Neural Networks - page 21

New comment
 

BPNN Predictor


Files:
BPNN.zip  53 kb
 
Prediction of stock market returns is an important issue in finance. The aim of this paper is to investigate the profitability of using artificial neural networks (ANNs). In this study, the ANNs predictions are transformed into a simple trading strategy, whose profitability is evaluated against a simple buy-hold strategy. We adopt the neural network approach to analyze the Taiwan Weighted Index and the S&P 500 in the States. Consequently, we find that the trading rule based on ANNs generates higher returns than the buy-hold strategy.
 
tampa_:

BPNN Predictor


Isn't it using future data?
 
Artificial neural networks can be most adequately characterized as computational models with particular properties such as the ability to adapt or learn to generalize or to cluster or organize data and which operation is based on parallel processing However many of the above mentioned properties can be attributed to existing nonneural models the intriguing question is to which extent the neural approach proves to be better suited for certain applications than existing models To date an equivocal answer to this question is not found
 
Electronic markets have emerged as popular venues for the trading of a wide variety of financial assets, and computer based algorithmic trading has also asserted itself as a dominant force in financial markets across the world. Identifying and understanding the impact of algorithmic trading on financial markets has become a critical issue for market operators and regulators. We propose to characterize traders’ behavior in terms of the reward functions most likely to have given rise to the observed trading actions. Our approach is to model trading decisions as a Markov Decision Process (MDP), and use observations of an optimal decision policy to find the reward function. This is known as Inverse Reinforcement Learning (IRL). Our IRL-based approach to characterizing trader behavior strikes a balance between two desirable features in that it captures key empirical properties of order book dynamics and yet remains computationally tractable. Using an IRL algorithm based on linear programming, we are able to achieve more than 90% classification accuracy in distinguishing high frequency trading from other trading strategies in experiments on a simulated E-Mini S&P 500 futures market. The results of these empirical tests suggest that high frequency trading strategies can be accurately identified and profiled based on observations of individual trading actions.
 
Attraction models are very popular in marketing research for studying the effects of marketing instruments on market shares. However, so far the marketing literature only considers attraction models with certain functional forms that exclude threshold or saturation effects on attraction values. We can achieve greater exibility by using the neural net based approach introduced here. This approach assesses brands' attraction values by means of a perceptron with one hidden layer. The approach uses log-ratio transformed market shares as dependent variables. Stochastic gradient descent followed by a quasi Newton method estimates parameters. For store-level data, neural net models perform better and imply a price response that is qualitatively different from the well-known multinomial logit attraction model. Price elasticities of neural net attraction models also lead to specific managerial implications in terms of optimal prices. (author's abstract)
 

Is there any code that program the process of optimization? so that we can automate the optimization.

logic.

0) do only at weekend.

1) set parameters in these range, 0. 200 and with step 1.

2) get the optimization result

3) round of the result of profits factor, to 1.0 digits, so that 7.4=7 and 7.5 = 8.

4) then select the lest trading number in the catagori of top 2 level of profits factor range, that's optimization result that I want.

5) put the new setting into the expert EA and run for next week.

Can the optimization part be coded?

 
Recent years have witnessed the advancement of automated algorithmic trading systems as institutional solutions in the form of autobots, black box or expert advisors. However, little research has been done in this area with sufficient evidence to show the efficiency of these systems. This paper builds an automated trading system which implements an optimized genetic-algorithm neural-network (GANN) model with cybernetic concepts and evaluates the success using a modified value-at-risk (MVaR) framework. The cybernetic engine includes a circular causal feedback control feature and a developed golden-ratio estimator, which can be applied to any form of market data in the development of risk-pricing models. The paper applies the Euro and Yen forex rates as data inputs. It is shown that the technique is useful as a trading and volatility control system for institutions including central bank monetary policy as a risk-minimizing strategy. Furthermore, the results are achieved within a 30-second timeframe for an intra-week trading strategy, offering relatively low latency performance. The results show that risk exposures are reduced by four to five times with a maximum possible success rate of 96%, providing evidence for further research and development in this area.
 
The study of Artificial Neural Networks derives from first trials to translate in mathematical models the principles of biological “processing”. An Artificial Neural Network deals with generating, in the fastest times, an implicit and predictive model of the evolution of a system. In particular, it derives from experience its ability to be able to recognize some behaviours or situations and to “suggest” how to take them into account. This work illustrates an approach to the use of Artificial Neural Networks for Financial Modelling; we aim to explore the structural differences (and implications) between one- and multi- agent and population models. In one-population models, ANNs are involved as forecasting devices with wealth-maximizing agents (in which agents make decisions so as to achieve an utility maximization following non-linear models to do forecasting), while in multi-population models agents do not follow predetermined rules, but tend to create their own behavioural rules as market data are collected. In particular, it is important to analyze diversities between one-agent and one-population models; in fact, in building one-population model it is possible to illustrate the market equilibrium endogenously, which is not possible in one-agent model where all the environmental characteristics are taken as given and beyond the control of the single agent. A particular application we aim to study is the one regarding “customer profiling”, in which (based on personal and direct relationships) the “buying” behaviour of each customer can be defined, making use of behavioural inference models such as the ones offered by Artificial Neural Networks much better than traditional statistical methodologies
 

How is the AI ea working out?

1...141516171819202122232425262728
New comment