Multi-input single-output (MISO) random system modeling using methods of system identification

The paper utilizes techniques commonly used in the system identification dynamic systems behavior using output-input data to an unknown dynamic system. The identification techniques are based on nine inputs and one output. The system is applied to a financial time series that represent the historical prices of gold. The nine inputs are the technical indicators calculates form the historical data of open, high, low, close, and volume of trading the gold while the output is the forecasted value of the closing price of gold. Nonlinear Identification techniques used in this paper include wavelet Network, Sigmoid Network and Tree Partition. The purpose of the identification techniques is come up with a dynamic system model “either a transfer function or State-Space model” that is capable of predicting the values of the output “close”. The data is split into estimation set and verification set. The estimation group is used in determining the best possible model that can predict the verification set of data. The highest match obtained was 92%. Details on the modeling techniques as well as the effect of each input on the output are also presented in this paper. Simulation results are utilized to examine the accuracy and integrity of the model proposed.