WO2020024456A1 - 一种量化交易预测方法、装置及设备 - Google Patents

一种量化交易预测方法、装置及设备 Download PDF

Info

Publication number
WO2020024456A1
WO2020024456A1 PCT/CN2018/111689 CN2018111689W WO2020024456A1 WO 2020024456 A1 WO2020024456 A1 WO 2020024456A1 CN 2018111689 W CN2018111689 W CN 2018111689W WO 2020024456 A1 WO2020024456 A1 WO 2020024456A1
Authority
WO
WIPO (PCT)
Prior art keywords
transaction
transaction data
quantitative
failed
tested
Prior art date
Application number
PCT/CN2018/111689
Other languages
English (en)
French (fr)
Inventor
毕野
黄博
吴振宇
王建明
肖京
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2020024456A1 publication Critical patent/WO2020024456A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
    • G06Q10/06375—Prediction of business process outcome or impact based on a proposed change
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange

Definitions

  • the present application relates to the technical field of financial transactions, and in particular, to a method, a device, and a device for quantitative transaction prediction.
  • Quantitative trading refers to the use of advanced mathematical models to replace human subjective judgments, and the use of computer technology to select a variety of "high probability" events that can bring excess returns from huge historical data to formulate strategies, which greatly reduces investor sentiment The impact of volatility, to avoid making irrational investment decisions in the extreme frenzy or pessimism of the market.
  • the present application provides a method, device and equipment for predicting quantitative transactions, the main purpose of which is to solve the problem that currently only manual analysis and query can be performed on failed quantitative transactions, so that the accuracy of quantitative transactions is low.
  • a quantitative transaction prediction method includes: obtaining a current market indicator and a current transaction indicator of a current transaction market, wherein both the current market indicator and the current transaction indicator will change over time. Changes continuously; the current market indicator and the current trading indicator are substituted into a quantitative trading prediction model, and corresponding parameters in the quantitative trading prediction model are updated, thereby ensuring the accuracy of the corresponding parameters.
  • the quantitative trading prediction model is obtained by learning and training based on historical failed transaction data.
  • the quantitative trading prediction model is provided with parameters corresponding to the current market index and the current trading index, respectively.
  • Data input the transaction data to be tested into the updated quantitative transaction prediction model to predict, and determine the failure probability that the transaction data to be tested belongs to failed transaction data; compare the failure probability with a predetermined probability, and when the When the failure probability is greater than or equal to the predetermined probability, the Easy data belonging to failed transaction data, or data belonging to a successful transaction.
  • a quantitative transaction prediction device includes: an obtaining unit for obtaining a current market indicator and a current transaction indicator of a current transaction market, wherein the current market indicator and the current transaction indicator Both will change continuously with the change of time; an update unit is configured to substitute the current market indicator and the current transaction indicator into a quantitative transaction prediction model, and update corresponding parameters in the quantitative transaction prediction model, and further To ensure the accuracy of the corresponding parameters, the quantitative trading prediction model is obtained by learning and training based on historical failed transaction data.
  • the quantitative trading prediction model is provided with parameters corresponding to the current market index and the current trading index, respectively.
  • a predicting unit configured to obtain the transaction data of the transaction to be tested, input the transaction data to be tested into the updated quantitative transaction prediction model, and determine a failure probability that the transaction data to be tested belongs to the failed transaction data;
  • a determining unit configured to compare the failure probability with a predetermined probability; In contrast, when the probability of failure is greater than or equal to the predetermined probability, then the test failed transaction's data belongs, otherwise to the success of the transaction data.
  • a non-volatile readable storage medium on which computer-readable instructions are stored, and when the computer-readable instructions are executed by a processor, the quantitative transaction according to the first aspect is implemented. method of prediction.
  • a computer device includes a non-volatile readable storage medium and a processor, and the non-volatile readable storage medium is configured to store computer-readable instructions.
  • the processor is configured to execute the computer-readable instructions to implement the quantitative transaction prediction method according to the first aspect.
  • a quantitative transaction prediction method, device, and device provided by the present application can use a quantitative transaction prediction model obtained through learning and training of historical failed transaction behaviors to perform predictive analysis on the user's current transaction behaviors and determine The transaction data to be tested belongs to the failure probability of the failed transaction data, and whether the transaction to be tested is a failed transaction behavior is determined based on the failure probability.
  • This prediction method does not require manual processing, effectively improves the speed of prediction and judgment of the transaction behavior, and quantifies the transaction.
  • the prediction model has the ability to learn and train, and can be updated in time with changes in market transactions, thereby effectively improving the accuracy of trading behavior prediction.
  • FIG. 1 is a flowchart of an embodiment of a quantitative transaction prediction method of the present application
  • FIG. 2 is a structural block diagram of an embodiment of a quantitative transaction prediction apparatus of the present application
  • FIG. 3 is a schematic structural diagram of a quantitative transaction prediction device of the present application.
  • the embodiment of the present application provides a quantitative transaction prediction method, which can learn and train according to historical failed transaction data to obtain a quantitative transaction prediction model, use the quantitative transaction prediction model to predict the user's transaction behavior, and determine whether the transaction data to be tested belongs to The failure probability of the failed transaction data, and whether the transaction to be tested belongs to the failed transaction behavior according to the failure probability, thereby improving the accuracy of the quantified transaction.
  • Quantitative trading refers to the use of advanced mathematical models to replace artificial subjective judgments.
  • Computer technology is used to select large amounts of data to formulate corresponding quantitative trading strategies (that is, the quantitative trading prediction models obtained in this application).
  • Prediction of investors' trading behaviors enables investors to make their own investment decisions based on the predictions, which can greatly reduce the impact of fluctuations in investor sentiment, and in the case of extreme fanaticism or pessimism in the market, investors make non- Rational investment decisions.
  • an embodiment of the present application provides a quantitative transaction prediction method, including:
  • Step 101 Obtain the current market index and the current transaction index of the trading market at the current moment, where both the current market index and the current transaction index will continuously change with time.
