FINANCIAL MARKET PREDICTION AND PORTFOLIO OPTIMIZATION USING FUZZY DECISION TREES

Preview:

Citation preview

ABHRA BASAK

KRISHNA KARNANI

FINANCIAL MARKET PREDICTION AND PORTFOLIO OPTIMIZATION USING FUZZY

DECISION TREES

SECURITY SCREENING AND SELECTIONSecurity Screening and Selection

Stock Classification

Stock Ranking

Stock Selection

STOCK CLASSIFICATION

• Security Evaluation using Technical Indicators

• Moving Average Convergence Divergence (MACD)

• Relative Strength Indicator (RSI)

• Commodity Channel Index (CCI)

• Bollinger Bands

• Momentum Oscillators

STOCK RANKING

• Corporate Evaluation using Fundamental Indicators

• Profitability – Returns on Assets and Equity

• Management Performance – Assets and Inventories Turnover

• Capital Structure – Assets to Liabilities, Liabilities to Equity

• Sales, Profit, Transaction Volume, Marginal Account

STOCK SELECTION

• Select 3 different stocks – one each showing uptrend, downtrend, and steady state

• Attempt to display different profit making strategies in stock trading

• All subsequent processes are applied on these 3 stocks

TRAINING PHASETraining Phase

TRAINING PHASE

• Gather Historical Stock data

• Obtain financial time series and price charts from data

• Determine technical indicators and momentum oscillators from charts

Historical Data

Financial Time Series

Price Charts

PIECEWISE LINEAR REPRESENTATION METHOD

Training Phase

PIECEWISE LINEAR REPRESENTATION METHOD

• Mining of trading points

• Points of begin (P) and end (Q) on a term of closing prices in the ascending order of dates

• Point K having longest straight line distance between P and Q

• K is the turning point resulting in 2 segments.

• Apply recursively in the resulting segments till minimum distance threshold

PIECEWISE LINEAR REPRESENTATION METHOD

• Trading signals transformation

• Convert PLR segments into trading signals

• Uptrend segment

• I <= L/2 : 0.5 – (I – 1) / L

• I <= L/2 : I / L – 0.5

• Downtrend segment

• I <= L/2 : 0.5 + (I – 1) / L

• I <= L/2 : 1.5 – I / L

• Ranges from 0 to 1

• Can also act as a potential technical indicator

STEPWISE REGRESSION ANALYSIS METHOD

Training Phase

STEPWISE REGRESSION ANALYSIS METHOD

• Data Preprocessing for Feature Selection

• Used to select important factors which affect forecasting results

• Sort out affecting variables to leave more influential ones in the model

• Adding or removing factors to find the fittest combination, decided by F-test statistical value (takes into account the PLR)

FUZZY RULES AND DECISION TREESTraining Phase

FUZZY RULES AND DECISION TREES

• Fuzzification

• Set of indicators selected by SRA fed into data fuzzification module

• This module transforms technical indicators to fuzzy values

• Adopt triangular and trapezoidal membership functions for the module

• Output decision is obtained as a Gaussian membership function

Fuzzy Inference

I3

I2I1

FUZZY RULES AND DECISION TREES

• Defuzzification

• Output from fuzzy inference scheme is transformed into a meaningful decision

• Implemented using the popular Center of Area (COA) methods in the Fuzzy Control Module’s algorithm

FUZZY RULES AND DECISION TREES

• Examples of Fuzzy decision rules

• If MACD above signal line, then BUY

• If RSI increases above 70, then market is BULLISH

• If Price increases above BBupper then market is BULLISH

• If MACD is LOW and RSIupper goes HIGH to LOW, then SELL

• If MACD is HIGH and CCIupper goes LOW to HIGH, then BUY

GENETIC ALGORITHMS AND REFINEMENT

Training Phase

GENETIC ALGORITHMS AND REFINEMENT

• Evolving the decision tree using GA

• Fitness function set as forecasting accuracy of the model

Selection

Crossover

Mutation

Replace

Termination

RESULT

• Decision of Stock price and transaction will be determined by the decision tree on the basis of trends and indicators

• Uptrend if hike in price is greater than 0.5%

• Downtrend if fall in price is less than -0.5%

• Steady state / hold if y is between -0.5% and 0.5%

CREDITS

• A Collaborative Trading Model by Support Vector Regression and TS Fuzzy Rule for Daily Stock Turning Points Detection – Wu, Chang, Chang, Zhang

• Evolving and Clustering Fuzzy Decision Trees for Financial Time Series Data Forecasting – Lai, Fan, Huang, Chang

• A Fuzzy Logic Based Trading System – Chueng, Keymak

• Nigerian Stock Market Investment using a Fuzzy Strategy – Neenwi, Kabari, Asagba

• Common Stock Portfolio Selection: A multiple criteria Decision making Methodology and an application to the Athens Stock Exchange – Xidonas, Askounis, Psarras

Recommended