TW200945236A - A method of predicting a financial market - Google Patents

Abstract

A method of predicting a financial market is provided. The method uses a PARK filter to predict a financial market, combining processes of statistical analysis, fuzzy clustering and grey prediction to determine a risk level corresponding to a transaction time.

Description

200945236 IX. Description of the Invention: [Technical Field of the Invention] The present invention relates to an information analysis and prediction method, and more specifically, the present invention is a novel method implemented in an artificial intelligence stock trend analysis and prediction system. Track and analyze the changes in stock market information, and propose operational strategies with different levels of risk as the basis for investors to enter and exit the stock market to improve investment efficiency and reduce risks. ❹ 【Prior Art】 According to the prior art, the following methods can be selected for stocks with investment value: 1. Basic analysis The basic analysis method is to consider the various factors that will affect the price to predict the future trend of the market. Investors must observe and analyze various political and economic factors that may affect market prices, and try to predict the trend of future price changes from the supply-demand relationship behind commodity prices. For example, investing in a company is as follows. : 1. Find out the history of the company you want to buy and the economic background of the industry. 2. Estimate the net worth surplus and dividends. 3. Estimate the stock production and sales situation and operational outlook and financial overview. 4. Estimate the future growth of the stock, whether the P/E ratio is reasonable, the dividend policy, and compare it with other similar types on the market. 5. Based on the above data or instructions, it can be used as an basis for investors to judge investment stocks. Of course, the basic analysis should not only consider the above factors, but also consider other external factors, such as: political factors, etc., must be combined with various factors before final decision. When looking for factors that affect the market, because other important factors may be missed, in the basic analysis, investors must have relevant financial knowledge to analyze the results, which is a big problem for the average investor. Obstacles. Generally speaking, the company's relevant information usually lags behind the market reaction. If the investor waits until the company publishes the information Φ, it will analyze and predict it. Usually, the market has already completed the reaction, and the result may cause chasing and killing, even Changes in economic structure or unanticipated changes may lead to imbalances in market supply and demand, affecting price forecasts. Second, the technical analysis The technical analysis is mainly to use the graphical or quantitative technical indicators to study the market sentiment or supply and demand of commodity transactions, and then decide the timing of trading. Because technical analysis can not only predict the price and judge the market _ investment timing, but also can be used as a warning signal, more importantly, the basic analysis is simple and easy to understand, so it is widely used by the general public. The following will introduce the commonly used in the market. Technical analysis methods: 1. Dow Theory Dow Theory is based on Trends and Patterns. The trend is composed of branch lines and pressure lines, while the shape trend is divided into There are two types of Continuation Patterns and Reversal Patterns. The Bullish Market, the Short Market 6 200945236 (Bearish Market), Support, Pressure, Head and Shoulders Formation, etc., which are well known to the public today, are all developed by Dow Theory. come out. After understanding the Dow's theoretical analysis method, combined with statistics, the results are derived from a variety of price analysis techniques. 2. Wave Theory was proposed by RN Elliott in 1943. The theory is that market prices are cyclical and volatility. There are eight waves in a complete cycle, including five main bands. And three back-shift waves, this theory combines trend lines, target area predictions and natural laws, and is one of the commonly used analytical tools in the market today. 3. Moving Average. The moving average (ΜΑ) is the most widely used technical analysis tool, mainly because it is easy to understand and easy to calculate. Since it calculates the average stock price over a period of time by statistical methods and then joins the points into a smooth curve, the moving average has the characteristics of smoothness, stability and trend. The moving average is a reference for determining the timing of buying and selling. The eight rules proposed by US investment expert J0seph Granville are set for the timing of buying and selling mobile averages. 4. Volume (V0L) Use the volume moving average to determine whether the trend is up or down. When the volume increases, it means the trend is upward, 7 200945236

