KR20090123352A - Method of predicting stocks price and sytem using the method - Google Patents

Abstract

PURPOSE: A method for predicting a trend of a stock market in the future based on past information about the stock market and stock market price predicting system using the same are provided to generate purchasing/selling signal. CONSTITUTION: An interval corresponding to a predetermined pattern type is extracted based on past market price information according to the time of a stock market(S110). Conditions corresponding to the type of patterns is set up based on the past quotation information(S150). A type presently related to the quotation of the stock market among the predetermined pattern types is selected by using a DTW(Dynamic Time Warping) algorithm(S170). The purchasing signal, selling signal or a holding signal are generated by using the set condition(S180).

Description

Stock quote forecasting method and stock quote forecasting system using the same {METHOD OF PREDICTING STOCKS PRICE AND SYTEM USING THE METHOD}

The present invention relates to techniques for predicting future stock market trends based on historical information about stock market tickers.

There have been many studies to predict the trend of the stock market. Since the trend of the stock market depends on many variables, there are clearly limitations in accurately predicting the trend of the stock market.

In order to predict future stock market trends, it is necessary to properly analyze past stock market trends. And there are various technical indicators that express the trend of the stock market. For example, many people used the various technical indicators as well as the technical indicators listed in Table 1 to analyze trends in the stock market and to predict future stock market trends.

A lot of studies are required to properly extract the technical indicators of the past stock market and to accurately predict the trend of the future stock market through the extracted technical indicators.

The stock quote prediction method and system according to an embodiment of the present invention accurately determine the type related to the current price of the stock market using the DTW algorithm, and appropriately generate a buy / sell signal corresponding to the determined type, thereby providing future Accurately predict stock quotes

In addition, the stock quote prediction method and the system according to an embodiment of the present invention sets the conditions for buying or selling from the past technical indicators, and actually performs the trade using the set conditions for buying or selling.

In addition, the stock quote prediction method and the system according to an embodiment of the present invention to optimize the conditions for buying or selling by calling a virtual experiment, and the conditions with high accuracy based on the results obtained by actually trading Maximize return by updating to.

The stock quote prediction method according to an embodiment of the present invention comprises extracting time sections corresponding to predetermined pattern types based on past quote information over time of a stock market, and technical indicators in the extracted time sections. Setting conditions corresponding to the predetermined pattern types based on the data and the past price information, a type related to the current price of the stock market among the predetermined pattern types using a dynamic time warping (DTW) algorithm. And selecting one of a buy signal, a sell signal, and a hold signal using the conditions set for the selected type.

In addition, the stock quote prediction method according to another embodiment of the present invention comprises the steps of extracting time intervals corresponding to the predetermined pattern types on the basis of the historical price information over time of the stock market, in the extracted time intervals Setting first conditions corresponding to the predetermined pattern types based on technical indicators and the historical quote information, calculating a rate of return of the first conditions through a virtual experiment, and returning the calculated first conditions Generating second conditions in which the first conditions are optimized, selecting a type related to the current price of the stock market among the predetermined pattern types using a dynamic time warping (DTW) algorithm, and the selected Generates either a buy signal, sell signal, or hold signal using a second condition set for the type It includes the steps:

In addition, the stock quote prediction system according to another embodiment of the present invention is a historical information analysis unit for extracting the time intervals corresponding to the predetermined pattern types based on the past quote information over time of the stock market, the extracted time interval A setting unit for setting first conditions corresponding to the predetermined pattern types based on the technical indicators and the past quote information in the first and second terms, and calculating a rate of return of the first conditions through a virtual experiment, Selecting a type related to the current market price of the stock market among the predetermined pattern types by using a dynamic time warping (DTW) algorithm, a database storing the second conditions in which the first conditions are optimized according to the rate of return of the 1 conditions. Buy signal, sell signal, or hold signal using a selector and a second condition set for the selected type It includes a signal generator for generating any one of the signals.

The stock quote prediction method and system according to an embodiment of the present invention accurately determine the type related to the current price of the stock market using the DTW algorithm, and appropriately generate a buy / sell signal corresponding to the determined type, thereby providing future You can accurately predict stock prices in.

In addition, the stock quote prediction method and the system according to an embodiment of the present invention can set the conditions for buying or selling from the past technical indicators, and can actually carry out the trading using the set conditions for buying or selling. have.

In addition, the stock quote prediction method and the system according to an embodiment of the present invention to optimize the conditions for buying or selling by calling a virtual experiment, and the conditions with high accuracy based on the results obtained by actually trading You can maximize your return by updating to.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a stock quote prediction method according to an embodiment of the present invention.

