KR20210097474A - Calculation System and Method for dealing a digital currency, and Program Recording Medium - Google Patents

Abstract

The present invention relates to a technology for predicting transaction information based on empirical machine learning using information of an automatic transaction system of virtual currency. Based on the moving average line and transaction volume, which are transaction information in virtual currency, an empirical method is learned through regression analysis, which is a machine learning method. Based on the learned information, it is possible to decide whether to buy, sell, or hold virtual currency, and allow determining mid- to long-term trading strategies in virtual currency transaction.

Description

A virtual currency transaction information calculation system and a method for calculating virtual currency transaction information using the same, and a recording medium of the program {Calculation System and Method for dealing a digital currency, and Program Recording Medium}

The present invention relates to a technology for predicting transaction information based on empirical machine learning using information of an automatic transaction system of virtual currency.

Recently, as interest in virtual currency (cryptocurrency) based on block chain encryption technology has rapidly increased, not only holding virtual currency for investment purposes, but also increasing its utility as a means of exchange in the real economy online and offline. Due to the increased interest in these virtual currencies, more than 1000 types of virtual currencies are currently being created and distributed. Representatively, 'Bitcoin', which was first developed in 2008, and 'Ethereum', which is relatively actively traded ', 'Ripple', 'Dash', 'Litecoin', etc.

Cryptocurrencies other than Bitcoin are collectively referred to as altcoins. According to Coin Market Cap, a website that provides information on cryptocurrencies, as of June 2018, the There are 1628 bars, and there are currently 1 bitcoin and 1627 altcoins in the world market. In the case of these altcoins, as of December 2017, 1,370 were traded in the global market, and virtual currencies are continuously being created and pouring into the market. However, in the case of virtual currencies including these bitcoins and altcoins, there are insufficient points to play a role as money in the current on/offline real economy. In the case of Bitcoin, it is used as currency in e-commerce such as the Internet in countries such as the UK, Germany, and the United States, or as cash in offline commerce. Although various attempts are being made to use Bitcoin as a payment method, such as using a credit card as a payment method, the increase is not large, and you can see that there are still many countries around the world that do not recognize Bitcoin as a currency. It is difficult to deny that there are still many difficulties in being settled as a payment method in real life like the actual currency of each country. In the case of users who own virtual currency, there are cases where only one type of virtual currency is owned and used. Bar, it is often the case that one person owns several types of cryptocurrencies. Recently, the number of exchanges that sell and buy the value of virtual currency through online transaction and exchange the value accordingly is increasing significantly. It is also a reality that it is positioned as a model of new value exchange through The need for a new technique for analyzing and predicting transaction information for these virtual currency transactions is growing at the same time.

Korean Patent Publication No. 10-2019-0143105 “CURRENCY PLATFORM SYSTEM FOR MULTIPLE CRYPTOCURRENCY VALUATION by Value Evaluation” Korean Patent Laid-Open Patent No. 10-2020-0003973 “CRYPTOCURRENCY TRANSACTION SYSTEM AND METHOD OF CRYPTOCURRENCY TRANSACTION SERVICE”

The present invention has been devised to solve the above needs, and an object of the present invention is to learn an empirical method through regression analysis, which is a machine learning method, based on a moving average line and transaction volume, which are transaction information in virtual currency, and learn It is to provide a forecasting system and method to determine whether to buy, sell, or hold a virtual currency based on the obtained information, and to determine a mid- to long-term trading strategy in a virtual currency transaction.

As a means for solving the above-described problem, in the embodiment of the present invention, as shown in FIG. 1 , in a system for calculating virtual currency transaction information for operating an electronic device including at least one processor, the processor comprises: It is a processor that collects information of a virtual currency exchange through wired/wireless communication, converts it into big data, and determines whether to trade a virtual currency that classifies big data, wherein the system for calculating the virtual currency transaction information includes at least one or more virtual currency transactions A plurality of virtual currency exchange server group 100 including a database for storing transaction information of virtual currency of the virtual currency exchange that generates information; a big data server 200 for receiving time series data information including transaction information for virtual currency from the virtual currency exchange server group 100 and converting it into big data; Based on the transaction information provided by the big data server 200, a learning data set is formed using time-series data including information on the transaction volume for virtual currency for the virtual currency, transaction chart, and moving average line as input values. and repeating machine learning by performing regression analysis using the learning dataset, calculating a cost function according to the regression analysis and a parameter that minimizes it, and using the calculated parameters, input new data To provide a virtual currency transaction information calculation system, including; a virtual currency transaction information calculation processor 300, which calculates transaction information for virtual currency for 'buy, sell, and hold'.

