Market Forecasting using(2D)2PCA + RBFNNBY: DANNY SANCHEZ
Current Models (market forecasting)
Auto-regressive Moving Average (ARIMA) Univariate
Benchmark used for comparisons
Predictor based regression model Multivariate
Various types of neural networks used
“A Stock Market Forecasting Model Combining Two-Directional Two-Dimensional Principal Component Analysis and Radial Basis Function
Neural Network” Paper published: April 7, 2015 Authors: Zhiq Guo, Huaiqing Wang, Jie Yang, David J. Miller
Subject: Chinese DOW
Multivariate model Proposed:
(2D)2PCA + RBFNN
“The proposed (2D)^2 PCA+RBFNN model use (2D)^2 PCA to remove the noise from the input raw data, the feature will contain less noise information and serve as the input of the RBFNN to predict the value or movement of the next day’s closing price.”
Hardware:
2.30 GHz CPU + 2Gb RAM + MATLAB software
General Outline
Data Source (Open, High, Low, Adj. Close)
-> Create Predictor variables
-> Reduce noise/extract features ((2D)2PCA )
-> Use reduced matrix as input for RBFNN
Predictor Variables
Using market data (usually in .csv format) containing historical:
Open – High – Low – Adj. Close
Predictors created using common technical indicator formulas. For the this research study 36 predictors were used including: MACD – Stochastic – Relative
Strength Indicator
Three of the most commonly used momentum indicators by technical traders
(2D)2PCA
Based on the idea of Principle Component Analysis Been used in signals processing for noise reduction for years
2D aims to reduce dimensionality while retaining usefulness of data by reducing variation
Ex.
RBFNN
Three-layered structure Input Layer -> Hidden Layer -> Output Layer
Results
“Replicating (2D)2PCA + RBFNN market forecasting model in R”
Authors: Danny Sanchez
Subject: S&P 500 (^GSPC) otherwise known as SPX
Multivariate model used: (2D)2PCA + RBFNN Similar to the original research but I have chosen my own predictor
variables based on market knowledge
Hardware: 2.30 GHz CPU + 2Gb RAM + MATLAB software
Predictor Variables
Using Yahoo Finance Historical data for S&P 500 & VIX (Volatility Index) containing:
Open – High – Low – Adj. Close
Based on previous trading knowledge, I chose 20 predictor vectors including: MACD – Stochastic – Relative
Strength Indicator
Three of the most commonly used momentum indicators by technical traders
Results
Results - Personal