KR102590232B1 - System for providing day trading stock picking service using bigdata based artificial intelligence - Google Patents

Abstract

A system to provide a recommendation service for short-term trading items using big data-based artificial intelligence is provided, and data including stock price information, stock index, and real-time issues for user terminals and domestic listed items that are recommended for stocks with rising potential among domestic listed items before the market starts. A big data department that builds big data by accumulating, an analysis department that analyzes big data with at least one artificial intelligence algorithm for short-term trading, and a department to minimize the margin of error of domestic listed stocks extracted from at least one artificial intelligence algorithm. A filtering unit that performs filtering using a pre-established filtering algorithm, recommends domestic listed stocks, excluding domestic listed stocks extracted from at least one artificial intelligence algorithm and filtered by the filtering algorithm, as the likely rising stocks of the day before the market starts. It includes a recommendation service providing server including a short-term recommendation unit.

Description

System for providing recommendation service for short-term trading items using big data-based artificial intelligence {SYSTEM FOR PROVIDING DAY TRADING STOCK PICKING SERVICE USING BIGDATA BASED ARTIFICIAL INTELLIGENCE}

The present invention relates to a system for providing a recommendation service for short-term trading items using big data-based artificial intelligence. It provides a system that analyzes big data on domestic listed items with an artificial intelligence algorithm, extracts stocks with rising potential, and provides them before the market starts. do.

With the advent of fintech, non-face-to-face account opening has become useful for individual investors, leading to a rapid increase in investment account openings. However, there are several limitations in the ability of general investors, who make up the majority of stock trading accounts, to make profits by investing in stocks against institutions and foreigners. This exists. There is information asymmetry, and it is almost impossible for ordinary investors who work both jobs to respond to the market in a timely manner. Accordingly, research is being conducted to analyze and predict the stock market to meet the needs of investors. The two main elements of stock prediction programs using artificial intelligence are big data and artificial intelligence models. Big data and artificial intelligence are closely related, and humans extract useful information by analyzing certain patterns in data and exercising insight based on experience. With the development of advanced computing capabilities and artificial intelligence models, artificial intelligence replaces humans. was able to fulfill this role. As computers with far superior computational capabilities than humans replaced humans, extensive and large-scale analysis of big data that could not be attempted with human capabilities became possible, giving humans the possibility of predicting the future.

At this time, a method of recommending stocks and evaluating recommended algorithms using big data and artificial intelligence was researched and developed. In this regard, Korea Patent Publication No. 2022-0097189 (published on July 7, 2022), which is a prior art, was developed. and Korea Registered Patent No. 10-1975448 (announced on May 8, 2019), in order to implement a stock trading recommendation method using big data-based valuation, financial information for listed companies is obtained from the electronic disclosure system based on the fiscal year. Statements are collected and stored in a database, and the per-share value of stocks for each listed company is evaluated based on the financial statements of each listed company. Among listed companies, listed companies are undervalued and have a current stock price that is lower than the per-share value evaluated by the stock valuation module. A structure that selects companies and recommends the purchase of stocks of selected listed companies, builds stock big data, and evaluates the hit rate based on artificial intelligence for the algorithm that recommends products with high investment value among products. The configuration of calculating the evaluation value of each element data, adding weight to the element data for each product, adding up the element data, setting the expected profit ranking in the order of the largest sum score, and calculating the hit rate of the actual expected profit order is initiated. It is done.

However, the former configuration does not reflect the current market situation and profit rate because it evaluates the value per share based on last year's financial statements, and the latter configuration only evaluates the accuracy of the algorithm and does not recommend stocks to purchase. Stock prices are affected by real-time issues due to internal and external reasons. The internal cause is that, under the premise that humans are not perfect, the confusion and fear that arise from a combination of human emotions, uncertainty, and a vast amount of information greatly deteriorates a person's judgment and makes rational thinking impossible. The external cause lies in the investment psychology combined with the investor's internal factors. Stock prices, which are the object of investment sentiment, are bound to show movements that reflect this. Accordingly, after analyzing the big data of domestic listed stocks with an artificial intelligence algorithm suitable for short-term trading, filtering is performed to minimize the margin of error for domestic listed stocks analyzed as suitable for short-term trading, thereby determining the day's upside potential in advance before the market starts. Research and development of methods to extract stocks are required.

One embodiment of the present invention analyzes big data of domestic listed stocks with an artificial intelligence algorithm suitable for short-term trading, then performs filtering to minimize the margin of error for domestic listed stocks analyzed as suitable for short-term trading, and starts the market. We can provide a short-term trading stock recommendation service system using big data-based artificial intelligence that can extract the day's likely rising stocks in advance. However, the technical challenge that this embodiment aims to achieve is not limited to the technical challenges described above, and other technical challenges may exist.

As a technical means for achieving the above-described technical task, an embodiment of the present invention provides a user terminal that receives recommendations for stocks with rising potential among domestic listed stocks before the start of the market, as well as stock price information, stock indexes, and real-time issues for domestic listed stocks. A big data department that builds big data by accumulating the included data, an analysis department that analyzes big data with at least one artificial intelligence algorithm for short-term trading, and minimizes the margin of error for domestic listed stocks extracted from at least one artificial intelligence algorithm. In order to do so, a filtering unit that performs filtering using a pre-established filtering algorithm, selects domestically listed stocks excluding domestic listed stocks extracted from at least one artificial intelligence algorithm and filtered by the filtering algorithm as the likely rising stocks of the day before the market starts. It includes a recommendation service providing server including a short-term recommendation unit that makes recommendations.

According to one of the means for solving the problem of the present invention described above, the big data of domestic listed items is analyzed with an artificial intelligence algorithm suitable for short-term trading, and then the error range of domestic listed items analyzed as suitable for short-term trading is minimized. By performing filtering, you can extract the day's likely rising stocks in advance before the market starts.

Figure 1 is a diagram illustrating a system for providing a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating a recommendation service providing server included in the system of FIG. 1.
Figures 3 and 4 are diagrams to explain an embodiment of a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment of the present invention.
Figure 5 is an operation flowchart illustrating a method of providing a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment of the present invention.