According to an embodiment of the present invention, an empirical method is learned through regression analysis, a machine learning method, based on the moving average line and transaction volume, which are transaction information in virtual currency, and buying and selling of virtual currency based on the learned information , it has the effect of determining whether to hold or not, so that mid- to long-term trading strategies can be determined in virtual currency trading.

1 is a block diagram showing the configuration of a virtual currency transaction information calculation system according to an embodiment of the present invention.
2 to 4 exemplify a screen displaying transaction information for a virtual currency of a virtual currency exchange.
5 is a flowchart of a method of applying the virtual currency transaction information calculation system according to an embodiment of the present invention.
6 is a conceptual diagram illustrating an example of configuring a learning data set of a virtual currency transaction information calculation system according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention for achieving the above object will be described with reference to the accompanying drawings. In this description, the same reference numerals mean the same components, and in order to facilitate the understanding of the features of the present invention to those of ordinary skill in the art, the secondary description may be omitted.

In the present specification, when one component forms a coupling relationship such as connection or coupling, or a transmission relationship such as transmission or transfer in a relationship with another component, unless there is a limitation of 'direct', a 'direct' coupling relationship It can exist not only in the form of a communication relationship or the like, but also in the form of a coupling relationship by an intermediary or a transmission relationship through which another component is related between them. Also, it is the same when terms that can connote the meaning of 'contact' such as 'on', 'above', 'under', 'below', etc. are included. Moreover, the terms indicating direction should be interpreted as relative concepts to the referenced element.

In addition, it should be noted that even if a component is expressed in a singular in the present specification, it may be used as a concept representing the entirety of a plurality of components unless it is interpreted contrary to or contradictory to the concept of the invention.

In addition, in this specification, the description of words such as 'comprises' and 'comprising' and terms derived therefrom indicate the possibility of addition, combination or combination of one or more other elements to the original element or element. It is not excluded, and furthermore, the description of words having the meaning of 'comprising', 'consisting' and the like and terms derived therefrom are addition, combination or combination of one or more other elements to the original elements or elements. As long as the original elements or elements do not lose their characteristics, functions and/or properties, the possibility of adding or combining one or more other elements should not be excluded. In the present specification, 'first', 'second', etc. should be understood as merely for distinguishing each of the terms defined by them, rather than referring to the order according to the operation, action, importance, etc. of the limited term.

1 shows a block diagram of a virtual currency transaction information calculation system (hereinafter referred to as 'the present invention') according to an embodiment of the present invention.

Referring to FIG. 1, in the present invention, in a system for calculating virtual currency transaction information for operating an electronic device including at least one processor, the processor collects information of a virtual currency exchange through wired/wireless communication and converts it into big data, It consists of a processor that classifies big data and decides whether to trade virtual currency.

In this case, the virtual currency transaction information calculation system includes a plurality of virtual currency exchange server groups ( 100), a big data server 200 that receives time series data information including transaction information on virtual currency from the virtual currency exchange server group 100 and converts it into big data, and the transaction provided by the big data server 200 Based on the information, a training data set is formed using time-series data including information on the transaction volume, transaction chart, and moving average line for the virtual currency for the virtual currency as an input value, and regression analysis is performed using the training data setter repeats machine learning to calculate a cost function or loss function according to regression analysis and a parameter that minimizes it, and using the calculated parameters, 'buy, sell, hold' for input new data It is configured to include a virtual currency transaction information calculation processor 300 that calculates transaction information for virtual currency for '.

Specifically, the plurality of virtual currency exchange server group 100 is an online or offline exchange for trading virtual currency, and as shown in FIG. 2 , various types of virtual currency (G) are traded, and for this Transaction information (transaction volume, price, etc.) is disclosed online, and information on the trend and status of transaction information is provided.