= Dan ^ and technical indicators nuclear (four) up or down. The Relative Strength Index (rsi) / RSI uses the stock price movements over a period of time to predict future price movements. The basic principle is that in a normal stock market, the strength of both buyers and sellers must be balanced and the stock price will be stable. RSI is the ratio of the average value of the stock price increase to the average of the total increase and decrease in a given period of time. Take the RSI value on the 6th as an example. If the value is 80 or more, it is overbought, 9〇 or M is the selling point, 20 or less is oversold, 1〇 or below is the Bet, and the RSI will appear before the stock price change. Ling or Bottom, can reflect the ups and downs of the stock price in advance, can be regarded as the leading indicator of the trend of the market index; if the RSI on the 6th through the 12th RSI from the bottom up, it can be regarded as a buy point; otherwise, when the RSI is on the 6th When it breaks through the 12th rsi, it can be regarded as a selling point; the trend of the RSI line and the trend of the broader market index indicates that the market is about to reverse. 6. Stochastic indicator (KD) A technical indicator that combines the advantages of the RSI indicator, the moving average, and the quantitative energy concept. If the market is a clear rise: the potential will drive the K and D lines to rise to 8. If the rally starts to slow, it will reflect the K and D values, causing the depreciation to fall below the D value. The situation is established; when the Κ line falls from the top to the D line, and the D value is above 80 (overbought area)' is the buy signal; when the Κ line breaks through the D line from the bottom up, and the value is below 20 (oversold) Zone) appears as a buy signal; when the devaluation is greater than 80 and the D value is greater than 70, it means that the closing price of the day is in the high price range, that is, the overbought condition; when the Κ value is less than 20 and the D value is less than 30, It means that the closing price of the day is in the low price range, that is, the oversold condition; the price is high or low, and KD does not have this phenomenon. This is a divergence phenomenon, which is an important precursor to possible reversal. The exponential smoothing similarity moving average formula (MACD) (DIF) uses the fast and slow two exponential smoothing moving averages (EMA) to calculate the difference divergence (DIF) between the two, and then use the difference divergence and The difference between the mean (DEM) and the divergence (separation) of the difference mean value (DEM) is used to judge the timing of the stock market buying or selling. If both MACD and DIF are positive, it can be regarded as a long market; MACD and DIF is negative, which can be regarded as a short market; DIF breaks through MACD and buys signals; DIF breaks down MACD and sells signals; DIF turns from negative to positive, and crosses MACD to buy signals; DIF value is forwarded Negative, and break through the MACD, sell the signal; when the DIF and the broader market index deviate from the trend, if the stock price continues to innovate low, and the DIF value does not hit a new low, 200945236 = positive deviation from the trend, as a buying opportunity; Continuous innovation high point, while DIF value # • point, this is the "negative divergence" trend, for the sale of two new in the stock market, there is a lot of information for the sale timing. Other Shanggu 11 # more commonly used technology refers to the news, investors 彳 * line, weekly line, monthly line, melting f, securities lending ... and other resources easy to calculate analysis Lai computer can be very easy to easily analyze various technical indicators, but single The backward trend and the purification of the indicator (4) occurred, and the invention can make the measurement result very inaccurate. I hope that wisdom (αι), so-to, not only the change of the situation V4:;= force, but also adapt to the various stock market in the field of knife prediction, gray sound application of a wide range of prediction methods, j-day connection type The advantages and disadvantages of the two methods are different, and the advantages and disadvantages of the two methods can be complementary. The gray Markov pre-comparison is a prediction method combining the advantages of gray and Markov models. Less and more widely used models are more suitable for a small amount of data (more than four two can be predicted, only a smooth index function curve = its fitting function is poor, the prediction accuracy is also better than the two The data