Below, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts that are not related to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout the specification, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected," but also the case where it is "electrically connected" with another element in between. . In addition, when a part is said to "include" a certain component, this does not mean excluding other components unless specifically stated to the contrary, but may further include other components, and one or more other features. It should be understood that it does not exclude in advance the presence or addition of numbers, steps, operations, components, parts, or combinations thereof.

The terms "about", "substantially", etc. used throughout the specification are used to mean at or close to that value when manufacturing and material tolerances inherent in the stated meaning are presented, and are used to enhance the understanding of the present invention. Precise or absolute figures are used to assist in preventing unscrupulous infringers from taking unfair advantage of stated disclosures. The term “step of” or “step of” as used throughout the specification of the present invention does not mean “step for.”

In this specification, 'part' includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more pieces of hardware, and two or more units may be realized using one piece of hardware. Meanwhile, '~ part' is not limited to software or hardware, and '~ part' may be configured to reside in an addressable storage medium or may be configured to reproduce one or more processors. Therefore, as an example, '~ part' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, and procedures. , subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and 'parts' may be combined into a smaller number of components and 'parts' or may be further separated into additional components and 'parts'. Additionally, components and 'parts' may be implemented to regenerate one or more CPUs within a device or a secure multimedia card.

In this specification, some of the operations or functions described as being performed by a terminal, apparatus, or device may instead be performed on a server connected to the terminal, apparatus, or device. Likewise, some of the operations or functions described as being performed by the server may also be performed in a terminal, apparatus, or device connected to the server.

In this specification, some of the operations or functions described as mapping or matching with the terminal mean mapping or matching the terminal's unique number or personal identification information, which is identifying data of the terminal. It can be interpreted as

Hereinafter, the present invention will be described in detail with reference to the attached drawings.

Figure 1 is a diagram illustrating a system for providing a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment of the present invention. Referring to Figure 1, the short-term trading item recommendation service providing system 1 using big data-based artificial intelligence includes at least one user terminal 100, a recommendation service providing server 300, and at least one institutional server 400. may include. However, since the short-term trading item recommendation service providing system 1 using big data-based artificial intelligence shown in FIG. 1 is only an embodiment of the present invention, the present invention is not limited to the interpretation of FIG. 1.

At this time, each component of FIG. 1 is generally connected through a network (Network, 200). For example, as shown in FIG. 1, at least one user terminal 100 may be connected to the recommendation service providing server 300 through the network 200. In addition, the recommendation service providing server 300 may be connected to at least one user terminal 100 and at least one organization server 400 through the network 200. Additionally, at least one institutional server 400 may be connected to the recommendation service providing server 300 through the network 200.

Here, the network refers to a connection structure that allows information exchange between each node, such as a plurality of terminals and servers. Examples of such networks include a local area network (LAN) and a wide area network (WAN). Wide Area Network, Internet (WWW: World Wide Web), wired and wireless data communication network, telephone network, wired and wireless television communication network, etc. Examples of wireless data communication networks include 3G, 4G, 5G, 3rd Generation Partnership Project (3GPP), 5th Generation Partnership Project (5GPP), Long Term Evolution (LTE), World Interoperability for Microwave Access (WIMAX), and Wi-Fi. , Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth network, NFC ( It includes, but is not limited to, Near-Field Communication (Near-Field Communication) network, satellite broadcasting network, analog broadcasting network, and DMB (Digital Multimedia Broadcasting) network.

In the following, the term at least one is defined as a term including singular and plural, and even if the term at least one does not exist, each component may exist in singular or plural, and may mean singular or plural. This should be self-explanatory. In addition, whether each component is provided in singular or plural form may be changed depending on the embodiment.

At least one user terminal 100 may be a terminal of a user who receives recommendations for stocks with rising potential before the start of the market using a web page, app page, program, or application related to a short-term trading item recommendation service using big data-based artificial intelligence. . In addition, the user terminal 100 may be a terminal that receives an emergency sale notification from the recommendation service providing server 300 when a real-time issue is analyzed in real time and a decline is predicted, and when an increase is predicted, the recommendation service providing server 300 It may be a terminal that receives a purchase notification from.

Here, at least one user terminal 100 may be implemented as a computer capable of accessing a remote server or terminal through a network. Here, the computer may include, for example, a laptop equipped with a navigation system and a web browser, a desktop, a laptop, etc. At this time, at least one user terminal 100 may be implemented as a terminal capable of accessing a remote server or terminal through a network. At least one user terminal 100 is, for example, a wireless communication device that guarantees portability and mobility, and includes navigation, personal communication system (PCS), global system for mobile communications (GSM), personal digital cellular (PDC), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) ) It may include all types of handheld-based wireless communication devices such as terminals, smartphones, smartpads, and tablet PCs.

The recommendation service providing server 300 may be a server that provides a short-term trading item recommendation service web page, app page, program, or application using big data-based artificial intelligence. In addition, the recommendation service providing server 300 may be a server that builds big data on domestic listed stocks and extracts stocks with rising potential by analyzing the big data with an artificial intelligence algorithm. At this time, the recommendation service providing server 300 may be a server that filters stocks with a potential to rise by considering the margin of error and extracts the remaining domestic listed stocks as stocks with a potential to rise. In addition, the recommendation service providing server 300 may be a server that extracts and filters using an artificial intelligence algorithm and transmits the remaining rising potential stocks to the user terminal 100 before the market starts.

Here, the recommendation service providing server 300 may be implemented as a computer that can connect to a remote server or terminal through a network. Here, the computer may include, for example, a laptop equipped with a navigation system and a web browser, a desktop, a laptop, etc.

At least one institutional server 400 is a terminal of an institution that sells or buys domestically listed items with or without using a web page, app page, program, or application related to a short-term trading item recommendation service using big data-based artificial intelligence. You can. Here, at least one institutional server 400 may be implemented as a computer capable of accessing a remote server or terminal through a network. Here, the computer may include, for example, a laptop equipped with a navigation system and a web browser, a desktop, a laptop, etc.

Figure 2 is a block diagram for explaining the recommendation service providing server included in the system of Figure 1, and Figures 3 and 4 show a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment of the present invention. This is a drawing to explain an implemented embodiment.

