KR100668417B1 - Method for detecting abnormal stock trade using A.I. - Google Patents

Abstract

The present invention relates to a method for detecting stock abnormal trade using artificial intelligence techniques to detect stock price manipulation to perform stock unfair trade.

According to the present invention, there is provided a method for detecting abnormal trading of stocks, comprising: a first step of a computer collecting stock price data of past stock manipulation items and generating an abnormal trading pattern using a heuristic model which is a field of artificial intelligence; A second step of receiving, by a computer, stock price data of a target item for which a stock price operation is to be detected, and determining a target period and a reference date; A third step of allowing the computer to adjust the stock price at the target date and the reference date of the detected target item and the abnormal trading pattern to enable one-to-one comparison of the stock price by date; A fourth step of the computer counting the number of matching days within an allowable range by comparing the stock price of the detected item and the stock price of the abnormal trading pattern for each corresponding date with respect to the detected period; And a fifth step in which the computer compares the number of days with a threshold to determine whether the detected target item is abnormally sold.

Stock, abnormal trading, stock manipulation, detection, artificial intelligence, heuristic model

Description

Method for detecting abnormal stock trade using A.I.}             

1 is a flowchart illustrating a method of detecting an abnormal trading of stocks according to an embodiment of the present invention;

2 is an operation flowchart illustrating an initial generation step of an abnormal traffic pattern of the present invention;

FIG. 3 is a flowchart illustrating an operation of adjusting the stock price at a target date and a reference date of the target stock and abnormal trading pattern according to the present invention.

The present invention relates to a method for detecting abnormal trading of stocks, and more particularly, to a method for detecting abnormal trading using artificial intelligence to detect a stock price manipulation for performing a stock unfair trade.

Stock manipulation is taking place in the securities markets of any country in the world. Stock manipulation is common even in the United States, where the world's most stringent securities market is regulated. Stock manipulation is not easily caught because it is buried in numerous securities transactions, and it is expected that stock manipulation will not go away as long as the stock market exists because success can bring huge profits in a moment.

However, the transparency and fairness of stock trading in the capitalist market is very important. If securities trading is unfair and the securities market is artificially influenced, the market economic order may collapse, losing investors' confidence in the market, which may put the social and national real economies of share price on supply. In addition, from the external side, foreign investors do not trust the domestic stock market, making it difficult to operate the national economy, such as reducing foreign investment and leaving foreign funds.

In order to prevent the damages of stock unfair trade such as the manipulation of stocks on the securities market, the state agency conducts stock trading supervision.

The stock trading process consists of a stock price monitoring process, a follow-up investigation process, a trading psychology process, and an action process. The stock price monitoring process is a process of finding stocks that are likely to be unfair trades in which fluctuations in stock prices and volume are out of the normal range. The follow-up process is a process of investigating and analyzing the probability of unfair trading found in the stock price monitoring process by the member, branch, and account. The trading psychology process is the process of setting up suspected consignee groups through the consignment linkage search for the items to be investigated further in the follow-up survey, and analyzing the quotations, transaction volume, and customs tariff rate of each group. The process of action is the process of disciplinary action or notification of the Financial Supervisory Commission according to the results of the trading hearing.

As a conventional method of detecting an abnormal trade in the stock price monitoring process, there is a method of detecting an abnormal trade when the fluctuation of the stock price is out of the normal range by comparing the past stock price movement with the current stock price movement of the corresponding item. In other words, the slope and intercept are calculated by regression analysis of stock index fluctuations and expected market returns over a certain period. The upper and lower limit criteria for abnormal trading are set by reflecting a certain weight in this figure. The allowable range is determined by subtracting the expected price fluctuation rate from the actual market price fluctuation rate, and if the stock price of the corresponding item is out of the allowable range, it is detected as an abnormal trade.

The conventional abnormal trade detection method detects an abnormal trade using one extraction criterion parameter calculated from past stock price movements. There is a problem that it is very difficult to accurately detect.

An object of the present invention devised to solve the above-mentioned problems of the prior art is to generate an abnormal trading pattern from the abnormal trading data in the past stock manipulation, and compare the abnormal trading pattern with the current stock price movement pattern. It is to provide a method for detecting marketing.

In accordance with an aspect of the present invention, there is provided a method for detecting abnormal trading of stocks, comprising: a first step of a computer generating an abnormal trading pattern using stock price data of past stocks; A second step of receiving, by a computer, stock price data of a target item for which a stock price operation is to be detected, and determining a target period and a reference date; A third step of allowing the computer to adjust the stock price at the target date and the reference date of the detected target item and the abnormal trading pattern to enable one-to-one comparison of the stock price by date; A fourth step of the computer counting the number of matching days within an allowable range by comparing the stock price of the detected item and the stock price of the abnormal trading pattern for each corresponding date with respect to the detected period; And a fifth step in which the computer compares the number of days with a threshold to determine whether the detected target item is abnormally sold.

According to the present invention, there is provided a computer-readable recording medium having recorded thereon a program for executing a method for detecting abnormal trafficking of stocks in a computer.

