Abstract: Financial markets are quite sensitive to unanticipated news and events. Identifying the effect of news on the market is a challenging task. In this demo, we present Forex-foreteller (FF) which mines news articles and makes forecasts about the movement of foreign currency markets. The system uses a combination of language models, topic clustering, and sentiment analysis to identify relevant news articles. These articles along with the historical stock index and currency exchange values are used in a linear regression model to make forecasts. The system has an interactive visualizer designed specifically for touch-sensitive devices which depicts forecasts along with the chronological news events and financial data used for making the forecasts.
For more information, please visit: http://people.cs.vt.edu/parang/ or contact parang at firstname at cs vt edu
Slides: Forex-Foreteller: Currency Trend Modeling using News Articles
1. Forex-foreteller: A News Based
Currency Predictor
Fang Jin, Nathan Self, Parang Saraf,
Patrick Butler, Wei Wang, Naren Ramakrishnan
Department of Computer Science
Virginia Tech
Aug 13, 2013
2. 2
EMBERS
• Funded
by
Intelligent
Advanced
Research
Projects
Ac:vity
(IARPA)
• Primarily
Interested
in
making
predic:ons
about
La:n
American
Countries
• The
primary
predic:on
areas
are
as
follows:
• Civil
Unrest
Events
• Influenza
Like
Illness
Events
• Rare
Diseases
Events
• Elec:ons
• Financial
Events
3. 3
Foreign Exchange Market
• Most
liquid
financial
market
in
the
world
• Average
daily
turnover
was
USD
3.98
trillion
in
April
2010
• Growth
of
approximately
20%
as
compared
to
2007
• United
States
GDP
is
around
USD
16.62
trillion
• Operates
24
hours
a
day
except
on
weekends
• Geographically
Dispersed
• Traders
include
large
banks,
central
banks,
ins:tu:onal
investors,
currency
speculators,
corpora:ons,
governments
and
retail
investors
• A
variety
of
factors
effect
exchange
rate:
• Economic
Factors
• Poli:cal
Condi:ons
• Market
Psychology
4. 4
Related Work
• Fundamental
Analysis
• Analyses
economic
health
of
a
country
• Employment
Reports
• Infla:on
• Produc:vity
• Trade
• Growth
• Technical
Analysis
• Mathema:cal
Techniques
like
VAR,
ARCH,
GARCH
etc
• Based
on
Past
Trends
of
financial
indicators
• Can’t
rely
on
just
one
type.
Have
to
use
a
combina:on
of
both
the
techniques
5. 5
Our Approach
Bloomberg
News
Interest
Rates
Infla:on
Unan:cipated
News
Past
Currency
Values
Past
Stock
Values
Linear
Regression
Model
Final
Predic:on
Fundamental
Technical
7. 7
Language Modeling
Different
Types
of
News
Latent
Dirichlet
Alloca:on
Model
to
iden:fy
different
topics
Top
30
topics
are
Iden:fied
Out
of
30
topics,
manually
iden:fy
topics
of
Interest
List
of
Interes:ng
topics
8. 8
Topic Clustering
Iden:fy
trending
topics
by
tracking
topic
distribu:on
movement
over
:me
10. 10
Linear Regression
Interest
Rates
Infla:on
Unan:cipated
News
Past
Currency
Values
Past
Stock
Values
Linear
Regression
Model
Final
Predic:on
Where:
• Δc
is
currency
change
• Δr
is
interest
rate
change
• Δf
is
interest
rate
change
• Δs
is
currency
change
• Δe
is
currency
change
• βr,
βf,
βs,
βe
are
respec:ve
weights