Referring to FIG. 2, the recommendation service providing server 300 includes a big data unit 310, an analysis unit 320, a filtering unit 330, a short-term recommendation unit 340, a real-time issue notification unit 350, and a natural It may include an analysis unit 360, a correlation analysis unit 370, an investment monitoring unit 380, a damage prediction unit 390, a market impact model unit 391, and a performance analysis unit 393.

The recommendation service providing server 300 or another server (not shown) operating in conjunction with an embodiment of the present invention provides big data-based artificial intelligence to at least one user terminal 100 and at least one institutional server 400. When transmitting a short-term trading item recommendation service application, program, app page, web page, etc., at least one user terminal 100 and at least one institutional server 400 perform short-term trading using big data-based artificial intelligence. You can install or open stock recommendation service applications, programs, app pages, web pages, etc. Additionally, a service program may be run on at least one user terminal 100 and at least one organization server 400 using a script executed in a web browser. Here, a web browser is a program that allows the use of web (WWW: World Wide Web) services and refers to a program that receives and displays hypertext written in HTML (Hyper Text Mark-up Language), for example, Netscape. , Explorer, Chrome, etc. Additionally, an application refers to an application on a terminal and includes, for example, an app running on a mobile terminal (smartphone).

Referring to FIG. 2, the big data unit 310 can build big data by accumulating data including stock price information, stock index, and real-time issues for domestic listed items. At this time, stock price information refers to information related to stock prices, including stock code, date, current stock price, fluctuation rate compared to the previous day, market price, high price, low price, trading volume, previous day's price, previous day's trading volume, upper limit price, lower limit price, PER (Price Earnings Ratio) ), sell price, buy price, 52-week high, 52-week low, PBR (Price to Book Ratio), foreigner ratio, par value, capital, number of listed stocks, market capitalization, listing date, selling balance, buying balance, etc. You can. Numerous stock items are traded in the stock market. In the case of stock trading, it is carried out by item, but when looking at the stock market as a whole, there are so many traded items that it is difficult to determine the price trends of the entire market. Therefore, in creating a stock index, If you create a comprehensive stock index that combines the stock price movements of various stocks, you can easily display the price trends of various stocks by grouping them by industry, size, or various other categories. Real-time issues include international, political, and financial issues. It may be an issue, but it is not limited to this.

The analysis unit 320 can analyze big data using at least one artificial intelligence algorithm for short-term trading. First, artificial intelligence algorithms can be used to recommend the timing of recommending stocks with rising potential by predicting the point in time when an upward trend transition occurs in a phase where the entire stock market is falling. A plummeting market cannot plummet indefinitely. Around the time when investors' fear reaches its peak, a sell-off occurs and the decline is maximized, a new buying trend based on rational reason, called smart money, is formed, ultimately leading to an upward trend shift. Neural network-based stock price prediction models have a strong tendency to assign high prediction scores to crash-type patterns. As the overall market decline progresses, the number of individual stocks corresponding to a slump-type pattern gradually increases, and as a result, the number of stocks assigned high prediction scores by the neural network-based stock price prediction model also increases. Accordingly, in one embodiment of the present invention, the statistical distribution of neural network prediction scores at the time of a crash and trend change can be analyzed, and based on this, the time of an upward trend change can be predicted for an unexperienced test period.

<Neural network-based trend reversal prediction>

From a technical analysis perspective, the main patterns in which stock prices turn from a downward trend to an upward trend are double bottom, triple bottom, inverse head and shoulders, and round bottom. There is a Rounded Bottom and a Spike Bottom. Among these, the Spike Bottom is also called the V-Formation. Among these various patterns, the most dramatic transition pattern is the V-shaped transition pattern, in which a short-term sharp decline is followed by a sudden surge accompanied by the sudden appearance of a buying trend. Most of the cases in which stock price prediction models developed based on neural network learning assign high prediction scores are crash-type patterns corresponding to the first half of a V-shaped pattern. Of course, the point in time when a crash-type pattern appears most often is when the entire stock market plummets due to global bad news or unexpected events.

For example, from early March to mid-March 2020, most of the world's stock markets plummeted, and what caused this crash was the coronavirus. The general process of a V-shaped trend reversal is as follows. After a decline in stock prices for a certain period of time, an excessive plunge in stock prices occurs in a short period of time, and a strong buying force emerges to resist this, leading to a V-shaped rise in stock prices, and then the trend itself turns into an upward trend. In this process, the excessive decline in stock prices started from fundamental factors such as a general recession in the company's industry or a temporary deterioration in corporate performance, and later was caused by a collective sell-off by investors whose investment sentiment had worsened over a long period of time. In many cases, it is due to A plunge in stock prices that excessively reflects crowd psychology leads to the emergence of a new crowd that identifies the excessive undervaluation of stock prices on rational grounds, ultimately leading to an upward shift in the stock price trend.

Even in a sideways market or a trend rising market rather than a crash market, there are almost always stocks that fall into a slump pattern among individual stocks, and according to trading simulations, the average profit per trade (PPT) exceeds 3%. It is possible to achieve a level of trading performance, but in a sideways market or a general uptrend rather than a plunge in the market as a whole, a V-shaped trend reversal does not necessarily occur just because the stock price of a specific individual company has fallen. For example, there are many cases where a company is designated as a managed stock or, in extreme cases, undergoes the process of delisting due to significant damage to corporate value, such as embezzlement and breach of trust by management, erosion of equity capital, and rejection of audit opinions. There are cases where trading of the stock in question was suspended as a preliminary measure before deciding whether to delist, and since the stock price fell significantly just before the trading suspension section, the existing prediction model assigned a fairly high prediction score to this section. do.

In this way, an excessive plunge in stock prices in the general market may lead to a V-shaped rebound and upward trend change, or conversely, it may cause enormous losses due to the uncertainty of the company's existence. Therefore, even if a neural network-based stock price prediction model can predict a V-shaped rebound pattern with a high probability, normal investment cannot be made in general market conditions without additional information other than technical analysis that can cope with the serious risks described above. In the end, it can be concluded that the best section to detect and utilize a V-shaped trend reversal is the section of a slumping market in which the entire market plunges. A technical rebound can be expected because the overall market crash, such as the coronavirus incident, is greatly influenced by psychological factors such as investor anxiety and fear. However, since the definition of a slump market is a period in which stock prices continue to fall rapidly, it is a very difficult task to find out where it ends.

