Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

Can social media

No description
by

Random ness

on 25 November 2013

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Can social media

Agenda
Model used in
published
papers
The model investigated against which
baseline
the number of
tweets about a certain topic
:
Emotional tweet predict stock market indicator
Image by Tom Mooring
Model Summary
Data
Sentiment Integration
Can social media
Predict Stock Price ?


worry
e.g.
The model is measured with
3 different baselines:


The number of
tweets
per day
The number of
followers
per day
The number of
retweets
per day
Pros and Cons
Simple
and
easy
to calculate and predict
Easy to craw tweets with emotional tags
“#hope, #fear, and #worry

The model only provides a
rough
prediction instead of a precise prediction on the stock market indicator.
Positive & Negative
Tweet Corpus

Keywords related
to company
Data Selection

Normalization
Han et al method
to normalize out of vocabulary issue

iphone, ipad, mac, ios
(OOV) words:
"Ipad 2 is very coooolllllll"
Remove noise data
Remove ad with linear Conditional Random Fields(CRF) model


Determine polarity for a certain
Tweet using a keyword- based algorithm

Twitter Semtiment tool
Bullish & Bearish
Normalized
Tweet
seperate words into
bullish and bearish
CMU Part-of-speech tagger
achieved the state of the art on tweet data
Semantic Orientation algorithm
calculate anchor words, measure the association between two words
SO(w):
PMI(w,"bullish") - PMI(w, "bearish")
A large p-value (> 0. 05)
indicates weak evidence
against the null hypothesis

P- value

Significance!
Pros:
High accuracy: 70 - 80%
Cons:
Predict
past performance only
;
Steps are
complicated
and

inefficient
Focus on
product
only
Products have
same importance weighting
Six kinds of Mood :

Calm, Alert, Sure,
Vital, Kind, Happy
Value of Dow Jones
Industrial Average (DJIA)
UP
DOWN
Bollen’s work
Volodymyr Kuleshov's work
Volodymyr Kuleshov. “Can twitter predict the stock market ?” 16.11.2011
Using only three previous DJIA values;
1
2
Only test set accuracies for 8 models reported;
Baseline algorithm already has a 73% accuracy.
3
Bollen’s work is
by luck
Twitter mood
predicts the stock market ?

Data from
iPhone Information Tweets
When
difference
between the
retweets
and

favorites

enlarges,
the price of stock is more likely to go down.
More
retweets
does not mean
increase
of stock price.
Data of daily tweets is simple
No reply mood analysis
Method used is doubtable
Cons:
Our proposed model
Improvement we can make
1st model:
Emotional tags like
eg. #hope, #fear and #worry
is not frequently used in normal tweets.
But every tweets/statement would have a sentimental polarity
(positive, negative, neutral)
larger sample size!
2nd model:
Our method is mainly based
on the 2nd model
+
weighting score
e.g.
#mac
has more tweets than
#iphone5
sentimental score of
#mac
should weight heavier
3rd model
emotion
"concern"
Data sample is
small
, result not significant
sample ~2000 tweets
per #tag every day, get the
total number
of tweets by Topsy
Data normalization and cleaning
Sentimental score(w) =
sum of polarity(w)
total number of tweets (w)
Weighted score (d) =
carry a
correlation test
between the sentimental score and the delta of open and close stock price of APPL
For both positive and negative correlation, strong correlation (0.6 to 1 or -0.6 to -1) is observed for +1day prediction.
Predicting rise and fall of stock
Rise or fall factor (f) =
∆ positive score of (t-1) +
∆ negative score (t-1)
if f=
+ve,
that means the stock would
rise
f=
-ve
, that means the stock would
fall
f=0, no changes
80%
accuracy with the sample data
Conclusion
Could only predict whether the stock price would
rise / fall tomorrow
.
predict next week/ month
Tweets is the
reflection / digest of news
, and
news affect the stock price
.
Like Lehman brothers, no one could ever predict an
AAA rank
company would suddenly go
bankrupt
.
Our proposed
model
1
2
Presented by:
HEUNG, Wai Kin Matthew,
KING, Sing,
LAI, Pik Yee Charlotte,
ZHAO, Qian

fear
Accuracy cannot be assured !
Move = OpenP - CloseP

Crawl product related data from twitter

-
Market Change
pre(i)
i
i
Difference < 0
Drop
Difference >= 0 Raise
retweets
favourite
reply
Official Tweets
six kinds of moods
The twitter
mood
towards certain products might be an useful tool to predict the trend of stock.
"Concern"
tweet account/ campaign
Find
related
tags using hotness!

Collect
data
from twitter API
use sentimental tool
to get the
polarity score
predict exact stock price
Thank you
e.g. LOL, ad
Full transcript