is intended to be W 4b 6^1 Φ, m ^ ^ The object of the Markovian type is a random machine_Daji system. '1', the basis is that the Markov process is applicable to the time series of the data with the number of eves of the presentation. It is the random process of the grass that sews with time as the time goes by. The process always deviates or oscillates with the trend of various random factors, so if you use different GMs (1,1) The model is first obtained in the process The trend is to use the Markov model for transition probability analysis. The combination of these two methods will improve the prediction accuracy of GM(1,1) and expand the application range of GM(1,1). In the stock market, Investors are almost always making big losses, resulting in funds often being stuck so that turnover is not working or forced to break their heads, etc. When investors buy and sell stocks, they often diversify their investment targets to avoid risks, but they cannot At the same time and long-term tracking of the target of investment, or lack of risk control ability, resulting in investment losses, but if the target concentration _ single stock, its investment risk is relatively high, may lose all chips at a time From the above description, we can understand that there are too many technical indicators in the stock market, and to understand the fundamentals of stocks, such as financial status, future growth, etc., but also to grasp the economic and political aspects of the entire environment. Such factors as the average investor are difficult to grasp. The stock market is often filled with all kinds of rumors and gossip, often resulting in investors. Trouble, which in turn affects investors to make correct judgments, various factors make a common investor's mind to form a difficult gap. In the stock market, the judgment of the trend is the most important, as long as the trend can be mastered The key to profitability. The stock market is nothing more than three states: “long”, “short” and “consolidated”. If the future stock market is predicted to be “long”, investors can enter the market to buy stocks to increase their shareholding ratio. If the future stock market is predicted to be a “short” trend, investors can profit as soon as possible to make a profit or a short-selling. If the future stock market is predicted to “consolidate”, it should avoid large inflows and exits and reduce the shareholding ratio, or even withdraw from the market. In order to reduce the risk of investment. If investors can grasp the trend of future stock market, the chances of investing in stocks will increase and the risk of investment will fall. 11 200945236 Low. Φ Φ The above-mentioned stock investment method, whether using technical analysis or basic analysis, is a backward indicator. Although it still has its usefulness, it is often slow and inferior. Instead, it causes investors to chase high and kill low, and become the target of people who are interested in killing. In addition, the influence of investors' judgments also includes various news situations that often appear in the stock market, as well as investors' own emotional problems, which often lead to negative judgments. In view of this, the present invention develops a method with artificial intelligence and makes the computer an artificial intelligence system with "human intelligence" through the implementation of the system of the present invention. This system can not only predict the results from the received information, but also avoid the problem of “people” as the decision-making center in the message and personal situations, and achieve the goal of “humanity” and “governance”. [Inventive content] In order to have the ability of artificial intelligence, f investors can make predictions, analysis and decision-making. The invention adopts a novel filtering method PARK, which has the power to hide _ (material pure, _ clustering, grey prediction) ) and excellent operational strategies, etc., applied to thousands of stock markets' in the hope that investors who can help Guang A will make more losses in the stock market. Among them, the PARK filter rule is a kind of filter that can filter out the necessary noise of the original trend of the market to produce a better trend signal, providing investors with a more delicate operation direction. Statistical analysis, fuzzy clustering, gray and ^ 孰 慧 特 特 的 ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' The majority determines the risk level, and the difficult clustering determines the fuzzy risk level by the maximum degree. The gray Markov prediction is based on the band variation of the recent PRK band of 12 200945236 1 kg to predict