Accordingly, in one embodiment of the present invention, a neural network-based prediction model detects a V-shaped trend reversal using statistical information of prediction scores assigned to each item and uses this for trading. Using a neural network-based prediction model, the sum of the prediction scores of stocks with high prediction scores can be used as a means of predicting the end of a slump. As the crash market progresses, the number of stocks that have crashed will increase, but as the extent of the crash will become larger, the prediction scores for stocks that have crashed will also gradually increase. The question is how much higher the sum of the predicted scores can be considered to be the end of the crash. In order to examine this issue, the process of changing the prediction scores of the prediction model learned using previously studied input features can be analyzed for past crash sections.

Let's define the sum of the prediction scores of the prediction model for the top i (Item, i) stocks with high prediction scores as SS(Score Sum, SS)i. For example, the change process of the KOSDAQ index can be used. The reason for looking at the KOSDAQ index instead of the KOSPI index is that the KOSDAQ index fluctuates more rapidly than the KOSPI index, and because of this characteristic, it is better to closely observe the trend transition process of the stock index. Because it is suitable. As a result of checking the data for the end of the crash market that hit the lowest, if SS30, which is the sum of the prediction scores of the top 30 stocks, is However, in two crash markets, the threshold can be set not only to The fact that the sum of the prediction scores at two different crash points is so different means that the deviation of the scores assigned to the crash pattern is quite large, which can be seen as a lack of consistency in the prediction scores assigned by the predictor. To improve this, the consistency of the prediction model for the crash pattern can be improved by expressing the crash pattern more precisely using additional input features.

<Neural network model>

The input features used in the learning of the neural network may first include binary features representing the rise and fall of the price moving average line. For example, when the 5-day moving average of item s on trading day t is MA5 ^s t, whether the 5-day moving average line rises can be obtained as in Equation 1.

In this way, five binary input features can be used to express whether the 5, 10, 20, 60, and 120-day price moving average lines are rising, and in the same way, the 5, 20, 60, and 120-day trading volume moving average lines can be used to express whether the price moving average line is rising. Four binary input features can be used to express whether is rising.

Next, you can use the feature to compare the i-day moving average and the j-day moving average. For example, when the 10-day moving average of item s on trading day t is MA10 ^s t, the characteristic comparing the 5-day moving average and the 10-day moving average can be calculated using Equation 2.

In this way, 10 binary input features can be used to express the results of size comparison for a total of 10 moving average combinations. In the same way, five binary input features are used to express the result of comparing the closing price (current price) and the 5, 10, 20, 60, and 120-day price moving averages, and the transaction volume and 5, 20, 60, and 120-day price moving averages are used. Four binary input features can be used to express the result of comparing.

Continuous features include Disparity, Gradient, and Rate of Change. The separation degree expresses the distance between the closing price and the i-day moving average. For example, the distance between the closing price of item s on trading day t and the 5-day moving average can be calculated using Equation 3 below.

In this way, five input features can be used to express the distance between the closing price and the 5, 10, 20, 60, and 120 day moving averages. The slope expresses the slope of the i-day moving average line. For example, the slope of the 5-day moving average line of item s on trading day t can be calculated using Equation 4.

In this way, five input features can be used to express the slope of the 5-, 10-, 20-, 60-, and 120-day moving average lines. The price change rate expresses the change ratio of the current closing price compared to the closing price i days ago. For example, the change ratio of the current closing price compared to the closing price 1 day ago of item s on trading day t can be calculated using Equation 5.

In this way, nine input features can be used to express the change ratio of the current closing price compared to the closing price 1, 2, 4, 7, 12, 20, 33, 54, and 68 days ago.

In addition to the input features described above, additional input features that can be used are as follows. First, features are used to compare the distance between the i-day moving average and the j-day moving average. For example, the distance between the 5-day moving average and the 10-day moving average of item s on trading day t can be obtained using Equation 6.

In this way, 10 features can be used to express the distance between moving averages for a total of 10 moving average combinations.

Trading volume separation expresses the distance between trading volume and i-day trading volume moving average. For example, if the trading volume of item s on trading day t is volume ^s t and the 5-day moving average of trading volume is VMA5 ^s t, the trading volume and trading volume The distance from the 5-day moving average can be calculated using Equation 7 below.

In this way, four input features can be used to express the distance between trading volume and the 5-, 20-, 60-, and 120-day moving averages of trading volume. The slope of the trading volume line expresses the slope of the i-day trading volume moving average line. For example, the slope of the 5-day moving average trading volume of item s on trading day t can be calculated using Equation 8.

In this way, four input features can be used to express the slope of the 5-, 20-, 60-, and 120-day trading volume moving average line. The rate of change in trading volume expresses the rate of change in the current trading volume compared to the trading volume i days ago. For example, the rate of change in the current trading volume compared to the trading volume 1 day ago of item s on trading day t can be calculated using Equation 9.

In this way, nine input features are used to express the change ratio between the trading volume 1, 2, 4, 7, 12, 20, 33, 54, and 68 days ago and the current trading volume. The target value target ^s t can be calculated using Equation 10.

γ is the discount factor, and T indicates whether stock price changes from the current date to a few days in the future will be reflected in the target value. If γ is close to 1, even changes in the distant future are reflected, and if γ is close to 0, only changes in the very near future are reflected.

Continuous features and target values are normalized to values between 0 and 1 and used as final input features and target values. Calculate the average and standard deviation of each input feature and target value from all data used for learning, with average + 3 x standard deviation as the maximum value and average - 3 x standard deviation as the minimum value, and the original value ranges from 0 to 1. Convert linearly to a number between. If the input feature or target value exceeds the maximum value, it is processed as the maximum value, and if it is less than the minimum value, it is processed as the minimum value.

<LSTM>

In addition to the technical analysis described above, in one embodiment of the present invention, as stock price prediction factors, all three types of indicators are used: technical indicators inherent in the stock price, market sentiment indicators indicating investors' interest in certain items in the market, and macroeconomic indicators. By mixing, you can find the optimal combination and use it to predict stock prices. Although various features have been used in many studies for technical indicators and macroeconomic indicators, in one embodiment of the present invention, two representative features can be selected.

