Predict digital asset trends through automated technical analysis and artificial intelligence(1)

Introduction

BitPeaks is an automated trading signal generation engine, supports in dealing with Digital Asset Exchanges in particular. Because of their particular nature in terms of the difficulty of making gains. And the continuing losses are illogical and unjustified to most investors. It was necessary to use modern software techniques to be able to determine the method of dealing and trading. To avoid losses as much as possible. By eliminating the human emotion which is usually the main reason for achieving "consecutive" losses that erode the capital, relying on the machine to analyze the data and nomination of assets without any human intervention, and use this system as a decision support tool or as a complete independent trader if connected to trading API of target digital asset exchanges (upcoming releases).

Abstract

Historical digital asset prices are used to predict future price trends. The prediction model uses two layers to analyze the data. Technical analysis in the first layer and then a second layer of thinking based on machine learning. This model complements the fund management strategy that uses the recommendations made by the model to determine the course of capital invested. It builds a portfolio of entry and exit signals resulting from this model, and concludes how far the forecasting model is, relative to performance of whole market.

Approach

Predicting the direction of future prices is a topic that has been widely studied in many fields including trade, finance, statistics and computer science. The fundamental motivation of course is to make gains. Professional traders usually use basic analysis and technical analysis to analyze markets and make investment decisions. Basic analysis is the traditional approach of studying the fundamentals of companies such as revenue and expenditure, market positioning, annual growth rates, coin technical potential, and so forth. Technical analysis, on the other hand, only examines historical price fluctuations. Technical analysis practitioners study historical prices to extract price action patterns using data at different time intervals to predict future price movements. Thus there is an inherent correlation between price and asset, which can be used to determine the times of entry and exit for each asset.

In the field of finance, statistics, computer science, and most traditional models, statistical models and neural network models are used(2)Derived from the price data of the forecast. Moreover, the dominant strategy in computer science uses evolutionary algorithms(3)، Neural networks, or a combination of both (advanced neural networks). The approach adopted in this system differs from the traditional approach in that it uses the first layer of technical analysis before applying a second layer of reasoning based on machine learning for further analysis and accuracy.

Overview of the steps of the forecasting model

The developed model uses a two-tier inference approach. The first layer of logic is to extract models and technical indicators from historical prices. The values ​​of the resulting technical indicators are then routed to the second layer of logic called the model classification unit, where different values ​​of technical indicators are compiled and their future results classified according to the most profitable models and used to teach the machine to enable it to match those indices and extract them if they are formed again on the same assets, This is called "artificial intelligence" and self-learning of the machine, the more information becomes available over time, the more it learns, the better it performs "experience", the result is the investment strategy that can be used to identify markets for trading, Investment in terms of profitability by simulating virtual wallet using trade signals generated by the investment strategy. In addition, the Unit includes a fund management strategy used to assess the strength of forecasts and to determine the amount of capital to invest in each signal generated.

Structure

Historical data

The function of this unit is to extract historical data for all markets listed.

Technical Analysis

The unit analyzes historical data by calculating various indicators, focusing on the most popular indicators that can be highly efficient and intuitive in interpretation (eg sma, ema, macd, rsi ...). Through a group of artificial agents where each agent acts as a specific subset of technical indicators. Output is a set of technical characteristics derived from price data.

Classification of models

This module will catalog and compile previously formed models and technical indicators and their future results, classify them based on results, favouring models followed by price spikes, sort and record in database

Matching models

The function of this module is to match all the models in the database with current models and extract them if they are formed on the same asset again

Signals generation

This module converts the extracted models into input signals and concludes exit signals from technical indicators of the shares held

Portfolio management

This unit translates the entry and exit signals generated into the sale and purchase orders and the inclusion of the proposed quantities according to a strategy that allows maximizing the profit with the distribution and reduction of the risk so as not to exceed certain ratios of available liquidity by means of proven equations on previous data

Performance evaluation

The function of this unit is to calculate the total return on the virtual portfolio over different periods of time and compare these figures against market returns to measure the performance of the model relative to the market

Definitions

(1) Artificial intelligence* Is the specific behavior and characteristics of software that makes it simulate human mental capacities and modes of action. The most important of these characteristics is the ability to learn, infer and react to situations that have not been programmed into the machine."wikipedia"

(2) Neural networks* Are computational techniques designed to mimic the way the human brain performs a task by massively distributed parallel processing consisting of simple processing units. These are only computational elements called Nodes, Neurons, In that it stores practical knowledge and empirical information to make it available to the user by adjusting the weights. "wikibooks"

(3) Evolutionary algorithms* Is a subset of evolutionary calculations, a general rule of the population to solve the general problem of algorithm examples. The evolutionary algorithm uses some mechanisms inspired by biological evolution: cloning, mutation, recombination, and selection. "wikibooks"

BitPeaks R&D Engineering dept.