2 is a table showing predetermined pattern types and corresponding time intervals according to an embodiment of the present invention.

FIG. 3 is a diagram illustrating historical quote information and technical indicators derived from past quote information according to an embodiment of the present invention.

4 is a diagram for explaining a DTW algorithm.

Referring to FIG. 1, the stock quote prediction method according to an embodiment of the present invention analyzes past quote information over time of a stock market and detects time sections corresponding to predetermined pattern types (S110).

Here, the historical price information may be related to the Korea Composite Stock Price Index (KOSPI) or the KOSPI 200 index, and may include information on the market price, the closing price, or the trading volume of the KOSPI 200 KOSPI index. That is, as will be described below, the stock price prediction method according to an embodiment of the present invention analyzes a past pattern of the KOSPI index or the KOSPI 200 index and analyzes the trend of the future KOSPI index or the KOSPI 200 based on the analyzed past pattern. You can predict accurately.

The table shown in FIG. 2 describes a plurality of pattern types. Pattern types include types such as short-term upswing, short-term downturn, bullish rise, bullish decline, ceiling sideways and bottom sideways. That is, these pattern types may be classified according to the pattern of the KOSPI index or the KOSPI 200 index. For example, stock experts can classify different pattern types from past KOSPI or KOSPI 200 indices using various applications or using their knowledge.

Each of the plurality of classified pattern types has a predetermined time interval. That is, referring to Figure 2, it can be seen that the starting point of the 'short-term rise' type is December 26, 1997, the last point is March 05, 1998. Similarly, short-term declines, bullish trends, bullish declines, ceiling sideways and bottom sideways also have their starting and ending points.

In conclusion, the stock quote prediction method according to an embodiment of the present invention may extract time intervals corresponding to a plurality of predetermined pattern types based on past quote information of the stock market through step S110.

In addition, the stock quote prediction method according to an embodiment of the present invention extracts the technical indicators in the extracted time intervals (S120).

That is, various technical indicators exist in the extracted time intervals. These technical indicators reflect the pattern of the stock market and can be used to predict future stock market trends.

Here, the technical indicators may include not only the indicators shown in Table 1, but also various indicators, and these technical indicators are generally used in technical analysis techniques for stock price prediction, and thus detailed description thereof will be omitted.

Referring to FIG. 3, past price information of a time interval corresponding to one pattern type is illustrated, and technical indicators extracted from the past price information are illustrated.

As a result, through step S120, technical indicators in time periods corresponding to the plurality of pattern types may be extracted based on past quote information. For example, the pattern type in the category of 'ceiling confinement' lasts from July 12, 1999 to February 03, 2000, and the time period from July 12, 1999 to February 03, 2000. Technical indicators in can be extracted through step S120.

In addition, the stock quote prediction method according to an embodiment of the present invention discretizes the extracted technical indicators (S130).

The purpose of step S130 is to cancel out the meaningless attributes between the technical indicators belonging to the same time interval, and to use only meaningful key indicators that well express the trend among the technical indicators belonging to the same time interval.

There are various techniques for discretization techniques. Representatively, there are various discretization techniques such as discernibility preservation, entropy minimization, equal frequency binning and various naive methods, and manual technique. Various discretization techniques are already well known, and thus detailed descriptions are omitted.

In addition, the stock quote prediction method according to an embodiment of the present invention filters discretized technical indicators by using a data mining technique (S140).

Discrete technical indicators may also be data that is not simplified or simplified. At this time, the stock quote prediction method according to an embodiment of the present invention may reduct discretized technical indicators using various reductance methods.

Here, the reduct methods may include genetic algorithm, manual reducer, dynamic reduct, approximate hitting set approaches.

In operation S150, first conditions corresponding to predetermined pattern types are set based on the technical indicators and the historical price information in the extracted time intervals, according to an exemplary embodiment.

First conditions for buying or selling may be set for each of the predetermined pattern types. Here, the first conditions may be configured as a function including the technical indicators in the extracted time intervals as variables. In this case, the first conditions may be a function of an if ~ then ~ form. These first conditions may later be used as trading rules for actual stock trading.

For example, suppose the 'short rise' pattern type occurred in the x time interval in the past. The technical indicators of the time interval x are a, b, c, d, e, and g, and the technical indicators a, b, c, d, e when the stock price rises and the stock price falls on the x time interval, respectively. , can be expressed as a function of g. At this time, the first conditions are technical indicators a, b, c to generate a buy signal before the rise in the stock price and a sell signal before the fall in the stock price based on the movement in the x time interval. It can be set as a function of, d, e, g.

