SOPS: Stock Prediction using Web Sentiment Presented by Vivek sehgal, Charles Song Department of Computer Science, University of Maryland ICDMW 2007 2009-05-29.

SOPS: Stock Prediction using Web SentimentSOPS: Stock Prediction using Web Sentiment

Presented by Vivek sehgal, Charles Song

Department of Computer Science, University of Maryland

ICDMW 2007

2009-05-29

Summarized by Jaeseok Myung

Copyright 2009 by CEBT

In this talk..In this talk..

Introducing some papers about sentiment analysis in finance

[1] 0Event and Sentiment Detection in Financial Markets (ISWC 08)

– Simple Architecture

[2] SOPS: Stock Prediction using Web Sentiment (ICDMW 07)

– Entire Process

[3] Yahoo! for Amazon: Sentiment Extraction from Small Talk on the Web (Management Science 07)

– An Idea that can improve prediction performance

We will focus on SOPS, but brief introductions about the others will also be presented

Center for E-Business Technology


Sentiment Analysis in Financial Sentiment Analysis in Financial MarketsMarkets

Sentiment analysis is one of my favorite research topic

I’ve conducted some researches by using product reviews

In my opinion, finance is more suitable domain than product

Product sales statistics is not publicly available

– Stock values are always opened

Financial markets are really related to investors’ sentiment

– ‘ 경제는 심리’

– Behavioral finance

– Lots of evidences

Interesting & Worth



Research Problem from [1][2][3]Research Problem from [1][2][3]

How can information from various, heterogeneous sources be integrated?

Different formats

How can the opinions in the documents be extracted?

Statistical, NLP ways

How can the important opinions be filtered?

Reliable Source(news, blog), Trusted Author, Promising Alg.

How can the users’ trading decisions be supported?

Finding out the relationships between investors’ sentiment and stock values



An Architecture from [1]An Architecture from [1]


Monitor a huge number of relevant sources

Monitor a huge number of relevant sources

Extract metadata and Make a single representation

Extract metadata and Make a single representation

Decide whether the information has to be analyzed or not

Decide whether the information has to be analyzed or not


SOPS: System OverviewSOPS: System Overview


Collect data from a message board

Collect data from a message board

Remove HTML tags and extract features

Remove HTML tags and extract features

Identify reliable users in order to filter noise

Identify reliable users in order to filter noise

Use several classifiersUse several classifiers


SOPS: Data CollectionSOPS: Data Collection

260,000 messages for 52 popular stocks on Yahoo! Finance

The messages covered over 6 month time period

A message board exists for each stock traded on major stock exchange such as NYSE and NASDAQ

Users must sign up before they can post messages

Every message posted is associated with the author



SOPS: Data CollectionSOPS: Data Collection



SOPS: Feature RepresentationSOPS: Feature Representation

After the relevant information has been extracted

Converting each message to a vector of words and author names

The value of each entry in the vector is then calculated using TFIDF formula


M : set of all messagesm : a messagew : a term

M : set of all messagesm : a messagew : a term

( 3.2, 1.6, 1.09, 3.37. 90, 0.5, …)

“good” “stop” “asdf” date % of change in stock price


SOPS: Sentiment PredictionSOPS: Sentiment Prediction


a message(undisclosed)

Classifier

Strong Buy Strong Sell

Buy Sell

Hold

What How

a message(disclosed)

Classifier(Training)


Buy Sell

Hold


SOPS: Sentiment PredictionSOPS: Sentiment Prediction

The sentiment for a message m at time instant i is modeled as follows:


m : a messageMi : set of all messagesSVi : Stock value

m : a messageMi : set of all messagesSVi : Stock value

Classifier

1.Naïve Bayes2.Decision Trees3.Bagging

Strong Buy, Buy, Hold, Sell, Strong SellStrong Buy, Buy, Hold, Sell, Strong Sell


Buy Sell

Hold

0.2 0.3 0.1 0.4


TrustValue CalculationTrustValue Calculation

Some authors are more knowledgeable than others about the stock market

Trusted author’s posts should carry more weight => TrustValue

TrustValue

Not only cares about the direction in which the stock price went, but also care about the magnitude

Takes into account the fact that a single author cannot be expert on all stocks => an author can be assigned different trust values for different stocks


PredictionScore : author’s prediction performance that is how closely does the author’s prediction follow the stock market

NumberOfPrediction : the total number of predictions made by the author

ExactPrediction : the number of exact predictions

ClosePrediction : the number of “good enough” predictions

ActivityConstant : a constant used to penalize low activity or predictions by the author

PredictionScore : author’s prediction performance that is how closely does the author’s prediction follow the stock market

NumberOfPrediction : the total number of predictions made by the author

ExactPrediction : the number of exact predictions

ClosePrediction : the number of “good enough” predictions

ActivityConstant : a constant used to penalize low activity or predictions by the author


SOPS: Stock PredictionSOPS: Stock Prediction


Classifier

Go up Go down


SOPS: Evaluation MetricsSOPS: Evaluation Metrics



SOPS: ExperimentsSOPS: Experiments



ConclusionConclusion

SOPS can predict Web sentiment with high precision and recall

SOPS introduced TrustValue which takes into account the trust-worthiness of an author

In my opinion, there are some points that are unclear

Presentation

– About Summarization

Users

Time Period



FurthermoreFurthermore

We have the paper [3]



Research Problem from [1][2][3]Research Problem from [1][2][3]

How can information from various, heterogeneous sources be integrated?

Different formats

How can the opinions in the documents be extracted?

Statistical, NLP ways

How can the important opinions be filtered?

Reliable Source(news, blog), Trusted Author, Promising Alg.

How can the users’ trading decisions be supported?

Finding out the relationships between investors’ sentiment and stock values


SOPS: Stock Prediction using Web Sentiment Presented by Vivek sehgal, Charles Song Department of Computer Science, University of Maryland ICDMW 2007 2009-05-29.

Documents

ebusiness technology

cebt sops

sentiment prediction

cebt sentiment analysis

stock values center

author center

stock price slide

sentiment detection