  • Market indicators include: stock indicators, business investment indicators, fund indicators, bond indicators, and stock indicators include: Relative Strength Index (RSI, Relative Strength Index), Stochastic Indicator (KDJ, Stochastics Oscillator), and Trend Indicator (DMI, Directional Movement) Index), moving similarity average (MACD, Moving Average / Convergence / Divergence), energy wave (OBV, On Balance Volume), psychological line, deviation rate, etc.
  • Trading indicators include: the profit ratio and profit amount of trading activities.
  • Step 102 Substituting the current market indicator and the current trading indicator into the quantitative trading prediction model, and updating the corresponding parameters in the quantitative trading prediction model, thereby ensuring the accuracy of the quantitative trading prediction model for the quantitative prediction.
  • the quantitative trading prediction model is It is obtained by learning and training based on historical failed transaction data.
  • the quantitative trading prediction model has parameters corresponding to the current market indicator and the current trading indicator, respectively.
  • the learning and training process of the quantitative trading prediction model is as follows: the successful trading behavior is eliminated from a large number of historical trading behaviors, the historical failed trading behaviors are filtered out, and the historical failed trading behaviors are used for learning and training. Determine whether the user's transaction behavior is a failed transaction behavior.
  • Step 103 Obtain the transaction data of the transaction to be tested, input the transaction data to be tested into the updated quantitative transaction prediction model, and determine the failure probability of the transaction data to be failed transaction data.
  • each transaction behavior has corresponding transaction data, for example, purchase time, transaction type, purchase quantity, transaction price, expiration time, and the like.
  • the financial product is the transaction behavior to be tested, so the acquired transaction data is the transaction data of the financial product.
  • use the updated quantitative transaction prediction model obtained above to perform predictive analysis on the transaction data to be tested, and analyze the failure probability that the transaction data to be tested belongs to the failed transaction data, so that it can be determined whether the transaction behavior to be tested is based on the failure probability. Release.
  • Step 104 Compare the failure probability with a predetermined probability. When the failure probability is greater than or equal to the predetermined probability, the transaction data to be tested belongs to the failed transaction data, otherwise it belongs to the successful transaction data.
  • the transaction data to be tested belongs to the failed transaction data, it proves that the risk of the user's transaction to be tested fails in the future market, and the transaction to be tested needs to be intercepted; when the transaction data to be tested belongs to successful transaction data , To prove that the user's test transaction behavior can get good returns in the future market, then release the test transaction behavior. For example, when a user wants to purchase a financial product, the transaction data of the financial product is input into an updated quantitative transaction prediction model. After the updated quantitative transaction prediction model is analyzed by prediction, the probability of failure is 45%. The preset predetermined probability is 50%, 45% ⁇ 50%, which proves that the transaction data of the wealth management product belongs to failed transaction data.
  • the wealth management product needs to be intercepted and the user is advised to give up the purchase of the wealth management product; if the obtained probability of failure It is 78% and 78%> 50%, which proves that the transaction data of the wealth management product belongs to successful transaction data.
  • the wealth management product is released, and the user is recommended to purchase the wealth management product.
  • a quantitative transaction prediction model obtained through learning and training of historical failed trading behaviors can be used to predict and analyze the user's current trading behavior, and the probability that the current trading behavior belongs to a failed transaction can be obtained, so that it can be judged based on the probability Find out whether the user's current trading behavior is a failed transaction, in order to provide users with the correct suggestions based on the judgment results.
  • This prediction method does not require manual processing, effectively improves the speed of trading behavior prediction and judgment, and the quantitative transaction prediction model has the ability to learn and train. , Can be updated in time with changes in market transactions, thereby effectively improving the accuracy of transaction behavior prediction.
  • the method further includes: step 1021, collecting historical transaction behaviors within a predetermined time period, and obtaining historical transaction data of each historical transaction behavior.
  • the predetermined time period may be a previous year, or a previous quarter, or another selected time period (for example, January 1, 2017 to May 8, 2017).
  • Step 1022 Back-test the historical transaction data using a quantitative strategy to obtain back-tested transaction data. The process of backtesting is to take historical transaction data according to a quantitative strategy and use the initial time of a predetermined time period as the start time to simulate transaction buying and selling.
  • the backtest transaction data corresponding to each historical transaction data (for example, Profit rate, risk value, maximum retracement rate), and store these backtested transaction data with the corresponding historical transaction data in a list for easy query and retrieval later.
  • the failed transaction data is filtered from the backtested transaction data. In order to be able to better learn and train failed trading behaviors, the successful transaction data in the backtested transaction data is eliminated, N failed transaction data is selected, and the N failed transaction data is correlated with the corresponding historical trading behavior. correspond.
  • each failed transaction data is input into a random forest model or a logistic regression model for learning and training to obtain a quantitative transaction prediction model.
  • the N failed transaction data obtained above is sequentially input in chronological order or randomly input to the random forest model or logistic regression model for learning training. Each time the failed transaction data is input, the random forest model or logistic regression model is adjusted accordingly. After the last failed transaction data is input, a quantitative transaction prediction model can be obtained.
  • the quantitative transaction prediction model can be used to accurately predict the failure probability of transaction data that belongs to failed transaction data. In this way, whether the transaction behavior belongs to can be determined based on the obtained failure probability. Failed transaction behavior.
  • the acquired historical failed transaction data is divided into M types, then there are M quantified transaction prediction models obtained after learning and training on these M types of historical failed transaction data.
  • M is a positive integer.
  • the corresponding step 102 specifically includes: substituting the current market indicator and the current transaction indicator into M quantitative transaction prediction models, and updating corresponding parameters in the M quantitative transaction prediction models.
  • the corresponding step 103 specifically includes: determining the failure probability of the transaction data to be tested belonging to various types of failed transaction data, each updated quantitative transaction model corresponds to a failure probability, and the number of obtained failure probabilities is M .
  • the corresponding step 104 specifically includes: comparing the M failure probabilities with the predetermined probability one by one. When any of the M failure probabilities is greater than or equal to the predetermined probability, the transaction data to be tested belongs to the failed transaction data. Otherwise it belongs to successful transaction data.