As a result, the stock quote prediction method according to an embodiment of the present invention may generate a plurality of trading rules (ie, first conditions) for each of the plurality of pattern types based on past quote information. In the future, when a pattern similar to any one of the plurality of pattern types occurs, a signal for buying or selling may be generated using the plurality of trading rules.

In addition, the stock quote prediction method according to an embodiment of the present invention calculates the returns of the first conditions through a virtual experiment, the second condition is optimized the first conditions in accordance with the calculated return of the first conditions Generate them (S160).

That is, the stock quote prediction method according to an embodiment of the present invention executes a virtual trade in accordance with the buy signal, sell signal or holding signal generated by the generated first conditions, and the actual trading result according to the first conditions. Can be optimized to optimize the first conditions.

For example, the virtual condition may be determined as the second conditions by extracting only the conditions having a yield of 10% or more among the first conditions.

In addition, the stock quote prediction method according to an embodiment of the present invention selects a type related to the current quote of the stock market from among the predetermined pattern types by using a dynamic time warping (DTW) algorithm (S170).

That is, there may be a pattern type having a pattern most similar to the current trend of the stock market among the predetermined pattern types. At this time, according to an embodiment of the present invention, a pattern type most similar to the current trend may be selected from among a plurality of predetermined pattern types by using the DTW algorithm.

Dynamic Time Warping (DTW) algorithm is an algorithm that compares two columns to determine similarity between two sequences. According to the DTW algorithm, the distance between all components of the two columns is set as cost. Then, on a two-column grid, two columns are compared, searching for a mapping function according to the dynamic programming method such that the cost from the starting component to the last component of each column is minimized.

At this time, the final cost in the last component can be interpreted as the similarity between the two rows. 4 conceptually illustrates a process of determining the similarity between the Q column and the C column using the DTW algorithm.

It is assumed that Q columns and C columns of lengths n and m can be expressed as Equation 1 below.

The Euclidean distance between component q _{i of} column Q and component c _{j of} column C can be expressed as d (q _i , c _j ), with the Euclidean distances between the components of column Q and the components of column C in the nxm matrix. Can be expressed. Here, the (i th, j th) component (q _i , c _j ) of the nxm matrix and the Euclidean distance d (q _i , c _j ) are d (q _i , c _j ) = (q _i , c _j ) ^Have a relationship of ^two .

When W is a set of components that define a mapping relationship between Q and C among the components of the n × m matrix, W denotes a warping path.

When k-th component of W is called w _k = (i, j) _k , W may be defined as in Equation 2 below.

Here, the warping path W has several limitations as follows.

Indicates that w ₁ = (i, j) and _{w K = (n, m)} K, this is the direction toward the last element from the starting components direction of warping path: threshold condition (boundary conditions).

이고,

이다. Continuity condition : w _k = (a, b), w _{k -1} = (a ', b'), where

ego,

to be.

이다. Forging (Monotonicity) condition: a _{w k = (a, b)} , w k -1 = (a ', b'), where:

to be.

With the above limitations, a warping path that minimizes the overall cost and the cost (similarity) of the warping path can be calculated as in Equation 3 below.

As a result, the stock quote prediction method according to the present invention may select a pattern type having the highest similarity to the current trend of the stock market using a DTW algorithm among a plurality of pattern types. In addition, the stock quote prediction method according to an embodiment of the present invention may generate a buy signal, a sell signal, or a hold signal according to trading rules (first conditions and second conditions) determined for the selected pattern type. .

In addition, the stock quote prediction method according to an embodiment of the present invention generates any one of a buy signal, a sell signal, or a hold signal according to the second condition, and actually trades according to any one of the generated signals. (S180).

The stock quote prediction method according to the present invention may be implemented in the form of program instructions that may be executed by various computer means and may be recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer readable recording media include magnetic media such as hard disks, floppy disks and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

5 is a block diagram illustrating a stock quote prediction system according to an exemplary embodiment of the present invention.

Referring to FIG. 5, a stock quote prediction system according to an embodiment of the present invention may include a historical information analyzer 510, a setter 520, a database 530, a selector 540, and a signal generator 550. ).

The historical information analyzer 510 extracts time sections corresponding to predetermined pattern types based on the historical price information of the stock market over time.

In addition, the setting unit 520 sets first conditions corresponding to the predetermined pattern types based on the technical indicators in the extracted time intervals and the past price information.

In addition, the database 530 calculates the rate of return of the first conditions through a virtual experiment, and stores the second conditions in which the first conditions are optimized according to the calculated rate of return of the first conditions.

In addition, the selector 540 selects a type related to the current market price of the stock market among the predetermined pattern types by using a dynamic time warping (DTW) algorithm.

