Transformation...Kimberly Jones
 
II-SDV 2012 Mining Opinion from Twitter
II-SDV 2012 Mining Opinion from TwitterII-SDV 2012 Mining Opinion from Twitter
II-SDV 2012 Mining Opinion from TwitterDr. Haxel Consult
 
{White Paper} Measuring Global Attention | Appinions
{White Paper} Measuring Global Attention | Appinions{White Paper} Measuring Global Attention | Appinions
{White Paper} Measuring Global Attention | AppinionsAppinions
 
Knime social media_white_paper
Knime social media_white_paperKnime social media_white_paper
Knime social media_white_paperFiras Husseini
 
Customers Sentiment on Life Insurance Industry
Customers Sentiment on Life Insurance IndustryCustomers Sentiment on Life Insurance Industry
Customers Sentiment on Life Insurance Industryzhongshu zhao
 
SENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATASENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATAanargha gangadharan
 
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATAREAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATAMary Lis Joseph
 

Ähnlich wie Can tweets help predict a stock's price movements? (20)

Twitter sentimentanalysis report
Twitter sentimentanalysis reportTwitter sentimentanalysis report
Twitter sentimentanalysis report
 
customer behavior analysis for social media
customer behavior analysis for social mediacustomer behavior analysis for social media
customer behavior analysis for social media
 
Stock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysisStock market prediction using Twitter sentiment analysis
Stock market prediction using Twitter sentiment analysis
 
Monitoring opinion on esop through social media and clustering its polarity
Monitoring opinion on esop through social media and clustering its polarityMonitoring opinion on esop through social media and clustering its polarity
Monitoring opinion on esop through social media and clustering its polarity
 
The impact of sentiment analysis from user on Facebook to enhanced the servic...
The impact of sentiment analysis from user on Facebook to enhanced the servic...The impact of sentiment analysis from user on Facebook to enhanced the servic...
The impact of sentiment analysis from user on Facebook to enhanced the servic...
 
How to Contextualize Data for Meaningful Insights
How to Contextualize Data for Meaningful Insights  How to Contextualize Data for Meaningful Insights
How to Contextualize Data for Meaningful Insights
 
Will facebook ever_drive_ecommerce
Will facebook ever_drive_ecommerceWill facebook ever_drive_ecommerce
Will facebook ever_drive_ecommerce
 
Will facebook ever drive ecommerce
Will facebook ever drive ecommerceWill facebook ever drive ecommerce
Will facebook ever drive ecommerce
 
Will facebook ever_drive_ecommerce-julio2011
Will facebook ever_drive_ecommerce-julio2011Will facebook ever_drive_ecommerce-julio2011
Will facebook ever_drive_ecommerce-julio2011
 
Quantifying online buzz and its impact on product launch
Quantifying online buzz and its impact on product launchQuantifying online buzz and its impact on product launch
Quantifying online buzz and its impact on product launch
 
THE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNING
THE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNINGTHE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNING
THE ANALYSIS FOR CUSTOMER REVIEWS THROUGH TWEETS, BASED ON DEEP LEARNING
 
Ryan Boehme Thesis Paper Final 8-11-2016
Ryan Boehme Thesis Paper Final 8-11-2016Ryan Boehme Thesis Paper Final 8-11-2016
Ryan Boehme Thesis Paper Final 8-11-2016
 
Consumer Segmentation with Bayesian Statistics
Consumer Segmentation with Bayesian StatisticsConsumer Segmentation with Bayesian Statistics
Consumer Segmentation with Bayesian Statistics
 
Data Matrix Of Cpi Data Distribution After Transformation...
Data Matrix Of Cpi Data Distribution After Transformation...Data Matrix Of Cpi Data Distribution After Transformation...
Data Matrix Of Cpi Data Distribution After Transformation...
 