the current pARK band. After '35, match some experience values to determine the risk level. After determining the η medium-low risk level, the long-term (long-term long-term + short-term and short-term (conservative and fierce) and other operational strategies = upper stop loss mechanism] are set according to different risk levels for investors to earn unlimited at low risk. Under the benefit of the experiment, it is confirmed by experiment that the use of the bend m to alleviate the trader's mind must incorporate the stop-loss mechanism into its transformation = the purpose of the vertical supply--the kind of change strategy for stock market information, the ancient squad &amp; Get stable under the operation of risk level =: The capitalist is in the stock market, and can provide a method to judge the trend of the noise in the low-risk and high-cost market. The law, no matter how the situation changes, can make the revised original trend closer to the generals. The purpose is to provide a risk cluster that can use the artificial intelligence model to attack and defend different risk levels for the system to set i:: = The method of strategy 'to help investors improve profitability and the effect of the other - the purpose is to provide a way to make the prediction model to find a better change of the two stocks; more adaptive by a systematic approach Sex to A disk f or habit change will make this 13 200945236 in the above information can be a lot of, the general investor's fundamental interest in the solution shows the relevant market intelligence and indicators, and is vulnerable to external: time t long-term tracking of all the information The change of the invention and the failure of the investment. The invention is swaying, causing the wrong method of judgment, which can be used for the operation strategy of the investor's insurance level or closing the position, and no longer needs the judgment and behavior of the poor (buy, sell and report.) Time-consuming energy to study the various unknowns and a large number of related Φ Lu formers, in which the relevant method of generating the required white h self-sounding is used to analyze the history of the market index in the stock market. The third part, the second of which is the original input data 'return method too; fφ a. The knife uses the park filter rule for investment ί==:=χ χ better trend signal - statistical analysis of intelligence, blur Clustering = part is the trend classification after processing with the artificial filtering rule, the square:, will pass the = class, as for the third part, it will be based on different risk levels to set the long-term (long + short) , long-term system, two-investment, and other operational strategies' plus stop loss or "ir i decided to take "buy", "sell" ^ thousand positions" to earn unlimited profits. The present invention will be fully described in the following description of the preferred embodiments and the accompanying drawings, but it should be understood that those skilled in the art can modify the embodiments described herein while The same effect as the present invention is obtained. Therefore, it is to be understood that the following description is a broad disclosure of the invention and is not intended to limit the invention. The following is an extended description of the technology related to the background of the present invention. Although readers with familiar experience and knowledge can choose to skip or not read the following background information, understanding this information can further understand the specific implementation of the present invention, and the recommendation should be read φ. Referring first to Figure 1, there is shown a schematic diagram of the method of the present invention. In an embodiment of the method of the present invention, the original trend 111 is first generated using the technical indicators, and a more reliable trend band is obtained by performing the filtering rule 110. The main purpose is to perform a more south-quality long/short band. The work of the series trend classification. These rules evolved from different perspectives and three main indicators: PCI < Phase Change Index>, PSAR_AF_P &lt; Possibility measure of Acceleration Factor ❿ about Parabolic Stop-And-Reversal> and K-Bband_Rate < Rate of K-Bollinger Band). Therefore, the term "PARK" consists of the prefix "P" of PCI, the end "AR" of PSAR, and the prefix "K" of K-BBand. In other words, PARK is a new trend indicator 115 filtered based on key features of PCI, PSAR and K-Bband. The algorithm for filtering rules is as follows and its associated decision tree is shown in Figure 2. Filter Rule 1 <PCI1>:

Raw Trend - + : 22Mom &gt; 0 and 22PCI &lt; 20

Raw Trend + — - : 22Mom &lt; 0 and 22PCI &gt; 80 15 200945236 Filter Rule 1 &lt;PCI2>:

Raw Trend --^+: 22Mom &gt; 0 , 22PCI &lt; 25 And 30Mom &lt; 25 , 30PCI &lt; 20 - [Double Bottoms] pattern Raw Trend 22Mom &lt; 0 , 22PCI &gt; 75 And 30Mom &gt; 0 , 30PCI 〉80 -【Double Tops】 pattern Filtering rule 2<PSAR_AF_P&gt;

When Raw Trend changes , verify as follows :

If PSAR Flag(day) x PSAR_Flag(week) &gt; 0 [PSAR1] 〇 If PSAR_AF_P(day)+PSAR_AF_P(week) &lt; 1 then

Reject Raw Trend End If

Else [PSAR2]

Accept PSAR—Flag of

Max(PSAR_AF_P(day), PSAR_AF_P(week))

End If filter rule 3 <Bband_Rate> _ Raw Trend +-^-: ❹

If lOBband Rate &gt;= 5% then Accept Raw Trend End If the statistical analysis 131 with the characteristics of the artificial intelligence model 130, the fuzzy cluster 132 and the gray Markov prediction 133 are used to determine the timing risk level. The statistical analysis 131 is to estimate the expected cut-off time and price fluctuations of the long-band by means of the mean and the mode to determine the degree of risk based on these empirical values. The histograms of the duration of the trend and the price fluctuations, as well as the associated risk levels, are shown in Figures 3 and 4. Fuzzy clustering 132 uses between 0 and 1 16 200945236 book; = the way the generation is explicitly assigned to a group 'in the description of the P sub-continuation period and the price fluctuations in the discussion universe defined model # fuzzy risk ^ 6th ) ' Then according to the maximum degree of ownership, the other is determined to be the other, the injury is gray Markov prediction 133, which is determined by the prediction model = Markov model, where (10) (10) input capital: k ARK band high / low point The data, and the original data sequence is obtained according to the number of days between the first and the first::, then the series can be modeled. The steps to establish a GMGJ) prediction model are as follows: 1. The data required for the calculation is obtained, so that % = (xd(1), 々(2),..."x.(8),..·.·,χ (8)) (1) 2, the operation of 1-AGO is performed to obtain the following equation V1) = (X(1)(1), X(1)(2), χ(1)(3),, χ(1)(8)) (2) where X(-k )~~ Σ^ο)5^= (3) 3· By performing a meam generating operation on 1(1), we get (1)=(^)(1)/)(2)/)(3) of Art(1), , (8)) (4) where Z (1)(A:) = 〇.5χ(1)(8) + 0·5χ(1) (Λ -1) (5) 4. The original equation for GM(1,1) modeling is as follows xi〇 )(k) + azm(k) = b, k = 2,3,....,n dxm dt

=b (6) 200945236 5. Generate the parameter ", the solution of 6 is based on the least squares method ij θ = \ϊ\ = (ΒτΒΤ'ΒτγΝ (7) b where Αν \ly \|7 \1|/ 1/ \ ) / 12 12 12 (/(. (--- ♦ -( /V zz:z 0)31 (( , .—..-)-(0&lt;2 = ^ \—/ 8 /IV 7 U 0) π) 6. After the operation of equation (7), the values of the parameters α and δ are obtained as follows: a Σ zS) Σ xw ~(η~ !)Σ ζ!Άχ!'ί k=2 k=2

A (9) (10)

Yx(0)Yz(1)2 -Yz(1) Yz(1)x(0) ZjX(*)ZjZ(*) ZuZ(k) ZaZ{k)X{k) k=2 k=2_k^2 k=2_ where δ = («-ι)Σ^!2-(Σ4!)2 (11) k=2 k=2 Substituting a, b values into the shadow equation (6) The resulting predicted value is (12) ) X (1) (众+1) = [x( 18 200945236 where branch (0)(1)=,(1) (13) 8. For i(1)(々+ 1), 1-IAGO is the last predictable value i ((V) ) is χ(0)(4)=f(1)(Α;) - ί(" (it -1) = (1 - 〇[x(0)(1) - soil K--1) ( 1 4 ) a 9. Carry out the actual value Analysis of error between the predicted value and Shanghai (8) e(8)= x(0\k)-x(0\k) xm(k) xlOO%, k = 2,3,4,...··,η. (15)

The main reason for PARK to refer to the high point number is that it can be simulated by past PARK training data, and the next unknown high point of PARK can be predicted and the number of expiration days from the next high point can be calculated at any time point. The number of due dates determines the level of risk. The Gray Markov Model GMM (1, 1) is a useful tool to improve the accuracy of GM (1, 1). The GM(1,1) prediction procedure must first be completed and its predictions treated as estimates. The second is to divide the state space by a certain percentage of the GM(1,1) estimate ® data and run out the most appropriate number of states with the simulator. By comparing the raw data with the GM(1,1) estimation data, the state to which the original data belongs can be known and the state transition probability matrix can be established at the same time. Through this transfer probability matrix, we can generate a prediction table and calculate the new prediction result by taking the upper and lower bounds. Finally, after the GMM (1, 1) correction procedure, a more accurate prediction result is obtained. The steps are outlined below: (1) Use GM (1, 1) to predict the results. (2) Split the state space. 19 200945236