Two technical indicators may include trading volume (V) and KOSPI index (K), and two macroeconomic indicators may include exchange rate (E) and interest rate (I). And as a market sentiment indicator, the word search frequency (W) on the Internet for a specific stock can be used. In addition, since the stock price (Pt) in the LSTM model is influenced by the past stock price (Pt-n), this can be expressed in a formula as shown in Equation 11.

Here, Pt and Pt-n are the current stock price at time t and the past stock price up to time t-n, respectively, Vt-n is the past trading volume up to time t-n, Kt-n is the past KOSPI index up to time t-n, and Wt-n. is the past word search frequency up to time t-n, Et-n is the past exchange rate up to time t-n, and It-n is the past interest rate up to time t-n. And Vt-n and Kt-n are technical indicators, Wt-n is a market sentiment indicator for a specific stock, and Et-n and It-n are macroeconomic indicators.

As described above, in one embodiment of the present invention, stock price prediction can be performed to extract stocks with rising potential using an LSTM (Long Short-Term Memory) model. Deep learning algorithms are based on ANN, such as DNN (Deep Neural Network), CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), DBN (Deep Belief Network), DQN (Deep Q-Network) can also be used. Among these, RNN is suitable for learning time series data because it remembers information about past data in the hidden layer, and LSTM, developed from this, has better learning ability the farther back in the past. It is an algorithm that improves the problems of deteriorating RNNs and shows good performance in predicting nonlinear long-term time series data such as stock prices. It is an algorithm that has recently been mainly used in stock price prediction research.

In addition, there are studies confirming that the performance of LSTM is better compared to the ARIMA model, a traditional time series technique, and studies showing that the LSTM model does not always show superior performance in comparison with the VAR model and that it varies depending on the characteristics of the data. The latest machine learning algorithms are not always excellent in building models for multivariate time series prediction, but the performance of the VAR model was excellent when the data length was relatively short, and the performance of the LSTM model was excellent when the data length increased. In one embodiment of the present invention, the LSTM model is applied for the purpose of using data for a relatively long N years and predicting stock prices using a deep learning algorithm.

<Measuring accuracy of prediction>

The accuracy of stock price prediction can be confirmed through the error value resulting from the difference between the predicted value and the actual value. In one embodiment of the present invention, RMSE (Root Mean Square Error) is used. RMSE is the root mean square error, which is a commonly used measure when dealing with the difference between an estimated or predicted value and a value observed in the actual environment. Since the residuals are squared, the slope can be calculated, so the model's performance can be improved using gradient descent. It is most widely used when evaluating the accuracy of machine learning algorithms that optimize . RMSE is a method of calculating the average square of the predicted value by subtracting the actual value and then taking the square root.

<Data and parameters>

In one embodiment of the present invention, the remaining daily stock price data excluding data excluded in the pre-processing process of N years of data can be used to analyze stock price prediction factors. In one embodiment of the present invention, the final purpose is not to find and use stock price prediction values, but to review the process of finding the appropriate combination of stock price prediction elements and the appropriate number of days of leading data for stock price prediction, so the verification data set for cross-validation is separate. It may not be classified, but it does not rule it out.

There are various classifications for classifying industries in the stock market, and in one embodiment of the present invention, WICS (Wise Industry Classification Standard), which is also used by Naver Securities, can be used. WICS consists of a total of 10 major categories, 28 middle categories, and 83 small categories. In one embodiment of the present invention, stock price data can be used by randomly selecting one representative KOSPI stock for each industry based on the major categories. Word search frequency can be determined by using keywords appropriate for stock items selected from Naver Data Lab and data obtained during the same period as the stock price data. There are several types of stock prices, including opening price, high price, low closing price, etc., and you can use the adjusted close price. For exchange rates and interest rates, exchange rate data provided by the Bank of Korea's Economic Statistics System and interest rates on 3-year maturity treasury bonds can be used. In addition, when running the LSTM model, the unit of measurement is different for each input variable and the stock price range is different for each industry, so in order to learn effectively and compare results by industry, the Min-Max Normalization method is used to calculate the input data and output RMSE. It can be normalized. Additionally, in one embodiment of the present invention, epochs can be set to avoid overfitting. After predicting the stock price in this way, the results can be used for evaluation. Alternatively, suitability can be predicted from the prediction result.

The filtering unit 330 may perform filtering using a pre-established filtering algorithm to minimize the error range of domestic listed stocks extracted from at least one artificial intelligence algorithm.

The short-term recommendation unit 340 may recommend domestically listed stocks, excluding domestically listed stocks extracted from at least one artificial intelligence algorithm and then filtered by a filtering algorithm, as the likely rising stocks of the day before the market starts. The user terminal 100 can receive recommendations for stocks with rising potential among domestic listed stocks before the start of the market. In other words, it can be done in the following order: [Building big data - Artificial intelligence algorithm modeling - Analysis of domestic listed stocks - Filtering considering the margin of error - Extracting and recommending stocks with rising potential].

The real-time issue notification unit 350 recommends stocks with rising potential and, if real-time issues change after the market starts, predicts stock price changes according to real-time issues and guides the user terminal 100 to the predicted stock price changes. . Real-time issues may include, but are not limited to, international issues, political issues, and financial issues. For example, if an article comes out that an employee at an implant company embezzled 200 billion won, the stock price of the implant company is expected to plummet, tourism industry stocks are expected to plummet as the number of visitors to Korea increases due to the THAAD deployment in China, and the government's ICO Since a ban is announced and institutional sanctions are imposed, the stock prices of companies preparing for ICOs or those preparing for them are expected to fall. If such an urgent issue exists, guidance can be provided to the user terminal 100 to provide a notification to remove it if it is a stock with a potential to rise, and to guide selling if it has already been purchased. Of course, there may be types of stock companies that are not affected by real-time issues, so such data can be provided by extracting only correlated items according to the results of big data analysis.