  • the transaction data to be tested is input into each quantitative transaction prediction model, and then the failure probability of the transaction data to be classified into various types of failed transactions is obtained.
  • a quantitative transaction prediction model corresponds to a category, and a corresponding category is predicted. Probability of failure, so there are M failure probabilities. If only one of the M failure probabilities has a failure probability value greater than or equal to a predetermined probability (for example, 40%), then the transaction data to be tested belongs to the failed transaction data and the user is recommended to abandon the transaction behavior.
  • a predetermined probability for example, 40%
  • the failure transaction category may be one or more. The greater the number of failed transaction categories, the higher the risk of failure of the transaction to be tested.
  • a user wants to purchase a certain type of futures (that is, trading behavior), obtain transaction data such as the buying time, selling time, and interest rate of the futures, and enter these transaction data into the quantification corresponding to type 1-type M failed transaction data.
  • the failure probability that the transaction data belongs to various types of failed transaction data is as follows: the failure probability that belongs to type 1 failed transaction data is 23%; the failure probability that belongs to type 2 failed transaction data is 67%; The probability of failure of M-type failed transaction data is 87%.
  • predetermined probability 50%
  • predetermined probability 50%
  • M the probability of Type 2
  • different predetermined probabilities can be set for various types of failed transactions. The specific values can be set according to the type and actual situation of failed transactions. The probability obtained in this way needs to be compared with the predetermined probability of the category corresponding to the quantitative transaction prediction model, and then it is determined whether the transaction to be tested belongs to a failed transaction of that category.
  • Step 1024 specifically includes step 1041, which uses an unsupervised classification algorithm to classify failed transaction data to obtain M-type failed transaction data.
  • the unsupervised classification algorithm can be used for classification based on the natural clustering of failed transaction data, and the number M after classification can be changed according to the actual clustering.
  • the unsupervised classification algorithm can classify while learning, find the same class through learning, and then distinguish this class from other classes to complete the classification of failed transaction data.
  • Step 1042 Substituting the obtained market index and corresponding transaction index at each moment in the predetermined time period into a random forest model or a logistic regression model to generate an initial model of quantitative transaction, wherein the initial model of quantitative transaction is generated separately from the market index and transaction index Corresponding parameter. Market indicators and trading indicators may change at different times.
  • the market indicators and corresponding trading indicators at each time and each time are used as feature factors and substituted into random forest models or logistic regression models to generate separate Corresponding parameters of market indicators and trading indicators to obtain the initial model of quantitative trading.
  • the M-type failed transaction data is learned and trained on the quantitative transaction initial model according to the category, and corresponding M quantitative transaction prediction models are obtained.
  • the type M failed transaction data is divided into M failed transaction categories according to the category. The obtained initial model of quantitative transaction is learned and trained, and each type of failed transaction data is trained to obtain a quantitative transaction prediction model.
  • each quantitative transaction model When the transaction data is input into each quantitative transaction model for predictive analysis, each quantitative transaction model will output the failure probability of the failed transaction data in which the transaction data belongs to the corresponding category. In this way, the respective output failure probabilities can be compared with corresponding predetermined probabilities, and then the category of failed transaction data to which the user's transaction data belongs can be determined.
  • multiple quantitative transaction prediction models capable of identifying each failed transaction category are obtained, so that the accuracy of the quantitative transaction prediction model in predicting and analyzing the transaction behavior can be effectively improved, and it can not only identify whether the transaction behavior is a failed transaction, but also Can identify the category of the corresponding failed transaction, easy to use.
  • Step 10241 specifically includes: step 102411, formulating a stock price fluctuation curve for each failed transaction data from the time of buying to the time of selling.
  • Step 102212 calculating stock price data characteristics of each stock price fluctuation curve.
  • the characteristics of stock price data include: average value, volatility, maximum volatility and minimum volatility.
  • Step 102413 Based on the characteristics of the stock price data, the K-means unsupervised classification algorithm is used to classify the failed transaction data to obtain M-type failed transaction data.
  • the K-means unsupervised classification algorithm is used to classify the failed transaction data to obtain M-type failed transaction data.
  • Step 1023 specifically includes: Step 1023A, obtaining profit data in the backtested transaction data, and determining the backtested transaction data whose profit data is negative within a predetermined period of time as failed transaction data.
  • step 1023B a double moving average is prepared for each backtesting transaction data, and the backtesting transaction data in the double moving average that neither meets the buying conditions nor the selling conditions within a predetermined period of time is determined as failed transaction data, where, when The backtested transaction data meets the buy condition when the double average has a golden fork, and the backtested transaction data meets the sell condition when the double average has a dead fork.
  • the double moving average golden fork mainly refers to the cross of the stock market indicator that crosses the long-term line upwards, which is called the golden fork.
  • the short-term line of the market indicator crosses the cross of the long-term lines downwards, which is called the death fork.
  • the double moving average appears a golden fork, it means that the stock is very strong and meets the trading conditions for buying; otherwise, when the double moving average appears a dead fork, it meets the selling trading conditions.
  • two schemes are used to filter out failed transaction data from backtested transaction data, which is convenient for learning and training for failed transaction data.
  • Step 1022 specifically includes: Step 10221, determining a backtest period corresponding to the historical transaction data.
  • the specific process is: setting stock index combinations for stock selection based on the stock strategy, entering the stock investment logic formed by investors based on experience into the computer, and according to the Stock investment logic simulates the rules of real financial market transactions during the backtest period.
  • the specific process is: input the logic of the income change formed by the income situation corresponding to the various transactions in the arbitrage strategy into the computer to obtain the historical transaction data.
  • Another embodiment of the present application proposes a method for predicting pocket transactions.
  • the steps include:
  • the specific backtesting process is: setting stock index combinations, based on historical transaction data, and selecting a backtesting time period (such as January to June 2017) to simulate trading buy and sell based on the historical transaction data. Out, get backtesting transaction data such as profit rate and maximum retracement rate of the transaction during the backtesting period.
