Forecasting the stock market using neural networks
@Cryptics
In the modern world, interest in the qualitative forecasting of financial markets is becoming more acute. This is due to the rapid development of high technology and, accordingly, the emergence of new tools for data analysis. However, the technical analysis that most market participants are accustomed to using is not effective. Forecasts based on exponential moving averages, oscillators and other indicators do not give a tangible result, because The economy is often irrational, because it is driven by irrational motivations of people.
In recent years, financial analysts have become very interested in the so-called artificial neural networks - these are mathematical models, as well as their software or hardware implementations built on the principle of the organization and functioning of biological neural networks - nerve cell networks of a living organism. This concept arose in the study of processes occurring in the brain in thinking, and when trying to simulate these processes. Subsequently, these models began to be used for practical purposes, as a rule, in forecasting problems. Neural networks are not programmed in the usual sense of the word, they are trained. The possibility of learning is one of the main advantages of neural networks over traditional algorithms. Technically, training is to find the coefficients of connections between neurons. In the process of learning, the neural network is able to identify complex dependencies between input data and output, and perform generalization. The neural network's ability to predict directly follows from its ability to generalize and isolate the hidden dependencies between input and output data. After training, the network is able to predict the future value of a certain sequence based on several previous values and / or some existing factors at the moment. It should be noted that forecasting is possible only when the previous changes do in some way predetermine the future. For example, the forecasting of stock quotes based on quotations for the past week may be successful, while forecasting the results of tomorrow's lottery based on data over the past 50 years will almost certainly not yield any results.
Neural networks are producing good results. This is largely due to the complexity and non-linearity of the structure of this series, whereas the classical methods are designed to apply to series with more visible and obvious structural patterns. But even in spite of all the visible positive qualities of neural networks, it is not worth considering them as a kind of "panacea". First, neural networks are a "black box", which does not allow to explicitly determine the type of dependencies between members of the series. Thus, a specific neural network can be "taught" to build a forecast only for a strictly fixed number of steps forward (which we specify in the specification of this network), therefore, there is a strong dependence on the type of problem. Secondly, in the presence of explicit linearity, the simplicity of the structure in the problem, the ability of neural networks to generalize is weaker with respect to classical methods. This is explained by the non-linear nature of the networks in their essence.
In general, to achieve the best result, it is necessary to use neural networks, together with a competent strategy for managing capital.
Source: https://m.geektimes.ru/post/279170/