In order to interpret real-time issues, the natural language understanding unit 360 can collect text about real-time issues and then analyze them using natural language understanding (NLU). To analyze real-time issues, you can use Natural Language Understanding (NLU), Intent, and Entity. NLU refers to understanding natural language used by people. In order to understand complex human language, you need to break sentences down into words, classify them, and understand their intent. Using natural language understanding technology, you can identify the grammatical structure of input sentences and classify the intent and entities of the news.

The correlation analysis unit 370 calculates the emotional score of each investor that can affect the stock price increase or decrease through natural language understanding, and analyzes the correlation between the emotional score and the stock price increase or decrease. If there is a correlation that exceeds a preset standard value, information on the rise or fall in stock prices for which the correlation exists may be transmitted to the user terminal 100. Since investment sentiment is also reflected in the rise or fall of stock prices, it is important to understand the investment sentiment of each investor.

The investment monitoring department 380 can monitor investment information using a previously established insolvency prediction model and credit rating calculation model to extract insolvent companies from domestic listed stocks. Here, the credit rating converts the credit rating assigned by a credit rating agency or its own credit rating system into a score according to the rating system, and provides credit rating history information so that it can be referred to in investment opinions.

At this time, the financial anomaly inspection system is a system that examines the possibility of fraud in a company's financial statements. It calculates analysis opinions and ratings by analyzing three things: Korea Federation of Banks' borrowings analysis, tax information summary table, and financial anomaly inspection model. First, we analyze loan discrepancies and omissions through comparison between the company's financial statement information and the Korea Federation of Banks' credit grant information. For example, comparative analysis using the following items: [Total borrowings (excluding private loans)], [Balance of transfer of unexpired sales receivables], [Dependency on borrowings (excluding private loans)], [Actual dependence on borrowings], [Total assets], and [Debentures]. You can calculate the corrosion sign results by doing this. In addition, the comprehensive tax information table comprehensively checks for signs of fraud by comparing the difference between the tax information amount for each analysis item and the financial statement amount. For example, you can calculate the results of fraud symptoms by comparing the differences between financial statements and tax information for [Sales], [Depreciation], [Doubtful write-down], [Retirement benefits], and [Loans] items. In addition, major financial information such as [total assets], [total capital], [sales], [operating profit], and [net profit before corporate tax] can be analyzed and reflected in the check for signs of fraud.

<Financial anomaly inspection model>

The financial outlier inspection model calculates the amount of analysis items, outlier signs, and grades for each industry. The analysis items are largely [financial costs], [transfer of trade receivables], [intangible assets], [non-operating assets and potentially non-performing assets], [working capital], [cash flow], [abnormal signs of rapid growth], and [earnings adjustment risk]. It is divided into seven categories, and detailed items are defined differently by industry. The analysis results of items by category are assigned a financial outlier rating according to the criteria for each industry (light industry, heavy industry, wholesale and retail, construction, and other service industries). The final financial outlier rating of a company is calculated based on the lowest rating for each item as the final financial outlier rating.

<Insolvency prediction model using big data>

News data, which is unstructured data, is data in text form. To analyze this data, you can first use text mining and quantification methods using artificial intelligence algorithms. For example, by using analysis modules such as machine learning-based Word2Vec and LDA to analyze and quantify the relationships between words mentioned in news articles, keywords of companies related to bankruptcy are registered and the number and ratio of reported articles are identified. Thus, an unstructured data indicator value can be assigned. At this time, keywords related to corporate bankruptcy are [insolvency], [workout], [accounting fraud], [sale], [bankruptcy], [insolvency], [workout], [court receivership], [bankruptcy], and [bankruptcy] ], [Closing of business], [Fraudulent accounting], [Delinquency], [Loss], [Debt management item], [Corporate rehabilitation], [Bill], [Fund shortage], [Exit], [Capital erosion], [Management difficulty] , [Collapse], [Loan], [Principal], [Debt], etc. can be registered and managed.

As above, the quantified number of negative cases and negative rate are calculated monthly, and the average monthly negative rate is calculated by weighting the number of positive and negative cases over the past N years. In addition, macroeconomic data for each industry is obtained and quantified to calculate the industrial bankruptcy probability of the relevant company.

division Macroeconomic source information shipbuilding shipping Korea Shipowners Association shipbuilding shipping Shanghai Shipping Exchange steel Korea Steel Association manufacturing Korea Automobile Industry Association service e-country indicators common trade association common economic statistics system common Economist Intelligence Unit (EIU) common Statistical Office financial information Securities company report financial information Naver Financial Index common Organization for Economic Cooperation (OECD) country information International Monetary Fund (IMF) etc ISM (Institute for Supply Management) financial information KIND (Korea Exchange public information)

Therefore, corporate bankruptcy prediction is not limited to financial information in the traditional financial information-based model represented by multivariate discriminant analysis and logit model, but also studies the excellence of corporate bankruptcy prediction based on market information and applies various methodologies. It can be constructed in a way that increases prediction performance. In this way, a new default prediction model that is different from the existing default prediction model can be implemented by calculating the probability of text insolvency using the indicators predicted based on financial information, the indicator values calculated from news data, and industry-specific macroeconomic indicators. Investment monitoring Unit 380 calculates the financial analysis model score based on the company's credit rating information and financial outlier analysis information, and calculates the score through the insolvency prediction model using big data and unstructured data corresponding to news. , based on a combined score that combines the financial analysis model score and the insolvency prediction model score, it is possible to provide results for the final investment decision considering the potential insolvency occurrence index. At the user's request, credit rating information, financial outlier information, and insolvency prediction information are obtained from external or internal systems to score the company's credit rating and financial outlier rating, and insolvency predictions obtained from a insolvency prediction system using big data. By scoring the calculated results, data can be constructed to be viewed in the monitoring system, and the system can provide investment decision results for the final investment decision.

The financial outlier rating is converted into scores for each financial outlier calculation item, and the combined score for all items is calculated as the financial outlier rating score. In addition, financial outlier rating history information can be provided for reference in analyzing investment opinions, and the financial outlier ratings (6-/6/6+) assigned by each bank can be added up for each financial outlier analysis item to form a score. For the failure prediction, the score calculated by converting the failure prediction score calculated from the failure prediction model using the big data described above into a percentile is calculated as the final failure prediction model score. The financial analysis model score and the final insolvency prediction model score are calculated by adding the combined analysis score by adding 60% weight to the non-analysis model score and 40% to the insolvency prediction model score. And depending on the combined analysis score, companies can be divided into companies with high insolvency prediction, medium insolvency prediction, and low insolvency prediction.