II-SDV 2012 Mining Opinion from Twitter
II-SDV 2012 Mining Opinion from TwitterII-SDV 2012 Mining Opinion from Twitter
II-SDV 2012 Mining Opinion from Twitter
 
{White Paper} Measuring Global Attention | Appinions
{White Paper} Measuring Global Attention | Appinions{White Paper} Measuring Global Attention | Appinions
{White Paper} Measuring Global Attention | Appinions
 
Knime social media_white_paper
Knime social media_white_paperKnime social media_white_paper
Knime social media_white_paper
 
Customers Sentiment on Life Insurance Industry
Customers Sentiment on Life Insurance IndustryCustomers Sentiment on Life Insurance Industry
Customers Sentiment on Life Insurance Industry
 
SENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATASENTIMENT ANALYSIS OF TWITTER DATA
SENTIMENT ANALYSIS OF TWITTER DATA
 
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATAREAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
REAL TIME SENTIMENT ANALYSIS OF TWITTER DATA
 

Kürzlich hochgeladen

The top 4 AI cryptocurrencies to know in 2024 .pdf
The top 4 AI cryptocurrencies to know in 2024 .pdfThe top 4 AI cryptocurrencies to know in 2024 .pdf
The top 4 AI cryptocurrencies to know in 2024 .pdfJhon Thompson
 
AnyConv.com__FSS Advance Retail & Distribution - 15.06.17.ppt
AnyConv.com__FSS Advance Retail & Distribution - 15.06.17.pptAnyConv.com__FSS Advance Retail & Distribution - 15.06.17.ppt
AnyConv.com__FSS Advance Retail & Distribution - 15.06.17.pptPriyankaSharma89719
 
2024 Q1 Crypto Industry Report | CoinGecko
2024 Q1 Crypto Industry Report | CoinGecko2024 Q1 Crypto Industry Report | CoinGecko
2024 Q1 Crypto Industry Report | CoinGeckoCoinGecko
 
Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...
Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...
Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...Amil baba
 
NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...
NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...
NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...Amil baba
 
NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...
NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...
NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...Amil baba
 
Banking: Commercial and Central Banking.pptx
Banking: Commercial and Central Banking.pptxBanking: Commercial and Central Banking.pptx
Banking: Commercial and Central Banking.pptxANTHONYAKINYOSOYE1
 
『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书
『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书
『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书rnrncn29
 
Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...
Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...
Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...Amil baba
 
Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...
Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...
Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...beulahfernandes8
 
The Inspirational Story of Julio Herrera Velutini - Global Finance Leader
The Inspirational Story of Julio Herrera Velutini - Global Finance LeaderThe Inspirational Story of Julio Herrera Velutini - Global Finance Leader
The Inspirational Story of Julio Herrera Velutini - Global Finance LeaderArianna Varetto
 
Gender and caste discrimination in india
Gender and caste discrimination in indiaGender and caste discrimination in india
Gender and caste discrimination in indiavandanasingh01072003
 
Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...
Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...
Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...Amil baba
 
Hello this ppt is about seminar final project
Hello this ppt is about seminar final projectHello this ppt is about seminar final project
Hello this ppt is about seminar final projectninnasirsi
 
Overview of Inkel Unlisted Shares Price.
Overview of Inkel Unlisted Shares Price.Overview of Inkel Unlisted Shares Price.
Overview of Inkel Unlisted Shares Price.Precize Formely Leadoff
 
Global Economic Outlook, 2024 - Scholaride Consulting
Global Economic Outlook, 2024 - Scholaride ConsultingGlobal Economic Outlook, 2024 - Scholaride Consulting
Global Economic Outlook, 2024 - Scholaride Consultingswastiknandyofficial
 
Stock Market Brief Deck FOR 4/17 video.pdf
Stock Market Brief Deck FOR 4/17 video.pdfStock Market Brief Deck FOR 4/17 video.pdf
Stock Market Brief Deck FOR 4/17 video.pdfMichael Silva
 