  • the trading mode of historical transaction data is stock trading, backtest the historical transaction data according to the stock strategy.
  • Set stock index combinations for stock selection according to the stock strategy input the stock investment logic formed by investors based on experience into the computer, and use the stock investment logic to simulate real financial market trading rules during backtesting periods for stocks in historical transaction data
  • the transaction is bought and sold, and backtesting transaction data such as profit rate and maximum retracement rate of the transaction during the backtesting period are obtained.
  • back-test historical transaction data in accordance with macro strategies. According to the price change law in the macro strategy, the price change logic formed is input into the computer to obtain the transaction price in the historical transaction data, and the transaction price is simulated in the back test period according to the price change logic to obtain the history in the back test period.
  • the price-income interest rate (that is, profit rate) corresponding to the transaction data is used as the back-tested transaction data.
  • the real financial market is simulated for trading, and the return rate (ie, profitability) corresponding to the historical transaction data is obtained as the backtesting transaction data.
  • the screening process is as follows: classify the backtested transaction data: First, obtain the profitability of the backtested transaction data within a predetermined period of time, determine the backtested transaction data that results in profit as successful transaction data, and cause a backtested loss.
  • the transaction data is determined as failed transaction data (the number of failed transaction data is N, and N is a positive integer).
  • the second is to formulate a double moving average for each backtested transaction data.
  • a double cross appears in the golden cross, it means that the stock is very strong and meets the conditions for buying; otherwise, when a double cross appears a dead fork, it meets the selling conditions for selling.
  • the backtest transaction data that satisfies the buying conditions and / or the selling conditions within a predetermined time is determined as the successful transaction data, and the backtest transaction data that neither meets the buying conditions or the selling conditions within the predetermined time is determined as Failed transaction data (the number of failed transaction data is N, N is a positive integer).
  • the double moving average golden fork mainly refers to the cross of the stock market indicator that crosses the long-term line upwards, which is called the golden fork.
  • the short-term line of the market indicator crosses the cross of the long-term lines downwards, which is called the death fork.
  • the screened successful transaction data is eliminated, and the remaining N failed transaction data are used as the training set.
  • the learning and training process is: for the N failed transaction data during the backtesting process, formulate the stock price fluctuation curve from the time of buying to selling and calculate the corresponding stock price data characteristics (for example, mean, volatility, maximum fluctuations) , Minimum fluctuation range, etc.); based on the above-mentioned stock price data characteristics, a k-means unsupervised classification algorithm is selected to classify the N failed transaction data into M different types of failed transaction data.
  • stock price data characteristics for example, mean, volatility, maximum fluctuations
  • Minimum fluctuation range etc.
  • the market index and transaction index of the user's current transaction behavior as the characteristic factors of the quantitative transaction prediction model, predict and judge the transaction data of the user's current transaction behavior, and determine that the user's current transaction data belongs to M different types of failed transactions.
  • the probability of the data If the probability of the failed transaction data of the M types obtained above is lower than a certain threshold, the model believes that the current transaction behavior of the user will not cause the transaction to fail, and the transaction behavior will be released; otherwise, the corresponding intelligent interception is recommended. Abandon the transaction.
  • an embodiment of the present application provides a quantitative transaction prediction apparatus, which includes an obtaining unit 21, an updating unit 22, a prediction unit 23, and a determination unit 24.
  • the obtaining unit 21 is used to obtain the current market index and the current trading index of the trading market at the current moment. Among them, the current market index and the current trading index are constantly changing with time.
  • the updating unit 22 is used to update the current market index. Substitute the current trading indicator into the quantitative trading prediction model and update the corresponding parameters in the quantitative trading prediction model to ensure the accuracy of the quantitative trading prediction model.
  • the quantitative trading prediction model is learned based on historical failed transaction data. Obtained through training.
  • the quantitative trading prediction model is provided with parameters corresponding to the current market index and the current trading index.
  • the prediction unit 23 is used to obtain the test transaction data of the test transaction behavior, and input the test transaction data into the updated
  • the quantitative transaction prediction model performs prediction to determine the failure probability of the transaction data to be tested; the determination unit 24 is used to compare the failure probability with a predetermined probability, and when the failure probability is greater than or equal to the predetermined probability, the transaction data to be tested Failed transaction data, no It belongs to a successful transaction data.
  • the device further includes: a collecting unit for collecting historical transaction behaviors within a predetermined period of time and obtaining historical transaction data for each historical transaction behavior; a backtesting unit for performing historical transaction data using a quantitative strategy Backtesting to get backtested transaction data; screening unit for screening failed transaction data from backtested transaction data; training unit for inputting each failed transaction data into a random forest model or logistic regression model for learning training Quantitative trading prediction model.
  • the M type of historical failed transaction data corresponds to one of the M quantitative transaction prediction models, where M is a positive integer; then the update unit 22 Specifically used for: substituting the current market indicator and the current trading indicator into M quantitative trading prediction models, and updating corresponding parameters in the M quantitative trading prediction models; the prediction unit 23 is specifically configured to: obtain the untested trading behavior to be tested Transaction data, using M updated quantitative transaction prediction models to predict the transaction data to be tested separately; determine the failure probability of the transaction data to be tested belonging to various types of failed transaction data, and each updated quantitative transaction model corresponds to a failure probability , The number of obtained failure probabilities is M; the determining unit 24 is specifically configured to compare the M failure probabilities with a predetermined probability one by one, and when any of the M failure probabilities is greater than or equal to the predetermined probability, the test is to be performed Transaction data belongs to failed transaction data, otherwise it belongs to successful transaction data.
  • the training unit specifically includes: a classification module for classifying failed transaction data using an unsupervised classification algorithm to obtain M-type failed transaction data, where M is a positive integer; a substitution module is used to substitute the obtained Market indicators and corresponding transaction indicators at various times within a predetermined period of time are substituted into a random forest model or a logistic regression model to generate an initial model of quantitative trading, wherein the initial models of quantitative transactions generate parameters corresponding to market indicators and transaction indicators, respectively; training The module is used for learning and training the initial model of quantitative transaction according to the category of the failed transaction data of type M to obtain corresponding M quantitative transaction prediction models.