The final result selected as the basis for selecting the final investment opinion result can be provided to the user based on the classification results of companies predicted to be insolvent calculated by the combined analysis score and the classification result calculated by the potential insolvency occurrence index. The potential insolvency index is information that determines the potential insolvency of corporate loans at the time of analysis, and is an indicator that determines potential insolvency factors by combining a company's financial information and news information. The financial analysis model score is subtracted from the insolvency prediction model score, and the result divided by the combined score of the insolvency prediction model score and the financial analysis model score is calculated as the potential insolvency occurrence index. If the calculated result is positive, the insolvency prediction model score is higher than the financial analysis model score, which can be leveled to strong or medium insolvency prediction depending on the preset level. If the calculated result is negative, the financial analysis model score is higher than the insolvency prediction model score. If it is higher than the score, it can be leveled by dividing it into high, medium, and low financial analysis.

The damage prediction unit 390 uses big data analysis that has a correlation between similar investment advisory services and consumer damage to preemptively predict consumer damage to similar investment advisory services, KOSPI, and stock market risk level. , it is possible to learn and verify the correlation between the search volume of search terms related to similar investment advisory services, the number of consumer consultations and applications for consumer damage relief for similar investment advisory services. Information on damages caused by these quasi-investment advisory services can be provided and information can be provided to prevent users from being harmed by quasi-investment advisory services by joining the KakaoTalk reading room.

The market impact model unit 391 is a market impact model that explains the extent to which the price rises when buying or the extent to which the price falls when selling, for ultra-short-term investment corresponding to HFT (High-Frequency Trading). You can use to predict transaction costs and evaluate transaction efficiency. The Capital Markets Act allows the establishment of Alternative Trading Systems (ATS) to reduce market shock costs, but despite this reality, there is no predictive model that considers the status of market shock costs. If investors buy at a higher price or sell at a lower price than they want, investors incur additional transaction costs. In addition, buying at a high price increases the current market price, or selling at a low price causes the market price to fall. Such a situation is highly likely to occur in large-scale transactions by institutional investors and foreign investors with large operating assets. The transaction costs that investors must bear when trading stocks can be divided into nominal transaction costs and implicit transaction costs. Nominal transaction costs include transaction taxes and transaction fees, while implicit transaction costs include spreads between buy and sell prices, opportunity costs, and market impact costs. Market shock cost is an implicit transaction cost that investors must bear additionally due to market shock caused by buying at a higher price and selling at a lower price than the current transaction price.

In terms of market microstructure, market impact cost is very important because it is also an important factor when deciding to invest in a specific stock. In general, investment decisions are made based on expectations about the future performance of the relevant stock. However, if the market shock cost is large when purchasing a stock expected to have a high expected return, a lower-than-expected rate of return may be realized, and if the market shock cost of purchasing a stock expected to have a low expected rate of return is small, a return higher than expected may be realized. Therefore, if you can estimate the market impact cost of the stock of interest, you can improve investment performance and get closer to composing an optimal portfolio. Accordingly, it is possible to compare and analyze investors' market shock costs through real-time quotes and sales in the Korean stock market. The order book contains information such as the time when investors submitted quotations, quotation type and conditions, quantity, price, order medium, encrypted account number, investor classification, and whether or not they are foreign investors, and the execution book ( Trade Book) contains various data showing investment decisions, including the market price corresponding to the transaction history between the bid price and the ask price, the execution price, and the quote serial number that distinguishes the leading order, divided down to the millisecond. there is.

The market impact cost can be calculated and analyzed for all listed items in the Korean stock market. If the market impact cost of investors has a negative correlation with the price, market capitalization, and liquidity of the traded item, the corresponding It can also predict transaction costs and efficiency to provide guidance for high-frequency trading. Individual investors are referred to as irrational or unsystematic investors who are suffering losses on average, and over-confidence, confirmation bias, and disposition effect are pointed out as the causes of this result. In other words, if excessive confidence and aggressive ordering patterns are increasing market shock costs and, as a result, are having a negative impact on investment performance, information on such high-frequency buying and selling can be provided in advance to provide guidance to increase investment efficiency.

The performance analysis unit 393 may monitor information on stocks with rising potential after the market starts and, if an error occurs, use the error pattern and error data to retrain at least one artificial intelligence algorithm. Because performance analysis is possible every day, if this data continues to be accumulated and re-learned, prediction accuracy can increase, and error patterns can be analyzed to re-train the algorithm to prevent errors from appearing.

Hereinafter, the operation process according to the configuration of the recommendation service providing server of FIG. 2 described above will be described in detail using FIGS. 3 and 4 as an example. However, it will be apparent that the embodiment is only one of various embodiments of the present invention and is not limited thereto.

Referring to Figure 3, (a) the recommendation service providing server 300 builds big data on domestic listed items, (b) models by learning and verifying artificial intelligence algorithms, and (c) before the market starts. It recommends stocks that are likely to rise, and (d) collects information on whether the stocks that are likely to rise that day are really likely to rise, and uses data for re-learning to train an artificial intelligence algorithm through pre-processing and normalization. When measuring prediction accuracy, as shown in (a) of Figure 4, the difference between the predicted value and the actual value is analyzed, and if an error pattern appears as shown in (b), relearning is performed to remove it, and (c) real-time issues are analyzed. By monitoring, information necessary for selling or purchasing can be provided to the user terminal 100 as a notification.

Matters that are not explained regarding the method of providing a short-term trading item recommendation service using big data-based artificial intelligence in Figures 2 to 4 are explained in Figure 1 regarding the method of providing a short-term trading item recommendation service using big data-based artificial intelligence. Since the content is the same as the content described or can be easily inferred from the content described, the description below will be omitted.

FIG. 5 is a diagram illustrating a process in which data is transmitted and received between components included in the system for providing a short-term trading item recommendation service using big data-based artificial intelligence shown in FIG. 1 according to an embodiment of the present invention. Hereinafter, an example of the process of transmitting and receiving data between each component will be described with reference to FIG. 5, but the present application is not limited to this embodiment, and the process shown in FIG. 5 according to the various embodiments described above It is obvious to those skilled in the art that the process of transmitting and receiving data can be changed.