That is, as shown in FIG. 2, in one virtual currency exchange, a list (G) of various types of virtual currency to be traded is listed up, and brief briefing information (price, fluctuation range, etc.) is presented. Various transaction information, such as sales volume, purchase volume, and transaction execution history, is also displayed. 2 shows transaction information for Bitcoin among virtual currencies traded in a domestic virtual currency exchange.

The most important factors for such virtual currency transaction information are price and transaction volume (chart information). Unlike the consideration of various financial information factors of a company that are considered in the existing stock transaction, the value of the virtual currency itself is a factor of price and transaction volume. It is different from the existing stock transaction information in that it is possible to grasp the trend of the transaction.

In particular, in the case of the transaction chart, it is possible to grasp the transaction information of a year in the Bitcoin transaction chart in FIG. 3 in general, and it is also possible to grasp the monthly transaction trend as in FIG. 4 .

In the present invention, transaction information, transaction volume, and chart information about the virtual currency stored in the server of the virtual currency exchange are collected, stored in the big data server, classified and analyzed, and the corresponding information is obtained during a specific period through machine learning. It enables machine learning through continuous regression analysis of information such as cryptocurrency trading volume and charts (daily chart, monthly chart, etc.).

That is, as shown in FIG. 1 , the big data server 200 collects and stores data on virtual currency provided from a plurality of virtual currency exchange server groups 100 , and the stored information in the big data server 200 . is transmitted from the virtual currency transaction information calculation processor 300 according to the present invention to provide predictive information on whether a transaction is made for virtual currency.

The electronic device in the present invention may omit at least one of the memory, communication module, input module, output module, and/or database components or may additionally include other components. For example, in one example, the virtual currency exchange server shown in FIG. 1 includes a database, which is provided outside the electronic device, and the electronic device is communicatively connected to the database to read data.

The 'processor' may, for example, execute an operation or data processing related to control and/or communication of at least one other component of the electronic device. For example, the processor may drive or execute software or programs stored therein or stored in a memory.

For example, the processor includes all of the modules shown in FIG. 1 , such as the time series data collection module 310 , the time series data classification module 320 , the learning data set generation module 330 , the regression analysis learning module 340 , etc. Alternatively, some operations may be performed. In addition, the processor may further perform the operations of the parameter calculation and storage module 350 , the new data input module 3560 , and the prediction value calculation module 250 shown in FIG. 1 .

In the present invention, "~ module" may be each or a part of a separate processor, program or software, or a combination thereof, or all may be a single processor, program or software, or a combination thereof.

An embodiment in which the time series data collection module 310 shown in FIG. 1 is omitted is possible. In this case, the function of the time series data collection module 310 may be performed by the time series data classification module 320 .

When two or more parts or all of the training data set generation module 330, the regression analysis learning module 340, the prediction calculation module 240, and the parameter calculation and storage module 350 are formed of one processor, each It can be divided into a method of driving different programs, or it can be divided into a method of driving different program modules in which separate operations belonging to the same program are performed. It may consist of a set of different program modules to be used, but is not limited thereto.

The virtual currency transaction information calculation processor 300 in the present invention collects time-series data including information on the transaction volume, transaction chart, and moving average line for the virtual currency provided by the virtual currency exchange server group 100. The time series data collection module 310 to store, and the data collected by the time series data collection module 310 are classified as chart information that displays transaction information of virtual currency, and the change form and transaction volume information of the daily chart, the monthly chart of Time series data classification module 320 for classifying into transaction volume information, based on the data classified in the time series data classification module 320, to calculate information on buying, selling, and holding based on the price and transaction volume information of virtual currency A learning data set generation module 330 that implements a learning data set that The regression analysis learning module 340 to calculate, a parameter calculation storage module 350 for storing the parameters calculated by the regression analysis learning module 340, and the newly input data and the parameter calculation storage module 350 It may be configured to include a prediction value calculation module 370 that calculates transaction information for virtual currency for 'buy, sell, hold' for new data input by applying the parameters stored in the .