  • the classification module specifically includes: a curve formulation module for formulating a stock price fluctuation curve for a period of time from buying to selling for each failed transaction data; a calculation module for calculating a stock price fluctuation curve Features of stock price data; The calculation module is also used to classify failed transaction data using the K-means unsupervised classification algorithm based on the characteristics of stock price data to obtain M-type failed transaction data.
  • the screening unit is specifically configured to: obtain profit data in backtested transaction data, and determine backtested transaction data with negative profit data within a predetermined period of time as failed transaction data; or, for each transaction, Measure the transaction data to formulate a double moving average, and determine the backtest transaction data of the double moving average that neither meets the buying conditions nor the selling conditions within a predetermined period of time as failed transaction data, of which, when the double average line appears a golden fork, backtest The transaction data meets the buying conditions. When the double moving average appears a dead fork, the backtesting transaction data meets the selling conditions.
  • the backtesting unit is specifically configured to: determine a backtesting period corresponding to historical transaction data; and within the backtesting period, use a stock strategy, and / or a macro strategy, and / or an arbitrage strategy to perform historical trading Simulate buying and selling backtesting to get backtesting transaction data during the backtesting period.
  • an embodiment of the present application further provides a non-volatile readable storage medium storing computer-readable instructions, which are implemented when the computer-readable instructions are executed by a processor.
  • the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), including several The instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of this application.
  • a computer device which may be a personal computer, a server, or a network device, etc.
  • an embodiment of the present application further provides a computer device, as shown in FIG. 3, including a non-volatile readable storage medium 32 And the processor 31, wherein the non-volatile readable storage medium 32 and the processor 31 are both disposed on the bus 33.
  • a non-volatile readable storage medium 32 is configured to store computer-readable instructions;
  • a processor 21 is configured to execute the computer-readable instructions to implement the foregoing quantitative transaction prediction method shown in FIG. 1.
  • the device can also be connected to a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, and so on.
  • the user interface may include a display, an input unit such as a keyboard, etc.
  • the optional user interface may also include a USB interface, a card reader interface, and the like.
  • the network interface may optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a WI-FI interface), and the like.
  • a computer device does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or arrange different components.
  • the non-volatile readable storage medium may further include an operating system and a network communication module.