Referring to Figure 5, the recommendation service providing server builds big data by accumulating data including stock price information, stock index, and real-time issues for domestic listed items (S5100).

In addition, the recommendation service providing server analyzes big data with at least one artificial intelligence algorithm for short-term trading (S5200), and is a pre-built server to minimize the margin of error of domestic listed stocks extracted from at least one artificial intelligence algorithm. Filtering is performed using a filtering algorithm (S5300).

In addition, the recommendation service providing server recommends domestic listed stocks, excluding domestic listed stocks extracted from at least one artificial intelligence algorithm and filtered by a filtering algorithm, as the likely rising stocks of the day before the market starts (S5400).

The sequence between the above-described steps (S5100 to S5400) is only an example and is not limited thereto. That is, the order between the above-described steps (S5100 to S5400) may change, and some of the steps may be executed simultaneously or deleted.

Matters that are not explained regarding the method of providing a recommendation service for short-term trading items using big data-based artificial intelligence in Figure 5 are explained in Figures 1 to 4 regarding the method of providing a recommendation service for short-term trading items using big data-based artificial intelligence. Since the content is the same as the content described or can be easily inferred from the content described, the description below will be omitted.

The method of providing a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment described in FIG. 5 is a method of providing a short-term trading item recommendation service using a recording medium containing instructions executable by a computer, such as an application or program module executed by a computer. It can also be implemented in form. Computer-readable media can be any available media that can be accessed by a computer and includes both volatile and non-volatile media, removable and non-removable media. Additionally, computer-readable media may include all computer storage media. Computer storage media includes both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data.

The method of providing a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment of the present invention described above includes applications installed by default on a terminal (this may include programs included in the platform or operating system, etc., which are basically installed on the terminal). It may be executed by an application (i.e., a program) that the user installs directly on the master terminal through an application providing server such as an application store server, an application, or a web server related to the service. In this sense, the method of providing a short-term trading item recommendation service using big data-based artificial intelligence according to an embodiment of the present invention described above is implemented as an application (i.e., program) installed by default on the terminal or directly installed by the user, and is implemented in the terminal. It may be recorded on a computer-readable recording medium such as .

The description of the present invention described above is for illustrative purposes, and those skilled in the art will understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. will be. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. For example, each component described as unitary may be implemented in a distributed manner, and similarly, components described as distributed may also be implemented in a combined form.

The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

Claims (10)

A user terminal that receives recommendations for stocks with rising potential among domestically listed stocks before the market begins; and
A big data storage unit that builds big data by accumulating data including stock price information, stock indexes, and real-time issues for domestic listed items, an analysis unit that analyzes the big data with at least one artificial intelligence algorithm for short-term trading, and the at least A filtering unit that performs filtering using a pre-established filtering algorithm to minimize the error range of domestic listed items extracted from one artificial intelligence algorithm, and filtered by the filtering algorithm after being extracted from the at least one artificial intelligence algorithm. A short-term recommendation department that recommends domestically-listed stocks, excluding domestically-listed stocks, as likely to rise on the day before the market begins, and uses big data analysis that shows the correlation between similar investment advisory services and consumer damage to consumers for similar investment advisory services. In order to preemptively predict the A recommendation service providing server including a unit;
A system that provides short-term trading item recommendation service using big data-based artificial intelligence, including.
According to claim 1,
The real-time issues include international issues, political issues, and financial issues,
a server providing the recommendation service;
A real-time issue notification unit that recommends the stocks with rising potential and predicts stock price changes according to the real-time issues when the real-time issues change after the market starts, and informs the user terminal of the predicted stock prices;
A system for providing short-term trading item recommendation service using big data-based artificial intelligence, further comprising:
According to claim 2,
The recommendation service providing server is,
In order to interpret the real-time issue, a natural language understanding unit that collects text about the real-time issue and then analyzes it using Natural Language Understanding (NLU);
A system for providing short-term trading item recommendation service using big data-based artificial intelligence, further comprising:
According to claim 3,
The recommendation service providing server is,
Through the natural language understanding, the emotional score of each investor that can affect the stock price increase or decrease is calculated, and the correlation between the emotional score and the stock price increase or decrease is analyzed to determine whether the stock price exceeds the preset standard value. If a correlation exists, a correlation analysis unit that transmits information on a stock price increase or stock price drop for which the correlation exists to the user terminal;
A system for providing short-term trading item recommendation service using big data-based artificial intelligence, further comprising:
According to claim 1,
The artificial intelligence algorithm is,
A system that provides short-term trading item recommendation service using big data-based artificial intelligence, characterized by LSTM (Long Short-Term Memory).
According to claim 1,
The recommendation service providing server is,
An investment monitoring department that monitors investment information using a pre-established insolvency prediction model and credit rating calculation model to extract insolvent companies among the domestic listed stocks;
A system for providing short-term trading item recommendation service using big data-based artificial intelligence, which further includes:
According to claim 6,
The investment monitoring department said,
The financial analysis model score is calculated based on the company's credit rating information and financial outlier analysis information, and the unstructured data corresponding to the news is used to calculate the score through the insolvency prediction model using the big data, and the financial analysis model A system that provides short-term trading item recommendation service using big data-based artificial intelligence, which provides results for final investment decisions considering the potential insolvency occurrence index based on a combined score that combines the score and the insolvency prediction model score.
delete According to claim 1,
The recommendation service providing server is,
For ultra-short-term investment in response to High-Frequency Trading (HFT), transaction costs are predicted using the Market Impact Model, which explains the extent to which prices rise when buying or the extent to which prices fall when selling. Market Impact Model Department, which evaluates transaction efficiency;
A system for providing short-term trading item recommendation service using big data-based artificial intelligence, further comprising:
According to claim 1,
The recommendation service providing server is,
A performance analysis unit that monitors information on the rising stock after the market starts and uses error patterns and error data to retrain the at least one artificial intelligence algorithm when an error occurs;
A system for providing short-term trading item recommendation service using big data-based artificial intelligence, further comprising:

Images