we implemented a model to predict the market price of a nonlinear chaotic time series,using reinforcement learning.
Our model is based on a multi-layer neural network, using reinforcement learning, in particular a learning algorithm called Stochastic Gradient Ascent (SGA). The proposed system includes 4 layers: input layer,hidden layer, stochastic parameter layerandoutputlayer. Inordertooptimizetheweightsandhaveapreciseprediction,we used a stochastic policy. In simulation, and to have a general model, we generate an artificial nonlinear data using Lorenz system. Then, to further test the efficiency of our model, we test it on real data. By the end, we used the simulation result, and compared short-term prediction accuracy of our proposed method with classical learning method. keywords : Forecasting, machine learning,Neural Network, Reinforcement Learning, stochastic gradient ascent,nonlinear data, time series
This work is based on the following paper: "Nonlinear Prediction by reinforcement learning". Authors: Takashi Kuremoto and Masanao Obayashi . Published: August 2005. You can use data from yahoo.finance