Introduction to Health Economics Dr. R. Kurinji Malar.pptx
Introduction to Health Economics Dr. R. Kurinji Malar.pptxIntroduction to Health Economics Dr. R. Kurinji Malar.pptx
Introduction to Health Economics Dr. R. Kurinji Malar.pptxDrRkurinjiMalarkurin
 
Financial analysis on Risk and Return.ppt
Financial analysis on Risk and Return.pptFinancial analysis on Risk and Return.ppt
Financial analysis on Risk and Return.ppttadegebreyesus
 
Financial Preparation for Millennia.pptx
Financial Preparation for Millennia.pptxFinancial Preparation for Millennia.pptx
Financial Preparation for Millennia.pptxsimon978302
 

Kürzlich hochgeladen (20)

The top 4 AI cryptocurrencies to know in 2024 .pdf
The top 4 AI cryptocurrencies to know in 2024 .pdfThe top 4 AI cryptocurrencies to know in 2024 .pdf
The top 4 AI cryptocurrencies to know in 2024 .pdf
 
AnyConv.com__FSS Advance Retail & Distribution - 15.06.17.ppt
AnyConv.com__FSS Advance Retail & Distribution - 15.06.17.pptAnyConv.com__FSS Advance Retail & Distribution - 15.06.17.ppt
AnyConv.com__FSS Advance Retail & Distribution - 15.06.17.ppt
 
2024 Q1 Crypto Industry Report | CoinGecko
2024 Q1 Crypto Industry Report | CoinGecko2024 Q1 Crypto Industry Report | CoinGecko
2024 Q1 Crypto Industry Report | CoinGecko
 
Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...
Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...
Uae-NO1 Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...
 
NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...
NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...
NO1 Certified kala jadu karne wale ka contact number kala jadu karne wale bab...
 
NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...
NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...
NO1 Certified Black Magic Specialist Expert In Bahawalpur, Sargodha, Sialkot,...
 
Banking: Commercial and Central Banking.pptx
Banking: Commercial and Central Banking.pptxBanking: Commercial and Central Banking.pptx
Banking: Commercial and Central Banking.pptx
 
『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书
『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书
『澳洲文凭』买科廷大学毕业证书成绩单办理澳洲Curtin文凭学位证书
 
Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...
Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...
Uae-NO1 Rohani Amil In Islamabad Amil Baba in Rawalpindi Kala Jadu Amil In Ra...
 
Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...
Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...
Unveiling Poonawalla Fincorp’s Phenomenal Performance Under Abhay Bhutada’s L...
 
The Inspirational Story of Julio Herrera Velutini - Global Finance Leader
The Inspirational Story of Julio Herrera Velutini - Global Finance LeaderThe Inspirational Story of Julio Herrera Velutini - Global Finance Leader
The Inspirational Story of Julio Herrera Velutini - Global Finance Leader
 
Gender and caste discrimination in india
Gender and caste discrimination in indiaGender and caste discrimination in india
Gender and caste discrimination in india
 
Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...
Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...
Uae-NO1 Pakistani Amil Baba Real Amil baba In Pakistan Najoomi Baba in Pakist...
 
Hello this ppt is about seminar final project
Hello this ppt is about seminar final projectHello this ppt is about seminar final project
Hello this ppt is about seminar final project
 
Overview of Inkel Unlisted Shares Price.
Overview of Inkel Unlisted Shares Price.Overview of Inkel Unlisted Shares Price.
Overview of Inkel Unlisted Shares Price.
 