  • An operating system is a program that manages the hardware and software resources of a computer device and supports the operation of computer-readable instructions for information processing and other software and / or computer-readable instructions.
  • the network communication module is used to implement communication between components in the non-volatile readable storage medium, and communication with other hardware and software in the computer device.
  • a quantitative transaction prediction model obtained through learning and training of historical failed trading behaviors can be used to predict and analyze the user's current transaction behaviors, so that it can directly determine whether the user's current transaction behaviors are failed transactions.
  • This forecasting method does not require manual processing, effectively improving the speed of trading behavior prediction and judgment, and the quantitative trading prediction model has the ability to learn and train, which can be updated in time with changes in market transactions, thereby effectively improving the accuracy of trading behavior prediction.

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Technology Law (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

一种量化交易预测方法、装置及设备，其中方法包括：获取当前时刻交易市场的当前市场指标和当前交易指标（101）；将当前市场指标和当前交易指标代入量化交易预测模型，对量化交易预测模型中相对应的参数进行更新，进而保证量化交易预测模型进行量化预测的准确性（102）；获取待测交易行为的待测交易数据，将待测交易数据输入更新后的量化交易预测模型进行预测，确定出待测交易数据属于失败交易数据的失败概率（103），并根据该失败概率判断该待测交易行为是否属于失败交易行为。该方法能够有效提高交易行为预测判断的速度，并且量化交易预测模型具有学习训练的能力，能够随着市场交易的变化及时更新，进而有效提高交易行为预测的准确率。

Description

一种量化交易预测方法、装置及设备
本申请要求与2018年08月01日提交中国专利局、申请号为2018108642029、申请名称为“一种量化交易预测方法、装置及设备”的中国专利申请的优先权，其全部内容通过引用结合在申请中。
技术领域
本申请涉及金融交易技术领域，特别是涉及一种量化交易预测方法、装置及设备。
背景技术
随着人们生活水平的提高，人们手中积累的资产越来越多，有些人会将这些资产放在银行中，但是银行的收益利率非常的低，因此现在很多人选择将资产进行金融投资交易。
由于金融投资交易的收益情况不固定，为了能够更好的帮助投资者进行理财投资，很多理财公司使用量化交易的方式为投资者进行投资检测。量化交易是指以先进的数学模型替代人为的主观判断，利用计算机技术从庞大的历史数据中海选能带来超额收益的多种“大概率”事件以制定策略，极大地减少了投资者情绪波动的影响，避免在市场极度狂热或悲观的情况下作出非理性的投资决策。
但是目前的量化交易的准确率还比较低，对于失败的量化交易只能人工分析和查询，进而来确定是否存在改进方案或者不合理的交易，然后再通过具体的策略代码进行修正，然而上述情况往往会造成过拟合的效果，而且对于量化策略的微调会带来负面的效果和影响，很难量化最终的效果。
发明内容
有鉴于此，本申请提供了一种量化交易预测方法、装置及设备，主要目的在于解决目前对于失败的量化交易只能进行人工分析和查询，使得量化交易的准确率较低的问题。
依据本申请的第一方面，提供了一种量化交易预测方法，所述方法包括：获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动；将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，进而保证所述相对应的参数的准确性，所述量化交易预测模型是根据历史失败交易数据进行学习训练获得，所述量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数；获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率；将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
依据本申请的第二方面，提供了一种量化交易预测装置，所述装置包括：获取单元，用于获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动；更新单元，用于将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，进而保证所述相对应的参数的准确性，所述量化交易预测模型是根据历史失败交易数据进行学习训练获得，所述量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数；预测单元，用于获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率；确定单元，用于将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
依据本申请的第三方面，提供了一种非易失性可读存储介质，其上存储有计算机可读指令，所述计算机可读指令被处理器执行时实现第一方面所述的量化交易预测方法。
依据本申请的第四方面，提供了一种计算机设备，所述计算机设备包括非易失性可读存储介质和处理器，所述非易失性可读存储介质，用于存储计算机可读指令；所述处理器，用于执行所述计算机可读指令以实现第一方面所述的量化交易预测方法。
借由上述技术方案，本申请提供的一种量化交易预测方法、装置及设备，能够利用经过历史失败交易行为进行学习训练得到的量化交易预测模型，对用户的当前交易行为进行预测分析，确定出该待测交易数据属于失败交易数据的失败概率，并根据该失败概率判断该待测交易行为是否属于失败交易行为，这种预测方式无需人工处理，有效提高交易行为预测判断的速度，并且量化交易预测模型具有学习训练的能力，能够随着市场交易的变化及时更新，进而有效提高交易行为预测的准确率。
上述说明仅是本申请技术方案的概述，为了能够更清楚了解本申请的技术手段，而可依照说明书的内容予以实施，并且为了让本申请的上述和其它目的、特征和优点能够更明显易懂，以下特举本申请的具体实施方式。
附图说明