(3) Obtain a state transition probability matrix. And (m) Ο/,),....·,px[m) p2(n•....,p2(km) p(n,) p(m) p(m) ^ k] ...... , factory Jki where P^m)=

Measuring table.曰期 state transition step probability reserve state 1 ------ 0.5 -··. state η 〇-^j〇5〇2 ”&quot; __?___ 0.333 ----- 0 ~~t~- • • Summer ------ 1.0 ----—* * · __ 0.2 — Total — — — — 1.833_^ 0.2 (5) Calculate the predicted result. = 〇·5 X (Ay. + By) or ^GMM = XGM X ^Pi+2Pj- Use the statistical analysis 131, fuzzy clustering 132 work wisdom method to determine the risk level, that is, with the gray prediction 133 and other "risk clustering" 』, 20, 200945236 Then use the “strategic analysis” including trend forecasting, market timing, capital management and stop loss mechanism to provide different ways for investors to operate, and finally use a lot of quantitative information to verify the performance of each operational strategy. To help investors improve investment efficiency and reduce investment risks. Any successful trading program must consider three important factors: price forecasting, market timing and location management. The strategy analysis 150 of the present invention has long-term [long + short] 151, long-term + short-term 152 and short-term [conservative and intense] 153 three. The strategy for long-term [long + short] 151 operation in this embodiment is to use the PARK indicator to distinguish the long/short band. Make four more in the long band and four in the short band. As for short-term [conservative and intense] operational strategies, the following can be subdivided into the following categories: a. Signal type: The trading operation is based on the risk level generated by the risk cluster mentioned above (see Figure 7). First, you must decide the upper limit of the short-term part. Secondly, if you encounter low risk, you will add two more, and the medium risk will be overweight. Finally, when the high-risk comes, the code is reduced one by one until the short-term part is all closed. In addition, we must also consider the stop loss mechanism, its operation rules are as shown in the table below. Table 1 Different risk levels of long and short strategy risk level long short head high risk waiting for closing position waiting for closing position risk for more shorting low risk strong for more powerful Venting b. Intense: Different from the "signal type" strategy, "aggressive" seems to be more active. When a low risk is encountered, the short-term part is immediately added to the full position. "Strong type" 21 200945236 "Powerful operation" and "operation" similar to the "signal type". The table is also slightly shown in Figure 8 and Figure 9. ', Guan Jia minus code part map shows that the short-term strategy has two important appeals: 1-2: still need to include the stop-loss mechanism to mitigate ❿ After the performance evaluation process 18 can be used, the performance measurement indicators of the present invention are as follows To determine (4) the net profit or loss of the performance (or the percentage of net profit or loss): Net profit and loss is the most common performance statistics. It is the total amount earned or lost during the backtesting. The strategy of losing money is treated with coffee. #读钱的略略 Total net profit/loss = Gross profit _ Gross loss ί时易成本' We can also use the net profit and loss percentage to table (7) Because of the percentage of gross profit = Gross profit percentage - the percentage of domain loss is also due to simple calculation Commonly built in computer programs, so the Bay == performance: estimated metrics. If the strategy is profitable, the child will be greater than one, and the strategy for loss will be less than one. Profit factor = gross profit / gross loss, <3> average profit and loss ratio: = ratio is similar to profit factor, but it is more meaningful for judging the profit factor. 22 200945236 number] average profit and loss rate = profit factor · [_ times / profit times <4> winning percentage: winning percentage is another rate that many people use to measure the success of the strategy. The number of profit is divided by the total number of transactions. Traders are too much to win, and there are myths that try to maximize winning percentage. , Emphasis on winning rate = [profits / total number of transactions] < 5 > Maximum decline: 跌 The decrease is the calculation of the net profit and loss from the highest point, the maximum value is the maximum decline. <6> Maximum continuous loss: The maximum continuous loss is the biggest risk of continuous trading losses. This concept is more in the futures market. The greater the maximum loss, the greater the required margin. The trader is a very good strategic performance measure. , mouth knife <7> ideal profit ·· ^ Another profit indicator is called ideal profit, defined as ·· ideal profit =100 χ Σ♦[咐·知)]·μ(0| ...· ZRoi~~~ Here is the actual price variation on t day, f is the variation of the compensation grid, and this sum is counted in the county W test observation window = ratio, this value falls in the interval of -] to +〗. The relevant information of the example 'first obtain the second order of the public order == price index (10)) is the Republic of China 85/]0/19 to the Republic of China _2/31, and 23 200945236 test interval for the Republic of China 90/1/1 93/12/31. According to the experimental data and related graphs, the performance of pARK is the best in the model (see Figure 1G). The long-term and short-term departments, ", 1 〇 mouth and short-term stop loss is set to 3%, and the higher the proportion of the short-term strategy part of the artificial intelligence climbing model Z, the better the performance is. I). The effects of strategy execution are different. I* never miss the same metrics and the results are not the same. However, the performance of Qu @Shi Lai active strategy is better than that of conservative one. The profit is increasing, This proves that the combination of PARK and tvb2s to promote the stability of the strategy can indeed enable investors to obtain a stable south salary in a low-risk state.