Global Economic Outlook, 2024 - Scholaride Consulting
Global Economic Outlook, 2024 - Scholaride ConsultingGlobal Economic Outlook, 2024 - Scholaride Consulting
Global Economic Outlook, 2024 - Scholaride Consulting
 
Stock Market Brief Deck FOR 4/17 video.pdf
Stock Market Brief Deck FOR 4/17 video.pdfStock Market Brief Deck FOR 4/17 video.pdf
Stock Market Brief Deck FOR 4/17 video.pdf
 
Introduction to Health Economics Dr. R. Kurinji Malar.pptx
Introduction to Health Economics Dr. R. Kurinji Malar.pptxIntroduction to Health Economics Dr. R. Kurinji Malar.pptx
Introduction to Health Economics Dr. R. Kurinji Malar.pptx
 
Financial analysis on Risk and Return.ppt
Financial analysis on Risk and Return.pptFinancial analysis on Risk and Return.ppt
Financial analysis on Risk and Return.ppt
 
Financial Preparation for Millennia.pptx
Financial Preparation for Millennia.pptxFinancial Preparation for Millennia.pptx
Financial Preparation for Millennia.pptx
 

Can tweets help predict a stock's price movements?

  • 1. 26 | The Journal of the CFA Society of the UK | www.cfauk.org Feature | Professional Investor Web users’ interactions and commentaries in the social media space can reflect current opinions, views and experiences, and therefore contain helpful information for market research. Consumers who profess stronger positive affinity with a certain brand are likely to have a higher customer lifetime value, a predictor of the net present value of profits from a customer over the entire future relationship with him/her. The evidence so far is too little to demonstrate consistent results. However, this new avenue demands further investigation with advanced statistical analysis and larger scale application. INTRODUCTION In the early days when internet search algorithms were being developed, who could have imagined that search data could be used to predict the future? Yet here we are in 2012 witnessing it. Many organisations have found that data extracted from specific searches can predict – or at least model the future. The Bank of England (BoE) is just one organisation that is convinced that appropriately interpreted search data can act as an indicator of future economic trends. In June 2011, a team of researchers from the BoE released a report illustrating how Fernan Flores asks whether the analysis of tweets or other social media postings could be a useful predictor of market movements, as it has been demonstrated in the case of Google search data. Tweet: “@cfauk – is it true that tweets can help predict a stock price movement?”
  • 2. The Journal of the CFA Society of the UK | www.cfauk.org | 27 Professional Investor | Feature results extracted from Google search data could predict changes in unemployment and even house prices. Being not a fan of social media sites, I had never used Twitter, a micro blogging site, until I read an article describing it as the new Google. As a market research analyst and consequently a fan of Google, I was intrigued and registered for Twitter to see what the buzz was about. Twitter has indeed a search function that allows anyone to browse through tweets, postings or status updates, sometimes in real time. In fact, the research results by seeking out key words from tweets proved to be very useful when I undertook some competitive intelligence work for a client to check about its competitor’s customer service. This was quite a revelation. Web users’ interactions and commentaries in the social media space can reflect current opinions, views and experiences and therefore contain helpful information for market research. Could the analysis of tweets or other social media postings be a useful predictor of market movements though, as it has been demonstrated in the case of Google search data? Derwent Capital, a company which was originally established as a hedge fund that used consumer tweets in its trading strategy but has now repositioned itself as a technology provider giving traders and investor access to its proprietary platform, said in an article published in August 2011 that based on its research and testing of randomly selected unstructured data from Twitter that its algorithm, which helps classify a tweet into a sentiment (e.g. alert, vital, happy), helped predict movements in liquid stocks. A similar strategy was replicated by the University of Manchester and Indiana University in a research paper (Bollen, Mao, and Zeng, 2010), showing that Twitter data analysed for sentiment predicted around 87.6% of the movements in the Dow Jones industrial average. The study was based on an assumption used in behavioural finance, which states that “financial decisions are significantly driven by emotion and mood… therefore, [it