Tiffany Iwantoro_Exchange Rate Directional Forecasting using Sentiment Analysis on Social Media in...

Post on 15-Jan-2017

21 views 0 download

Transcript of Tiffany Iwantoro_Exchange Rate Directional Forecasting using Sentiment Analysis on Social Media in...

Final PresentationTuesday, August 24th, 2015

Presented byTiffany Iwantoro - 19012057

SupervisorDeddy Priatmodjo Koesrindartoto, PhD

Tutorial 1E SBM ITB10 AM – 11 AM

ISBN 978-0-9942714-2-6

Social Media Trends in Indonesia

Source: eMarketers, Forbes

- User-friendly- Accessible- Low-cost- Show real time event

Social Media Exposure

Exchange rateSupply

Demand

• Self-fulfilling prophecy• Behavioral finance

Fluctuation of Exchange Rate

Source: Yahoo! Finance

USD/IDR

Aug 1st, 201411.760

May 31st, 201513.219

12,40%

Fluctuation of Exchange Rate

Source: Yahoo! Finance

EUR/IDR

Aug 1st, 201415.794

May 31st, 201514.494

8,96%

Fluctuation of Exchange Rate

Source: Yahoo! Finance

JPY/IDR

Aug 1st, 2014114,6051

May 31st, 2015106,4546

7,66%

Government Loans

Corporate Loans

Government Balance of Payment

Export and Import Forex Trading Investor or

Individual

Uses of Foreign Currencies

1.2.1 Is there any relations between market sentiments and exchange rate?

1.2.2 How significant is the relations between market sentiments and exchange rate?

Problem Identification

Research Objectives

1.2.1 To find out the relations between market sentiments and exchange rate

1.2.2 To find out the significance between market sentiments and exchange rate

Previous Findings

Research Objectives

Lucia Russo (2013)

• “Can Social Micro-blogging be Used to Forecast Intraday Exchange Rate?”

• Time series analysis

• Can predict in short run

Fang Jin (2012)

• “Forex Foreteller: Currency Trend Modelling using News Articles”

• Other factor : inflation, interest rate

K.S. Madhava Rao (2013)

• “Exchange Rate Market Sentiment Analysis of Major Global Currencies”

• EUR, GBP, SRD, YEN, ZAR/USD

• Time series and probabilistic mode

Alexander Kurov (2008)

• “Investor Sentiment, Trading Behavior, and Informational Efficiency in Index Future Markets”

• Using intraday frequency trading

Vasilios P. (2007)

• “Market Sentiment and Exchange Rate Directional Forecasting”

• Efficient Market Hypothesis in Weak Form

• EUR, GBP, AUD/USD

• Indonesian Rupiah (IDR)• Major global currencies (USD, EUR, JPY)• Twitter in Indonesia

• Sentiment in Indonesian language• Linear regression• Semantria

Research Contributions

Research Design

Problem Identification

Research Objectives

Literature Study Data Collection

Sentiment Dictionary

Classical Assumption Test

Linear Regression

Conclusion and Recommendation

Journals

Textbook

Sentiment Analysis

Exchange Rate

Tweets

Exchange Rate

• USD/IDR• EUR/IDR• JPY/IDR

Jan 1st, 2015 – May 31st, 2015

Tweets

Data CollectionFilteringKeywords

• USD IDR• Dollar rupiah

• EUR IDR• Euro rupiah

• JPY IDR• Yen rupiah

EUR/IDR JPY/IDRUSD/IDR10,247tweets

2,314tweets

1,545tweets

Problem Identification

Research Objectives

Literature Study Data Collection

Sentiment Dictionary

Classical Assumption Test

Linear Regression

Conclusion and Recommendation

Journals

Textbook

Sentiment Analysis

Exchange Rate

Tweets

Research Design

Sentiment Dictionary

Score Description

+1- Indicates positive sentiment- High buying or selling interest towards current exchange rate- Potentially influence others to buy or sell currencies

0- Neutral sentiment towards exchange rate- Author aims only to spread news/information

-1- Indicates negative sentiment- Low buying or selling interest towards current exchange rate- Potentially influence others not to buy or sell currencies

Indonesian Word English Translation Sentiment Score

Menguat Strengthened +1

Naik Up, rise, increase, escalate +1

Perkasa Strong +1

Penguatan Strengthening, reinforcement +1

Kuat Strong +1

Positif Positive +1

Melemah Weakened, fall off -1

Turun Down, sink, subside -1

Anjlok Drop -1

Pelemahan Weakening, impairment -1

Tertekan Oppressed -1

Ambruk Collapse, crumble, tumbling -1

Terpuruk Worsen -1

Stabil Stable 0

Bertahan Survive, last, withstand 0

Tetap Still, consistently, constant 0

Problem Identification

Research Objectives

Literature Study Data Collection

Sentiment Dictionary

Classical Assumption Test

Linear Regression

Conclusion and Recommendation

Journals

Textbook

Sentiment Analysis

Exchange Rate

Tweets

Research Design

Sentiment Analysis

0.51 until 1 Positive-0.5 until 0.5 Neutral-0.51 until -1 Negative

𝑹=𝑷𝒕−𝑷𝒕−𝟏

𝑷𝒕−𝟏 Average sentimentper day

Problem Identification

Research Objectives

Literature Study Data Collection

Sentiment Dictionary

Classical Assumption Test

Linear Regression

Conclusion and Recommendation

Journals

Textbook

Sentiment Analysis

Exchange Rate

Tweets

Research Design

Classical Assumption Test

USD/IDR Classical Assumption Test

EUR/IDR Classical Assumption Test

JPY/IDR Classical Assumption Test

USD/IDR EUR/IDR JPY/IDR

Sample 129 151 151

Normality Test(Kolmogorov-Smirnov Test)

Asymp. Sig (2-tailed)

0.509 0.121 0.167

Heteroscedasticity Test Sig. Score 0.271 0.425 0.831

Auto Correlation Test(Durbin Watson)

DW Score 2.206 1.986 1.954

Result Pass Pass Pass

Classical Assumption Test Summary

Problem Identification

Research Objectives

Literature Study Data Collection

Sentiment Dictionary

Classical Assumption Test

Linear Regression

Conclusion and Recommendation

Journals

Textbook

Sentiment Analysis

Exchange Rate

Tweets

Research Design

Linear Regression

USD/IDR Linear Regression

EUR/IDR Linear Regression

JPY/IDR Linear Regression

Linear Regression Summary

USD/IDR EUR/IDR JPY/IDR

Sample 129 151 151

R 0.433 0.361 0.318

R Square 0.188 0.131 0.101

F-stat 29.314 22.368 16.740

Regression model Y = 0.005 X Y = 0.011X Y = 0.006X

T-stat 5.414 4.730 4.092

Sig. 0.000 0.000 0.000

Result H0 rejected H0 rejected H0 rejected

H0 = there are no significant relations between market sentiment and exchange rateH1 = there are significant relations between market sentiment and exchange rate

Conclusion

1. The are significant relations between market sentiment and exchange rate in Indonesia.

2. Indonesian’s Twitter can show significant relations between market sentiment and exchange rate as much as

USD/IDR EUR/IDR JPY/IDR18,8% 13,1% 10,1%

Recommendation

• Various types of social media such as Facebook, blog, websites, and discussion groups

• Online news• Time-series analysis• Intra-day trading data for the tweet and the currency rate• Broaden the sentiment words into local language and social media

language instead of formal language in Indonesia.