The time series data collection module 310 performs a function of collecting and storing time series data among the transaction information of virtual currency stored in the big data server 200 . In this case, the time series data may refer to data arranged into a series according to a predetermined standard by observing events that occur continuously over time at regular time intervals. The time series data may include, for example, data in various formats, such as chart data, text data, and relational database (RDB) for virtual currency transactions. The time unit of the time series data may be various time intervals, such as a monthly unit, an annual unit, a semi-annual unit, a quarter unit, a weekly unit, and a daily unit, but is not limited thereto. In the following, time series data at monthly intervals will be described as an example, but this is exemplary.

The collected time series data is classified as information necessary for the factor of the learning data set for analyzing the transaction information of the virtual currency in the time series data classification module 320 . In particular, transaction information of virtual currency includes transaction volume and chart information (daily and monthly chart information).

The method implemented by the prediction system for the transaction information of the virtual currency implemented in the present invention is a mid- to long-term trading method based on the moving average line and the transaction volume among the transaction information in the virtual currency, but the empirical method is a machine learning method regression analysis ( LOGISTIC REGRESSION).

In other words, it uses information on the shape and volume of daily fluctuations formed around the 5, 10, 20, 60, and 120 lines of the daily chart among the moving averages for a specific virtual currency, and also uses the 5, 10, 20, 60, It is summarized in a way of making three decisions in the mid- to long-term by taking the form of monthly wage fluctuations at the 120 line as input. Of course, it can be configured with only two methods of deciding whether to buy or sell.

Meanwhile, as a method of realizing whether a transaction is made, transaction data is built through machine learning that reflects the regression analysis of the transaction volume and chart of transaction information for a certain period of time for a specific virtual currency.

That is, in the learning data set creation module 330, in the time series data for virtual currency transactions (for example, FIG. 3), 5, 10, 20, 60, 120 lines of the daily chart among the moving average lines for a specific virtual currency A learning data set can be constructed by using the form of daily wage fluctuations and trading volume information formed around .

Taking Figure 6 as an example, in the time series data for virtual currency transactions, ten price indicators X0, X1 to X9 including the lower low (2 or 3) and the upper high (2 or 3) After setting the , including transaction volume information, and having a data set, it is possible to repeatedly learn the experience value to determine whether to sell or not, and buy (1), sell (0), and hold (2) as labels. By forming multiclass classification values like In this case, a minimum of about 200 to 300 data and a maximum of about 1000 to 2000 data of the low point and the high point may be formed. As an example, the training data set can be formed so that daily chart values (price and trading volume) of a certain length become X values, and 1 (buy), 0 (sell), or 2 (hold) are assigned as labels. Of course, it is obvious that the meaning of the label number is only an example, and is not a fixed value, and can be assigned a meaning in 0 (buy), 1 (sell), 2 (hold) or other ways.

Such machine learning is implemented through the regression analysis learning module 340, and in LOGISTIC REGRESSION (regression analysis), HYPOTHESIS (estimated hypothesis) is expressed as follows.

{Formula 1}

Also, the corresponding cost function is expressed as follows.

{Equation 2}

In addition, the theta(θ) value of the parameter that minimizes this can be obtained by the gradient descent method as follows.

{expression 3}

If there are multiple classification targets, the one-vs-all multiclass classification method can be used as follows.

{Equation 4}

The parameters learned and calculated by the above regression analysis learning module 340 are stored in the parameter calculation and storage module 350, and then, the virtual currency based on new information input through the new data input module 360. Strategic determination of mid- to long-term selling, buying, or holding may be calculated through the prediction value calculation module 370 .

The method of determining transaction information of virtual currency using the system of the present invention as described above includes the first step of collecting and classifying time-series data on virtual currency transaction information provided from a plurality of virtual currency exchange servers, and the virtual currency Step 2 of implementing a learning data set that calculates information on buying, selling, and holding based on the price and transaction volume information of virtual currency by taking time-series data on transaction information, and performing regression analysis with the learning data set as input A method performed including a third step of calculating and storing a parameter that minimizes the cost function through can be implemented as

In the above, the above-described embodiments and the accompanying drawings are illustratively described to help those of ordinary skill in the art for the present invention, rather than limiting the scope of the present invention. That is, various embodiments of the present invention may be implemented in various modified forms according to various combinations of the above-described components without departing from the essential characteristics of the present invention. Accordingly, the scope of the present invention should be interpreted according to the invention described in the claims, and includes various modifications, alternatives, and equivalent embodiments by those skilled in the art as well as the above-described embodiments.