is] reasonable to assume that the public mood and sentiment can drive stock market values as much as news.” ANALYSIS In order to explain unstructured tweets, many social media monitoring and analytics (SMMA) firms like Derwent Capital have developed algorithms that categorise tweets (or any social media postings) as positive, neutral or negative. The tweets are further classified so that words that express stronger emotions are classified at the extreme ends of a Likert scale such as the ones illustrated in Chart 1 above. The hypothesis that social media can be a strong indicator of financial performance is based on the principle that consumers who profess stronger positive affinity with a certain brand will have a higher customer lifetime value, a predictor of the net present value of profits from a customer over the entire future relationship with him/her. If a brand or organisation has more customers with stronger positive (or less negative) affinity, it should have a positive financial outlook, which is reflected through a strong stock performance. To prove this relationship at a basic level, I plotted the proportion of positive and negative sentiments against the closing stock price of Apple (see Figures 1 and 2) and Microsoft (see Figures 3 and 4). Because of the volatility of the data, especially the sentiments, I used the data’s three-day moving average standardised with z-scores in order to compare the movements in the stock price and the sentiments more evenly. Chart 1: Likert Scale “Financial decisions are significantly driven by emotion and mood…therefore, it is reasonable to assume that the public mood and sentiment can drive stock market values as much as news.” Apple annoys me! I will never buy an iPhone again. My iPhone is getting problematic. My iPhone is working ok. I enjoy using my iPhone. I love my new iPhone! I strongly recommend that everyone buys one too! 1 2 3 4 5 Positive sentimentsNegative sentiments
  • 3. 28 | The Journal of the CFA Society of the UK | www.cfauk.org Apple (Jan - Dec 2011) Microsoft (Jan - Dec* 2011) Microsoft (Jan - Dec* 2011) Apple (Dec 2011 - Jan 2012) Feature | Professional Investor As can be seen in Figures 1 and 2, the correlation coefficient between the stock price and sentiments is very weak for Apple and actually counter-intuitive as the positive sentiments trend is negatively correlated with the stock price. For Microsoft, a relationship seems to exist especially for positive sentiments. As highlighted in Figures 3 and 4, there are days when either the positive or negative sentiments clearly moved along with the changes in stock price (as highlighted by the blue vertical lines). While the accuracy of the technology developed by SMMA firms in data mining has considerably improved over the years, removing spam or filtering only relevant information remains a challenge with the best technology achieving only an accuracy level of between 75%-85% and the majority achieving an accuracy level of between 50%-60%. A deep-dive analysis of Apple verbatims reveals that a considerable number of statements analysed refers to either apple, the fruit, or apple juice. It is therefore not surprising that the relationship between the sentiments and Apple’s stock price hardly exists at all. In contrast, the data mining technology has more accurately analysed Microsoft given the uniqueness of the brand as a term. The resulting correlation for Microsoft over a year, however, remains weak. It could be possible that while verbatims for Apple include irrelevant information, analysis for Microsoft may have excluded tweets that refer to Microsoft but have been omitted because consumers may have used their own jargon when spelling the brand or have unintentionally misspelled it (e.g. MS, Macrosoft, Mikrosoft, Microsof). Figures 3 and 4: Microsoft full year 2011 (positive and negative sentiments) Apple (Jan - Dec 2011) Apple (Jan - Dec 2011) Microsoft (Jan - Dec* 2011) Microsoft (Jan - Dec* 2011) Apple (Dec 2011 - Jan 2012) Source: Yahoo! Finance and Twitter Figures 3 and 4. Z-scores of Microsoft’s closing stock price in NASDAQ versus z-scores of positive sentiments (Figure 3) and negative sentiments (Figure 4) from Twitter. (Note that z-scores of negative sentiments are shown in reverse order as a decrease in negative sentiment is expected to have a positive impact while an increase in negative sentiment is expected to have a negative impact on a stock’s performance.) Figures 1 and 2: Apple full year 2011 (positive and negative sentiments) Apple (Jan - Dec 2011) Apple (Jan - Dec 2011) Microsoft (Jan - Dec* 2011) Microsoft (Jan - Dec* 2011) Apple (Dec 2011 - Jan 2012) Source: Yahoo! Finance and Twitter Figures 1 and 2. Z-scores of Apple’s closing stock price in NASDAQ versus z-scores of positive sentiments (Figure 1) and negative sentiments (Figure 2) from Twitter. (Note that z-scores of negative sentiments are shown in reverse order as a decrease in negative sentiment is expected to have a positive impact while an increase in negative sentiment is expected to have a negative impact on a stock’s performance.) Apple (Jan - Dec 2011) Microsoft (Jan - Dec* 2011) Microsoft (Jan - Dec* 2011) Apple (Dec 2011 - Jan 2012) r = -0.26 r = 0.39 r = 0.14 r = -0.26
