Abstract

Abstract: The stock markets all over the world have many uncertaintieswhich contain a large amount of data with jumps, noises, and movements inleading to the highly non-stationary time series.  Here, introduce real-time financial signalrepresentation and trading techniques as RDNN (Recurrent Deep Neural Network)for the environment sensing and recurrent decision making for the onlinefinancial assert trading. In which the RDNN have two parts, first one is DNN(Deep neural network) function of this DNN is feature learning and the secondone is RNN (Recurrent Neural Networks) to predict rapidly changing marketcondition. This proposed model based on two biological related learningconcepts such as Reinforcement Learning (RL) and Deep Leaning and DL (DeepLearning) is used to represent financial signals and self-taught reinforcementtrading. Here, RL is used to interact with deep representations and makestrading decisions to accumulate the ultimate rewards in an unknown environment.It improves the robustness of market summarization using fuzzy learningconcepts it reduces the uncertainty of input data.  Fuzzy MLP will be used to provide theprediction of the stock market.

Keywords: Deep NeuralNetwork, Direct reinforcement Learning, Fuzzy logic, Multilayer PerceptronModel.