In addition, the signal generator 550 generates one of a buy signal, a sell signal, or a hold signal by using the second condition set for the selected type.

Although not shown in FIG. 5, the contents described with reference to FIGS. 1 to 4 may be applied as it is, and detailed description of FIG. 5 will be omitted.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims below but also by the equivalents of the claims.

1 is a flowchart illustrating a stock quote prediction method according to an embodiment of the present invention.

2 is a table showing predetermined pattern types and corresponding time intervals according to an embodiment of the present invention.

FIG. 3 is a diagram illustrating historical quote information and technical indicators derived from past quote information according to an embodiment of the present invention.

4 is a diagram for explaining a DTW algorithm.

5 is a block diagram illustrating a stock quote prediction system according to an exemplary embodiment of the present invention.

Claims (15)

Extracting time intervals corresponding to predetermined pattern types based on past market price information of the stock market; Setting conditions corresponding to the predetermined pattern types based on the technical indicators in the extracted time intervals and the past quote information; Selecting a type related to the current market price of the stock market from among the predetermined pattern types using a dynamic time warping (DTW) algorithm; And Generating any one of a buy signal, sell signal, or hold signal using a condition set for the selected type; Stock quote prediction method comprising a. The method of claim 1, Selecting a type related to the current price of the stock market Determining similarity between the predetermined pattern types and the pattern of the current market price of the stock market using the DTW algorithm, and selecting a type related to the current market price of the stock market according to the determined similarity. Stock quote forecasting method. The method of claim 1, Setting conditions corresponding to the predetermined pattern types Discretizing the technical indicators in the extracted time intervals; Filtering the discretized descriptive indicators using a data mining technique; And Determining conditions corresponding to the predetermined pattern types using the filtered discretized technical indicators. Stock quote prediction method comprising a. The method of claim 1, Setting conditions corresponding to the predetermined pattern types Performing virtual experiments on the conditions and extracting conditions that yield a higher yield than a predetermined level Stock quote prediction method comprising a. The method of claim 1, Performing trading according to the generated one signal Stock quote forecasting method further comprising. The method of claim 1, Updating the conditions corresponding to the predetermined pattern types based on the trading result performed according to the generated one signal Stock quote forecasting method further comprising. The method of claim 1, Wherein the predetermined pattern types comprise at least one of a short-term rise, a short-term fall, a trend-high rise, a trend-down, a ceiling sideways or a bottom sideways. The method of claim 1, The stock quote information prediction method, characterized in that the market price, closing price and trading volume. The method of claim 1, The stock quote information according to the time of the stock market is related to the Korea Composite Stock Price Index (KOSPI) or KOSPI 200 index. Extracting time intervals corresponding to predetermined pattern types based on past market price information of the stock market; Setting first conditions corresponding to the predetermined pattern types based on the technical indicators in the extracted time intervals and the past quote information; Calculating yields of the first conditions through a virtual experiment and generating second conditions in which the first conditions are optimized according to the calculated return rates of the first conditions; Selecting a type related to the current market price of the stock market from among the predetermined pattern types using a dynamic time warping (DTW) algorithm; And Generating one of a buy signal, sell signal, or hold signal using a second condition set for the selected type; Stock quote prediction method comprising a. The method of claim 10, Selecting a type related to the current price of the stock market Determining similarity between the predetermined pattern types and the pattern of the current market price of the stock market using the DTW algorithm, and selecting a type related to the current market price of the stock market according to the determined similarity. Stock quote forecasting method. The method of claim 10, Performing a trading according to the generated one signal; And Updating the second conditions based on a trading result performed according to the generated one signal Stock quote forecasting method further comprising. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 1 to 12. A historical information analyzer configured to extract time intervals corresponding to predetermined pattern types based on past market price information of the stock market; A setting unit configured to set first conditions corresponding to the predetermined pattern types based on the technical indicators in the extracted time intervals and the past price information; A database configured to calculate a rate of return of the first conditions through a virtual experiment and store second conditions in which the first conditions are optimized according to the calculated rate of return of the first conditions; A selection unit for selecting a type related to the current market price of the stock market among the predetermined pattern types using a dynamic time warping (DTW) algorithm; And A signal generator for generating any one of a buy signal, a sell signal, or a hold signal by using a second condition set for the selected type. Stock quote prediction system comprising a. The method of claim 14, The selection unit Determining the similarity between the predetermined pattern types and the pattern of the current market price of the stock market using the DTW algorithm, and selecting a type related to the current market price of the stock market according to the determined similarity. Stock quote prediction system.

Images