100: virtual currency exchange server
200: big data server
300: cryptocurrency transaction information calculation module
310: time series data collection module
320: time series data classification module
330: training data set generation module
340: regression analysis learning module
350: parameter calculation and storage module
360: new data input module
370: prediction value calculation module

Claims (7)

In the virtual currency transaction information calculation system for operating an electronic device including at least one processor,
The processor is a processor that collects information of a virtual currency exchange through wired/wireless communication, converts it into big data, and classifies the big data to determine whether to trade virtual currency,
The virtual currency transaction information calculation system,
a plurality of virtual currency exchange server groups 100 including a database for storing transaction information of virtual currency of a virtual currency exchange that generates transaction information for buying and selling at least one virtual currency;
a big data server 200 for receiving time series data information including transaction information for virtual currency from the virtual currency exchange server group 100 and converting it into big data;
Based on the transaction information provided from the big data server 200,
Forming a learning data set with time-series data including information on the transaction volume, transaction chart, and moving average line for the virtual currency as an input value,
Regression analysis is performed using the learning data set and machine learning is repeated, and a cost function according to the regression analysis and a parameter to minimize it are calculated and stored as parameters,
a virtual currency transaction information calculation processor 300 for calculating transaction information on whether to buy, sell, or hold virtual currency based on the new data that is input by using the calculated parameter;
Cryptocurrency transaction information calculation system comprising a.
The method according to claim 1,
The virtual currency transaction information calculation processor 300,
a time-series data collection module 310 for collecting and storing time-series data including information on a transaction volume, a transaction chart, and a moving average line for virtual currency provided by the virtual currency exchange server group 100;
A time series data classification module 320 that classifies the data collected by the time series data collection module 310 and classifies it into chart information displaying transaction information of virtual currency into daily chart fluctuation type and transaction volume information, and monthly chart transaction volume information );
From the data classified by the time series data classification module 320, a learning dataset creation module 330 that implements a learning data set that allows information on buying, selling, or holding to be calculated based on the price and transaction volume information of virtual currency. ;
a regression analysis learning module 340 that performs regression analysis based on the training data set and repeats machine learning to calculate a cost function according to the regression analysis and a parameter for minimizing it as parameters;
a parameter calculation storage module 350 for storing the parameters calculated by the regression analysis learning module 340;
a prediction value calculation module 370 for predicting whether a virtual currency will be traded by applying the newly input data and the parameters stored in the parameter calculation and storage module 350;
Cryptocurrency transaction information calculation system comprising a.
3. The method according to claim 2,
The regression analysis learning module 340,
A virtual currency transaction information calculation system in which the hypothesis of the following {Equation 1} is defined, and the cost function is defined by {Equation 2}.
{Formula 1}

{Equation 2}

4. The method according to claim 3,
The theta (0) value of the parameter that minimizes the cost function is a virtual currency transaction information calculation system calculated by the following {Equation 3}.
{expression 3}

5. The method according to claim 4,
The training data set is
A virtual currency transaction information calculation system, characterized in that the daily chart value of a certain length is used as the X value, and any one of 1 (buy), 0 (sell), and 2 (hold) is assigned as a label.
In the method of calculating the transaction information of virtual currency by applying the virtual currency transaction information calculation system of any one of claims 1 to 5,
Step 1 of collecting and classifying time-series data on virtual currency transaction information provided by a plurality of virtual currency exchange servers;
a second step of implementing a learning data set that calculates time-series data on the virtual currency transaction information to calculate information on buying, selling, and holding based on the price and transaction volume information of virtual currency;
a third step of calculating and storing parameters for minimizing a cost function through regression analysis with the learning data set as input;
a fourth step of determining whether to buy, sell, or hold a virtual currency by comparing the newly input data value with the parameters of the third step;
A method of calculating cryptocurrency transaction information, including a.
A computer-readable recording medium in which a program for implementing the method of calculating the virtual currency transaction information according to claim 6 is recorded.

Images