通过阅读下文优选实施方式的详细描述，各种其他的优点和益处对于本领域普通技术人员将变得清楚明了。附图仅用于示出优选实施方式的目的，而并不认为是对本申请的限制。而且在整个附图中，用相同的参考符号表示相同的部件。在附图中：
图1为本申请的量化交易预测方法的一个实施例的流程图；
图2为本申请的量化交易预测装置的一个实施例的结构框图；
图3为本申请的量化交易预测设备的结构示意图。
具体实施方式
下面将参照附图更详细地描述本公开的示例性实施例。虽然附图中显示了本公开的示例性实施例，然而应当理解，可以以各种形式实现本公开而不应被这里阐述的实施例所限制。相反，提供这些实施例是为了能够更透彻地理解本公开，并且能够将本公开的范围完整的传达给本领域的技术人员。
本申请实施例提供了一种量化交易预测方法，能够根据历史失败交易数据进行学习训练，得到量化交易预测模型，利用量化交易预测模型对用户的交易行为进行预测，确定出判断待测交易数据属于失败交易数据的失败概率，并根据该失败概率判断该待测交易行为是否属于失败交易行为，进而提高量化交易的准确率。
量化交易是指以先进的数学模型替代人为的主观判断，利用计算机技术对大量的数据进行海选以制定相应的量化交易策略(即本申请得到的量化交易预测模型)，并根据该量化交易策略对投资者的交易行为进行预测，使投资者能够根据预测结果决定自己的投资决策，能够极大地减少了由于投资者情绪波动的影响，以及在市场极度狂热或悲观的情况下，投资者作出非理性的投资决策。
如图1所示，本申请实施例提供了一种量化交易预测方法，包括：
步骤101，获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动。市场指标包括：股票指标、商业投资指标、基金指标、债券指标，股票指标又包括：相对强弱指标(RSI，Relative Strength Index)、随机指标(KDJ，Stochastics oscillator)、趋向指标(DMI，Directional Movement Index)、平滑异同平均线(MACD，MovingAverage Convergence/Divergence)、能量潮(OBV，On BalanceVolume)、心理线、乖离率等。交易指标包括：交易行为的盈利比例、盈利金额等。
步骤102，将当前市场指标和当前交易指标代入量化交易预测模型，对量化交易预测模型中相对应的参数进行更新，进而保证量化交易预测模型进行量化预测的准确性，其中，量化交易预测模型是根据历史失败交易数据进行学习训练获得，量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数。在上述技术方案中，由于市场指标和交易指标会随着时间的变动而产生变化，因此为了提高对用户交易行为(即，待测交易行为)预测的准确率，需要将当前时刻交易市场的当前市场指标和当前交易指标作为量化交易预测模型的基准，对量化交易模型中与当前市场指标和当前交易指标相对应的参数进行更新。这样，在利用更新后的量化交易预测模型来对用户的交易行为进行预测时，能够有效避免量化交易预测模型使用过旧的市场指标和交易指标产生预测的准确率降低的情况。量化交易预测模型的学习训练过程为：从大量的历史交易行为中将成功的交易行为剔除， 筛选出历史失败交易行为，并利用这些历史失败交易行为进行学习训练，这样得到的量化交易预测模型能够判断用户的交易行为是否是失败的交易行为。
步骤103，获取待测交易行为的待测交易数据，将待测交易数据输入更新后的量化交易预测模型进行预测，确定出待测交易数据属于失败交易数据的失败概率。
在上述方案中，每个交易行为都有对应的交易数据，例如，买入时间、交易种类、购买数量、交易价格、到期时间等。当用户想要购买某个理财产品时，该理财产品即是待测交易行为，这样获取到的待测交易数据就是该理财产品的交易数据。然后利用上述得到的更新后的量化交易预测模型对待测交易数据进行预测分析，分析出该待测交易数据属于失败交易数据的失败概率，这样就可以根据该失败概率确定是否对该待测交易行为放行。
具体确定过程如下：
步骤104，将失败概率与预定概率进行对比，当失败概率大于等于预定概率时，则待测交易数据属于失败交易数据，否则属于成功交易数据。
当该待测交易数据属于失败交易数据时，证明用户的待测交易行为在未来市场中失败的风险比较高，需要对该待测交易行为进行拦截；当该待测交易数据属于成功交易数据时，证明用户的待测交易行为在未来市场中能够得到良好的收益，则对该待测交易行为进行放行。例如，当用户想要购买某个理财产品时，将该理财产品的交易数据输入更新后的量化交易预测模型，经过更新后的量化交易预测模型的预测分析之后，得到的失败概率是45％，预先设定的预定概率是50％，45％＜50％，证明该理财产品的交易数据属于失败交易数据，需要对该理财产品进行拦截，并建议用户放弃购买该理财产品；若得到的失败概率是78％，78％＞50％，证明该理财产品的交易数据属于成功交易数据，将该理财产品进行放行，并建议用户购买该理财产品。通过上述技术方案，能够利用经过历史失败交易行为进行学习训练得到的量化交易预测模型，对用户的当前交易行为进行预测分析，得出该当前交易行为属于失败交易的概率，这样可以根据该概率判断出用户当前的交易行为是否是失败的交易，以便根据判断结果给用户提供正确的建议，这种预测方式无需人工处理，有效提高交易行为预测判断的速度，并且量化交易预测模型具有学习训练的能力，能够随着市场交易的变化及时更新，进而有效提高交易行为预测的准确率。
在步骤102之前，方法还包括：步骤1021，收集预定时间段内的历史交易行为，并获取各个历史交易行为的历史交易数据。其中，该预定时间段可以是上一年度、或者上一季度、或者其他选定时间段(例如，2017年1月1日至2017年5月8日)。步骤1022，利用量化策略对历史交易数据进行回测，得到回测交易数据。回测的过程就是，将历史交易数据按照量化策略以预定时间段的初始时间作为开始时间，模拟交易买入和卖出，模拟完 成后得到每个历史交易数据对应的回测交易数据(例如，盈利率、风险值、最大回撤率)，并将得到的这些回测交易数据与对应的历史交易数据进行列表存储，便于后期的查询和调取。步骤1023，从回测交易数据中筛选出失败交易数据。为了能够更好的对失败的交易行为进行学习和训练，将回测交易数据中成功的交易数据剔除出去，选出N个失败交易数据，并将N个失败交易数据与相应的历史交易行为进行对应。步骤1024，将各个失败交易数据输入随机森林模型或逻辑回归模型中进行学习训练，得到量化交易预测模型。将上述得到的N个失败交易数据，按照时间顺序依次输入或者随机输入随机森林模型或逻辑回归模型进行学习训练，每输入一个失败交易数据，就对随机森林模型或逻辑回归模型进行相应的调整一次，使得输入最后一个失败交易数据之后，能够得到量化交易预测模型，利用量化交易预测模型能够准确预测交易数据属于失败交易数据的失败概率，这样，就可以根据得到的失败概率判断该交易行为是否属于失败交易行为。通过上述技术方案，减小了对历史交易数据中的失败交易数据进行人工筛选和分析时产生的误差，有效提高对交易行为进行预测的准确率。
为了能够更加精细准确地对交易行为进行预测分析，将获取到的历史失败交易数据划分为M类，则对这M类历史失败交易数据进行学习训练后得到的量化交易预测模型有M个，其中，M为正整数。
根据上述描述，对应的步骤102具体包括：将当前市场指标和当前交易指标代入M个量化交易预测模型，对M个量化交易预测模型中相对应的参数进行更新。
根据上述描述，对应的步骤103具体包括：确定出待测交易数据属于各个类型的失败交易数据的失败概率，每个更新后的量化交易模型对应一个失败概率，得到的失败概率的数量为M个。
根据上述描述，对应的步骤104具体包括：将M个失败概率与预定概率进行一一对比，当M个失败概率中任一失败概率大于等于预定概率时，则待测交易数据属于失败交易数据，否则属于成功交易数据。
