A Method of Predicting Moments of Market Trend Reversal for Decision-Making Block of Computer-Based Intelligent Financial Agent-Avatar

A Method of Predicting Moments of Market Trend Reversal for Decision-Making Block of Computer-Based Intelligent Financial Agent-Avatar

Vsevolod Chernyshenko (Financial University Under the Government of the Russian Federation, Russia), Vladimir Soloviev (Financial University Under the Government of the Russian Federation, Russia), Vadim Feklin (Financial University Under the Government of the Russian Federation, Russia), Mikhail Koroteev (Financial University Under the Government of the Russian Federation, Russia) and Nikita Titov (Financial University Under the Government of the Russian Federation, Russia)
DOI: 10.4018/978-1-7998-1581-5.ch001

Abstract

The chapter formalizes the financial task for the definitions and properties of financial indicators under study. A wide range of traditional approaches used for predicting economic time series were reviewed. Investigated as well were the advanced algorithms for predicting moments of reversals of market trends based on machine learning tools. The chapter discusses the effectiveness of different kinds of approaches, which is illustrated with related examples. Described is an original securities price dynamics trend classification algorithm, based on the use of the sliding window methodology and financial agents. General scheme of the classification algorithm to identify market phases is analyzed and results of computer modeling are presented. Selection of initial and resulting metrics is grounded.
Chapter Preview
Top

Introduction

Design of intelligent computer systems capable to lead to effective economic (particularly financial) decisions is a topical problem that is still far from being solved. It encounters a number of difficulties, one of which is a high randomness of the dynamics of economic indicators, immanently inherent in this class of dynamic processes. On the other hand, they nature may not be considered as a completely chaotic: there are, of course, certain patterns that can be recognized by high-intelligent algorithms. It should be mentioned that even a slight increase of the accuracy of a volatility forecast may provide an investor with a quite significant yield.

Obviously, algorithms that make reliable forecasts regarding market trends dynamic cannot be based just on simple mathematical models with fixed properties. Recent trends in this filed – solutions based on machine learning that collect and analyze big statistical data in real time (including data for an evaluation of the quality of these models previous prognoses and the effectiveness of the corresponding recommended solutions). Such solutions may be represented as computer agents or avatars – pieces of a program code, to be separate objects with their inputs and outputs, interacting with/in a common software environment and having access to the relevant databases. Such computer “financial agents” are widely used in a financial sphere and partly determine financial markets dynamics themselves. That fact, for sure, does not contradict the effectiveness of use of such algorithms, since the described approach doesn`t link to a particular market model but is able to adapt to any conditions.

Obviously, one of the key quality of an agent (determining its success mostly) is its ability to predict the market trend in relation to a certain indicator (e.g. share quotation of certain companies). In this chapter, one of such an approaches to create an “avatar self-learning” algorithm will be considered, allowing user to effectively predict changes of financial markets` trends.

Complete Chapter List

Search this Book:
Reset