Hereinafter, a method for detecting abnormal sales of stocks according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

The present invention is a computer related invention, and typically operates in a computer having an input / output device, a memory device, and a central processing unit. Input devices include direct input devices including keyboards and mice, and communication devices for inputting necessary data from a remote external system. The memory device stores a program for operating the central processing unit and temporary data according to program driving while the central processing unit executes the method for detecting abnormal trafficking of stocks according to the present invention.

In the abnormal trading method of the stock according to the present invention, first, an abnormal trading pattern is generated by using stock price data of stocks manipulated in the past (S101), and the generated abnormal trading pattern is stored in a database. The storage format of the abnormal trading pattern is composed of the number of the abnormal trading pattern, the period, the day, and the stock price of the corresponding day as shown in Table 1 below. The abnormal traffic pattern initial generation process will be described later in detail with reference to the flowchart of FIG. 2.

Abnormal traffic pattern number term X value (days) Y value (share price) One 60 One 800 One 60 2 1000 One 60 3 1200 One 60 4 1350 One ... ... ... One 60 60 980

The stock price data of the current stock manipulation detection target item is input (S102), and the D-day moving average value is calculated for the stock price of the detected target item (S103). The reason for calculating the moving average is to reduce the fluctuation of the stock price by alleviating the fluctuation of the stock price, and the D value is obtained for each item through the training and validation process. The equation for calculating the D-day moving average value is shown in Equation 1 below.

In Equation 1, j represents an item and C _jt is an item price in which item j is formed on date t. t means a date in sequential meaning, taking into account only stock trading dates. The D-day moving average is an indicator that continuously displays the average of the stock prices for the past D days, and uses the simple averaging method. In the present invention, a three-day moving average value is calculated. If the closing price of today is 10,000 won, the closing price of 9,000 won and the closing price of the previous day is 8,000 won, then the moving average of three days is (10,000 + 9,000 + 8,000) / 3 = 9,000 won. Becomes

Next, the detection target period and the reference date are determined (S104), and the stock price of the detection target item and the abnormal trading pattern and the target price of the detection target period and the reference date are the same so that the stock price of the detection target item and the abnormal trading pattern can be compared one-to-one. In conversion (S105). The stock price adjustment step at the target date and the target date of the detected item and the abnormal trading pattern will be described in detail later with reference to the flowchart of FIG. 3.

Next, the coincidence counter and the comparison counter are initialized (S106), and the stock price of the item to be detected and the stock price of the abnormal trading pattern are compared one-to-one on a daily basis (S107). If the two stocks coincide within the allowable range (S108), the coincident counter is increased by 1 (S109), and the comparison counter is increased by 1 (S110). If the two shares do not match in step S108, only the comparison counter is increased by one without increasing the coincidence counter (S110). If the comparison counter is not larger than the detection target period (S111), the process returns to step S107. If the comparison counter is larger than the detection target period (S111), the threshold value is compared with (matching counter / detection target period) (S112). In this embodiment, 0.95 is used as the threshold.

If (matching counter / detection target period) is greater than or equal to the threshold value 0.95 in step S112, it is determined that there is a possibility of the stock price manipulation (S113). If the item to be detected is likely to be manipulated, the investor may conduct a intensive analysis of the amount of trading or involvement during the relevant period to determine whether it is an abnormal trading. The abnormal trading pattern is further learned by using the pattern of the detected target item (S114). If (matching counter / detection period) is smaller than the threshold value 0.95 in step S112, it is determined that the corresponding detection target item has no possibility of stock price manipulation (S115).

The stocks that are determined to have the possibility of stock price manipulation in step S113 are included in the follow-up process for investigating and analyzing the involvement ratios of the quotations and transaction volume by member, branch, and account, and the stocks to be further investigated in the follow-up survey. It is judged whether or not an abnormal sale is made through the sales psychology process that analyzes the group of suspected consignees, the quotations of each group, the volume of transactions and the rate of customs tariff.

2 is a flowchart illustrating an operation of generating an abnormal traffic pattern initial step (S101) according to an exemplary embodiment of the present invention.

First, by collecting the stock price data of the stock items manipulated in the past (S211), the D-day moving average value of the stock prices of the collected pattern generation target items is calculated (S212). The reason for calculating the moving average for the stock price is to reduce the fluctuation of the stock price by easing the fluctuation of the stock price. Equation for calculating the D-day moving average value of the stock price of the target stock items is the same as Equation 1 above. In the present invention, a three-day moving average value is calculated.

Next, the heuristic model of the stock price manipulation item is generated by calculating the α% truncated average for each day (S213). α% truncated average is calculated as in Equation 2 below.

Here, n is the number of events, M _ji is the moving average value of the week price of day j of the event j, k is the value of n multiplied by α and divided by 100 (k = n * α / 100).

Explaining the process of calculating the α% truncated average, the i-day week of all n items sorts the moving average in ascending order and cuts the numerical area as low as α% and the numerical area as high as α%. .