According to the best of the previous day (4), the noise in the beginning trend of Xiang PARK to produce a better trend of the operation direction of the trend = the trend of the statistical i with artificial intelligence; eight categories, gray prediction, etc. The winds will be processed by the filtering rules to generate (4) wind-level winds. (High, head + short) According to different risk levels, long-term (long-term) long-term + short-term and short-term (conservative and intense) operations are set. = Field loss t system, for investors to decide at a low risk that they should be produced "or thousands of positions" to earn unlimited profits. The predicted values generated by different gray Markov models (10) can be repaired: and = the best predictive model m A &quot; to produce the best predicted value. The amount of data η, the number of states i and the number of transition steps m of the rolling model of the predicted value of the predicted value are not fixed 24 200945236, and will follow the prediction model. The amount of data is automatically changed, that is, the prediction model GMM (1, 1) can automatically generate the best prediction result along with the amount of data used by the optimal prediction model. In a preferred embodiment of the invention, Select the market (TAIEX) and the original driving line as the source of the original data, all the analysis is for the two items. Investors can change it to other technical indicators, such as MACD, RSI and other technical indicators, or in the analysis In addition to other moving averages or indicators, it may be possible to obtain better prediction results, which of course has to be verified experimentally. _ The method of implementation of the present invention has been described in detail in the foregoing embodiments, and anyone skilled in the art can follow the present invention. It is to be understood that the invention may be modified and modified as needed without departing from the spirit and scope of the invention, and therefore, other embodiments are also included in the scope of the present invention. Using fuzzy clustering and gray Markov model to predict the trend of the market, that is, using artificial intelligence to predict the turning point of the future stock market, not only allows investors to understand the future stock market trend and grasp the investment newsletter first, and thus effective Reducing investment risk and loss _ loss. It can be understood that the present invention has many excellent characteristics, which can not only solve the lack and inconvenience of some practical applications, but also provide a cost-effective solution, and also allow investors to be in a short time. Relevant investment information is obtained to avoid the influence of external environment and human factors, and the risk of effective control of investment has been met. It has already met the application requirements for invention patents, and the bureau can be reviewed and patented to provide preferential treatment. 25 200945236 [Simple description of the diagram] Figure 1 is a schematic diagram of the analysis and prediction method of the present invention. Figure 2 is the reliability of the original trend. Figure 3 is the histogram of the duration of the trend and the risk level. Figure 4 shows the histogram of price fluctuations and risk levels. Figure 5 shows the blurring of risk levels during the trend period. Generic function. Figure 6 is the fuzzy attribution function of the price fluctuation risk level. Figure 7 is the PARK long/short band risk map. Figure 8 is the "aggressive" strategy plus coded part map. Figure 9 is the "aggressive" strategy. The code reduction part map. Figure 10 is a comparison of the total net income of various trend models. Figure 11 is the part operation diagram of the strategy “Statistics_Signal” (long-term + short-term = 10 ports). [Main component symbol description] Original trend ~111; artificial intelligence model ~130; fuzzy clustering ~132; operation strategy ~150; long-term + short-term ~ 152; performance evaluation process ~ 18. Filtering rules ~ 110; PARK trend ~ 115;

