Transcript Presentation(PowerPoint)

Stock Movement
Prediction
Deepathi Lingala
Sathindra K. Kamepalli
Sudhir K. V. Potturi
Agenda
Introduction-Goal
Domain Description
Method
Implementation
Results
Experiences and Challenges
Questions
Stock Movement Prediction
Goal
Apply trend analysis to stock data of
in order to predict the direction of movement
of stock value with time.
Stock Movement Prediction
Domain
Continuous Valued
Time series: Data for a period of ten years
(1992-2002)
Data size: 2587 rows
Data Attributes: Open, Close, High, Low
stock values and Volume
Stock Movement Prediction
Data Mining Technique Used
Association Rule Mining Technique has been
used for the Prediction
Why Association Rule Mining Technique?
Association rule mining helps in finding
interesting association relationships among
large set of data items. The discovery of such
associations can help develop strategies to
predict.
Stock Movement Prediction
Implementation
Data Preparation
Data Cleaning
Data Transformation
Data Discretization
Data Partition
Association Rule Mining
Stock Movement Prediction
Data Preparation
Data Cleaning
Not much data cleaning was required.
Missing data was replaced by the correct one
obtained from the internet. The data was
searched for any steep changes in it which
might have occurred by stock splits etc., but
did not find any
Stock Movement Prediction
Data Preparation
Attributes used:
Closing stock price (Decision attribute)
Volume
Derived Attributes
Two-day average
Five-day average
Ten-day average
Average True Range (ATR)
Absolute Price Oscillator (APO)
Stock Movement Prediction
Data Preparation
Data Transformation
The data has been transformed into percentage
rate of change, wherein the percentages are
obtained according to the increase or decrease
with respect to the previous day.
The decision attribute was generalized to 0’s and
1’s according the increase or decrease of the
close stock price compared to its previous day
price.
Stock Movement Prediction
Data Preparation
Data Discretization
Software Used: ROSETTA
Algorithm Used: Equal Frequency Binning
The data is discretized and put into bins.
Each bin was given a separate name for the
purpose of increasing the ease of
understanding when the rules are developed.
Stock Movement Prediction
Data Partitioning
The data tuples are analyzed, the training
data set(1000 records), is selected from the
data set. This learned model is represented
in the form of association rules. This step is
the supervised learning step. A test data set
(150 records) is selected and this is
independent of the training data set.
Stock Movement Prediction
Association Rule Mining
Software used: LERS
The Training data set has been fed into the
LERS system to build the association rules
(Machine Learning)
Total No. of Rules: 1059
Certain Rules: 532
Possible Rules: 527
Stock Movement Prediction
Association Rule Mining
Support for all the Certain and Possible rules
was determined.
A threshold support value was chosen.
The rules were filtered based on the
threshold support value.
Stock Movement Prediction
Association Rule Mining
After filtering
Total number of rules: 55
Certain Rules: 27
Possible Rules: 28
These rules were applied to the test data to
predict the decision value
Stock Movement Prediction
Example Rules
Certain Rules:
(vol,a9) & (5day,c2) & (2day,b3) -> (close,1)
(apo,f6) & (5day,c0) -> (close,0)
Possible Rules:
(vol,a9) & (atr,e3) & (2day,b4) -> (close,1)
(5day,c6) & (apo,f7) & (10day,d7) -> (close,0)
Stock Movement Prediction
Results
No. of Records in the Test Data Set = 150
Total No. of correct matches Found = 77
Accuracy = 51.33%
No. of correct Full matches = 20 out of 36
Accuracy = 55.55%
No. of correct Partial matches = 57 out of 114
Accuracy = 50%
Stock Movement Prediction
Results
No of Records
200
150
100
50
77
57
73
57
20
16
1
2
3
0
Total
Partial
Stock Movement Prediction
Full
Results
Full Match
No Match
Partial
Match
Stock Movement Prediction
Experiences & Challenges





Manual for LERS
Huge Data sets
Support & Confidence Measures
Rule Filtering Tools
Time Constraint
Stock Movement Prediction
QUESTIONS ??
Stock Movement Prediction
THANK YOU !
Have a Happy Thanks Giving!
Stock Movement Prediction