For example, if the week on day 12 of 12 events is sorted in ascending order, the 10% truncated mean is calculated at 2.28, 3.96, 5.60, 9.51, 10.55, 11.25, 17.44, 20.62, 23.20, 23.95, 31.43. Explain. In this example n = 12 and k = n * α / 100 = 12 * 10/100 = 1.2. Therefore, the share price is cut by 1.2 in both the upper and lower regions of the moving average value, and the highest and lowest values are cut one by one, and then the highest and lowest values are multiplied by 1-0.2. In the above example, the 10% truncated average is [(0.8) * 3.96 + 5.60 + 9.51 + 10.55 + 11.25 + 17.44 + 20.62 + 23.20 + 23.95 * (0.8)] / (12-2 * 1.2) = 13.6570833.

As described above for the entire stock price manipulation period, the aberration trading pattern is obtained by obtaining the α% truncated average for all stocks, and as shown in Table 1, the α% truncated average value is stored for each day (S214).

3 is an operation flowchart showing a stock price adjustment step (S105) in a detection target period and a reference date of the detection target item and the abnormal traffic pattern of the present invention.

First, the stock prices of the stock to be detected are proportionally converted to be the same as the stock price of the start date of the abnormal trading pattern (S241). For example, if the stock price of the start date of the abnormal trading pattern is 100 won and the stock price of the target day of the detected item is 10 won, the stock price of the other day of the detected item is adjusted while adjusting the stock price of the detected item 10 won to 100 won. Adjust up by 10 times together.

Then, the least common multiple of the detection subject period and the period of the abnormal traffic pattern is calculated (S242). If the abnormal trading pattern is 60 days and the detection target period is 90 days, the least common multiple is 180 days. When the minimum common multiple is obtained, the period of the detection target period and the abnormal traffic pattern is extended so that the period of the detection target period and the abnormal traffic pattern is the minimum common multiple (S243).

Then, the period of abnormal trading pattern should be tripled, and the period to be detected should be doubled. If you triple the period, insert two median median values between two adjacent days, and double the period. Is obtained by inserting the median value of one stock price between two adjacent days. For example, when (1,800), (2, 1000) ..., if you need to triple the period, replace (1,800), (2,1000) with (1,800), (4,1000) Insert the median stock price of (2,867) and (3,933).

Although the technical spirit of the present invention has been described above with reference to the accompanying drawings, it is intended to exemplarily describe the best embodiment of the present invention, but not to limit the present invention. In addition, it is obvious that any person skilled in the art may make various modifications and imitations without departing from the scope of the technical idea of the present invention.

As described above, according to the present invention, since the possibility of stock price manipulation (abnormal sales) can be determined by generating an abnormal trading pattern from past abnormal trading data and comparing the current stock price flow for a certain period with the abnormal trading pattern, Groups can be easily and accurately detected and punished, and finally, the transparency and fairness of stock trading can be restored.

Claims (6)

A first step in which a computer receives stock price data of past stock price-manipulated items, calculates a?% Truncated average for each day, and generates an abnormal trading pattern using a heuristic model; A second step of receiving, by a computer, stock price data of a target item for which a stock price operation is to be detected, and determining a target period and a reference date; A third step of allowing the computer to adjust the stock price at the target date and the reference date of the detected target item and the abnormal trading pattern to enable one-to-one comparison of the stock price by date; A fourth step of the computer counting the number of matching days within an allowable range by comparing the stock price of the detected item and the stock price of the abnormal trading pattern for each corresponding date with respect to the detected period; And a fifth step in which the computer compares the number of days with a threshold value and determines whether the detected target item is abnormally traded. delete The method of claim 1, The first and fifth steps are performed by using the moving average value of D (D is a natural number greater than 2) of the stock price data for generating the abnormal trading pattern and the stock price data of the detected target item. Abnormal trafficking detection method. The method of claim 1, wherein the third step, Proportionally adjusting the stock price of the detected target item so that the stock price of the detected target item is the same as the stock price of the start date of the abnormal trading pattern; Obtaining a least common multiple of the detection subject period and the abnormal trading pattern period; And inserting a median stock price value between two adjacent days such that the detection target period and the period of the abnormal trading pattern are the least common multiple. The method of claim 1, further comprising the step of further learning the abnormal trading pattern by applying the stock price pattern of the detecting target item to the abnormal trading pattern when the detected target item is determined to be a stock price manipulation. Detecting abnormal stock trading. A computer-readable recording medium having recorded thereon a program for executing a stock abnormality trading method on a computer, The stock error trading method, A first step of the computer receiving the stock price data of the stock price-manipulated items of the past and calculating an α% truncated average for each day and generating an abnormal trading pattern using a heuristic model; A second step of receiving, by the computer, stock price data of a target item for which a stock price operation is to be detected and determining a target period and a reference date; A third step of allowing the computer to adjust the stock price at the target date and the reference date of the detected target item and the abnormal trading pattern to enable one-to-one comparison of the stock price by date; A fourth step in which the computer compares the stock price of the detected item and the stock price of the abnormal trading pattern one-to-one for each corresponding date with respect to the detected object period and counts the number of matching days within an allowable range; And a fifth step in which the computer compares the number of days with a threshold to determine whether the detected target item has been abnormally traded.

Images