Statistical analysis ~131, gray Markov prediction ~ 133; long term [long + short] ~ 151; short [conservative, intense] ~ 153; 26

Claims (1)

200945236 X. Patent application scope: 1. A method for predicting the market trend by using fuzzy clustering and gray Markov model, which is used to evaluate the market trend of the stock market and provide information on the operational strategies of different risk levels, including the following procedures: Historical data of different technical indicators in the market, the historical data including the market index, MACD and different original trend values; using the filtering rule PARK to filter the noise of the original trend value Off, to generate new trend indicators, as statistical analysis of artificial intelligence models, fuzzy clustering and gray prediction GM (1,1) to generate risk clustering modeling data; modified by gray Markov model GMM (1,1) pARK Predicting the predicted value of the model GM(1,1) pARK to obtain more accurate prediction results; using the aforementioned risk clustering to set short-term (conservative, intense), short-term + long-term and long-term (long + short) And other operational strategies to provide investors with when to spend more or short and how many chips to make; and to use custom assessment indicators (net profit and loss, profit factor, average profit and loss ratio, winning rate, maximum decline and maximum continuous loss) to evaluate Operational performance. 2. The method for predicting the market trend by using fuzzy clustering fish ash 如 型 as described in the scope of application for patents, wherein the technical indicators collected include the market index, MACD and different original trend values. 3 soil as described in the scope of the patent application, using fuzzy clustering and Markov model to predict the trend of the market, where 廯 3〇-l〇-10 MACD is the difference between the 30th and 1st day index moving flat two. , g can go to the method of using fuzzy clustering and gray: letter τ Γ 制 A , , , , , , , , , , : : : : : : : 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用 利用2 27 200945236 is produced by AIFTS (Artificial Intelligence Financial Trading System). 5. The method of predicting the market trend using fuzzy clustering and gray Markov model as described in item 1 of the patent application scope, wherein the filtering algorithm PARK has four algorithms: PCI1, PCI2, PSAR-AF_P and Bband_Rate. 6. The method for predicting the market trend using fuzzy clustering and gray Markov model as described in item 1 of the patent application scope, wherein the methods for generating risk clustering are statistical analysis, fuzzy clustering and grey prediction. 7. The method for predicting the market trend using fuzzy clustering and gray Markov model as described in item 1 of the patent application scope, wherein the optimal modeling point of the gray prediction model GM(1,1) PARK is 4 points, ie four Pen data. 8. For the method of predicting the market trend using fuzzy clustering and gray Markov model as described in item 1 of the patent application, in which the predicted value of the gray model GM(1,1) PARK is corrected, GMM(1,1) PARK can be used. The predictive model predicts trends. H 9. The method for predicting the market trend using fuzzy clustering and gray Markov model as described in item 1 of the patent application scope, wherein the measurement indicators for measuring performance include net profit and loss, profit factor, average profit-loss ratio, winning rate, The biggest decline and the largest continuous loss. 28