  • 4. I conducted regression analysis and made various combinations of analysis accounting for potential lag, comparing the weighted average score of all sentiments (i.e. rating extremely positive statements a 5, a relatively positive statement a 4, a neutral statement a 3, a relatively negative statement a 2 and an extremely negative statement a 1) and comparing the net sentiment (i.e. the resulting proportion of sentiments when negative is deducted from positive) but none of the resulting analysis proved that the sentiments have a strong relationship with a brand’s stock price. With some effort, I manually cleaned hundreds of Apple tweets (i.e. removing tweets that refer to apple, the fruit, or apple juice) from December 2011 until January 2012. The resulting comparison as shown in Figure 5 illustrates that tweets that are more accurately filtered can potentially be more effective in predicting a brand’s stock price, achieving a correlation coefficient of 0.85. CONCLUSION While manually cleaned Twitter sentiments, at least for Apple in this example, shows that consumer sentiment movements movements can have a strong correlation to a company’s stock price movements, the evidence so far is too little to demonstrate consistent results. Clearly, this new avenue consisting of exploitating Twitter or other social media websites demands further investigation with advanced statistical analysis and application on a larger scale to ascertain the relationship between the two data sets. With the rapid progress of technology in this field, especially with search algorithms becoming more and more clever, it is likely that the capability to demonstrate a correlation will improve across time. Can this work for non-consumer brands (e.g. BHP Billiton)? Can sentiments on brands really have an impact on the stock price of the company that owns them (e.g. PG tips, Bovril and Persil owned by Unilever)? Can tweets from non-English speaking countries and consumers, which are continuously increasing in share as a proportion of total global tweets, weaken or strengthen the relationship between sentiments and stock price? These are just a few of the questions that we have not even begun to address. Yet as technology develops, this will spread into other compatible areas, geographies and cultures. Given these issues, using tweets or any social media data for trading strategy needs further exploration to strengthen the case for it. But perhaps, based on Everett Rogers’ theory of “Diffusion of Innovation” this may not be necessary for innovators and early adopters – the consumer segments which adopt technology ahead of the rest of the population. Given the speed of technological innovation in data mining, combined with advanced statistical analysis, I am confident that using social media as a highly reliable predictor of stock price movements can be achieved much sooner than expected. However, when this point happens and when everyone else starts to use insights from tweet sentiments for trading, then the opportunity for arbitrage will have disappeared. ■ The Journal of the CFA Society of the UK | www.cfauk.org | 29 Professional Investor | Feature Profile Fernan Flores Fernan Flores is a freelance market research analyst and director at Zapienza, a Canary Wharf-based market research consulting firm that specialises in the technology and finance sectors, which he established after completing his MBA degree from the Cambridge Judge Business School. Apart from the technology and finance sectors, he also does a considerable amount of work in the not-for-profit sector and specialises in the deployment of technology to solve healthcare issues in developing markets. He has passed the Level I exam of the CFA Program and is a member of the CFA UK marketing and communications committee. Source: Yahoo! Finance and Twitter Figure 5: Apple 2 months December 2011 - January 2012 Microsoft (Jan - Dec* 2011) Microsoft (Jan - Dec* 2011) Apple (Dec 2011 - Jan 2012) Source: Yahoo! Finance and Twitter Figure 5. Z-scores of Apple’s closing stock price in NASDAQ versus z-scores of positive sentiments using data that are further filtered manually. r = -0.85