在上述技术方案中，将待测交易数据分别输入各个量化交易预测模型，进而得到待测交易数据属于各类失败交易的失败概率，一个量化交易预测模型对应一个类别，并预测得出一个对应类别的失败概率，因此得到的失败概率有M个。如果这M个失败概率中只要有一个失败概率值大于等于预定概率(例如，40％)，则证明该待测交易数据属于失败交易数据建议用户放弃该交易行为。通过上述还可以确定出有哪些类别的失败概率大于等于预定概率，进而确定出该待测交易数据的失败交易类别，该失败交易类别可以是一个或多个。失败交易类别数量越多证明该待测交易行为的失败风险越高。
例如，用户想要购买某类期货(即，交易行为)，获取该期货的买入时间、卖出时间、收益利率等交易数据，将这些交易数据输入1类-M类失败交易数据对应的量化交易预测模型中后，得到该交易数据属于各类失败交易数据的失败概率如下：属于1类失败交易数据的失败概率是23％；属于2类失败交易数据的失败概率是67％；……属于M类失败交易数据的失败概率是87％。则设置的预定概率为50％，则将上述失败概率与该预定概率进行比对，得出2类、5类、M类的概率大于50％，则确定该购买期货的行为属于2类、5类、M类对应的失败交易行为，对该购买期货的行为进行拦截，并告知用户不要购买该期货。另外，还可以为各类失败交易设定不同的预定概率，具体数值可以根据失败交易的种类和实际情况进行设定。这样得到的概率需要与对应量化交易预测模型的类别的预定概率进行比对，然后再确定待测交易行为是否属于该类别的失败交易。
步骤1024具体包括：步骤10241，利用无监督分类算法对失败交易数据进行分类，得到M类失败交易数据。利用无监督分类算法能够根据失败交易数据的自然聚类情况进行分类，分类后的数量M可以根据实际聚类情况进行变动。无监督分类算法能够一边学习一边分类，通过学习找到相同的类别，然后将该类与其它类区分开，进而完成对失败交易数据的分类。步骤10242，将获取的预定时间段内各个时刻的市场指标和对应的交易指标代入随机森林模型或逻辑回归模型，生成量化交易初始模型，其中，量化交易初始模型中生成分别与市场指标和交易指标相对应的参数。不同时刻市场指标和交易指标都可能会发生变化，为了保证学习训练的准确性，将各个时刻的各个时刻的市场指标和对应的交易指标作为特征因子，代入随机森林模型或逻辑回归模型，生成分别与市场指标和交易指标相对应的参数，进而获得量化交易初始模型。步骤10243，将M类失败交易数据按照类别对量化交易初始模型进行学习训练，得到对应的M个量化交易预测模型。在上述技术方案中，M类失败交易数据会根据类别分成M个失败交易类别，对得到的量化交易初始模型进行学习训练，每类失败交易数据都会被训练得到一个量化交易预测模型。当将交易数据输入各个量化交易模型进行预测分析，各个量化交易模型就会输出该交易数据属于对应类别的失败交易数据的失败概率。这样就可以将输出的各个失败概率与相应的预定概率进行比对，进而确定用户的交易数据属于哪个类别的失败交易数据。通过上述技术方案，得到能够识别各个失败交易类别的多个量化交易预测模型，使得量化交易预测模型对交易行为进行预测分析的准确率能够有效提高，不仅能够识别出交易行为是否是失败交易，还能识别出对应失败交易的类别，方便使用。
步骤10241具体包括：步骤102411，为各个失败交易数据制定从买入到卖出时间段内的股票价格波动曲线。步骤102412，计算各个股票价格波动曲线的股票价格数据特征。其 中，股票价格数据特征包括：均值、波动率、波动最大幅度以及波动最小幅度等。步骤102413，基于股票价格数据特征，利用K-means无监督分类算法对为失败交易数据进行分类，得到M类失败交易数据。上述技术方案中，每个失败交易数据从买入后，股票价格都会随着时间的变化而变化，制定相应的股票价格波动曲线能够更准确的观察和计算股票价格数据特征，这样使得到的失败交易数据的类别也能更加准确。
步骤1023具体包括：步骤1023A，获取回测交易数据中的盈利数据，并将在预定时间段内盈利数据为负值的回测交易数据确定为失败交易数据。或者，步骤1023B，为各个回测交易数据制定双均线，将双均线中在预定时间段内既不满足买入条件又不满足卖出条件的回测交易数据确定为失败交易数据，其中，当双均线出现金叉时回测交易数据满足买入条件，当双均线出现死叉时回测交易数据满足卖出条件。其中，双均线金叉主要指股票行情指标的短期线向上穿越长期线的交叉，称之为金叉，反之，行情指标的短期线向下穿越长期线的交叉，称之为死叉。当双均线出现金叉时，表示股票很强势，满足买入的交易条件；反之当双均线出现死叉时候，满足卖出的交易条件。通过上述技术方案，利用两种方案从回测交易数据中筛选出失败交易数据，为针对失败交易数据进行学习训练提供便利。
步骤1022具体包括：步骤10221，确定历史交易数据对应的回测时间段。步骤10222，在回测时间段内，利用股票策略、和/或宏观策略、和/或套利策略对历史交易进行模拟买入和卖出回测，得到回测时间段内的回测交易数据。其中，若历史交易数据是股票交易，则利用股票策略进行回测，具体过程为：根据股票策略设定股票指标组合进行选股，将投资人根据经验形成的股票投资逻辑输入计算机，并根据该股票投资逻辑模拟回测时间段内真实金融市场交易的规则对历史交易数据中的股票交易进行买入、卖出，得出回测时间段内交易的盈利率、最大回撤率等回测交易数据。若历史交易数据注重价格的变化，则按照宏观策略进行回测，具体过程为：根据宏观策略中的价格变化规律，形成的价格变化逻辑输入计算机，获取历史交易数据中交易价格，将交易价格根据价格变化逻辑在回测时间段内进行模拟交易，得出回测时间段内历史交易数据对应的价格收益利率(即盈利率)作为回测交易数据。若历史交易数据对应的产品是固定收益类产品，则按照套利策略进行回测，具体过程为：将套利策略中各种交易对应的收益情况形成的收益变化逻辑输入计算机，获取历史交易数据中买入价格、到期时间、利率、通胀比例、信用利差等数据，对这些数据根据收益变化逻辑在回测时间段内模拟真实金融市场进行交易，得出历史交易数据对应的收益率(即盈利率)作为回测交易数据。
本申请的另一个实施例提出了一种零花交易预测方法，步骤包括：
1)通过量化策略对历史交易数据进行回测，得到相应的回测交易数据。
收集近期内各种交易模式下的历史交易数据，利用股票策略、宏观策略或套利策略对这些历史交易数据进行回测。具体回测过程为：设定股票指标组合，基于历史交易数据，在历史交易数据对应的时间内，选定出一个回测时间段(例如2017年1月至6月)模拟交易买入和卖出，得出该回测时间段内交易的盈利率、最大回撤率等回测交易数据。一，若历史交易数据的交易模式是股票交易，则按照股票策略对历史交易数据进行回测。根据股票策略设定股票指标组合进行选股，将投资人根据经验形成的股票投资逻辑输入计算机，并根据该股票投资逻辑模拟回测时间段内真实金融市场交易的规则对历史交易数据中的股票交易进行买入、卖出，得出回测时间段内交易的盈利率、最大回撤率等回测交易数据。二，若历史交易数据的交易模式注重价格的变化，则按照宏观策略对历史交易数据进行回测。根据宏观策略中的价格变化规律，形成的价格变化逻辑输入计算机，获取历史交易数据中交易价格，将交易价格根据价格变化逻辑在回测时间段内进行模拟交易，得出回测时间段内历史交易数据对应的价格收益利率(即盈利率)作为回测交易数据。三，若历史交易数据对应的理财产品是固定收益类产品，则按照套利策略对历史交易数据进行回测。根据套利策略中各种交易对应的收益情况形成的收益变化逻辑输入计算机，获取历史交易数据中买入价格、到期时间、利率、通胀比例、信用利差等数据，对这些数据根据收益变化逻辑在回测时间段内模拟真实金融市场进行交易，得出历史交易数据对应的收益率(即盈利率)作为回测交易数据。
2)基于上述回测交易数据筛选训练集，并进行学习和训练得到量化交易预测模型。
筛选过程为：将回测交易数据进行分类：其一，获取回测交易数据中在预定时间段内的盈利情况，将造成盈利的回测交易数据确定为成功交易数据，将造成亏损的回测交易数据确定为失败交易数据(失败交易数据数量为N，N为正整数)。或者，其二，为各个回测交易数据制定双均线，当双均线出现金叉时候，表示股票很强势，满足买入的条件；反之当双均线出现死叉时候，满足卖出的交易条件。将在预定时间内满足买入条件和/或卖出条件的回测交易数据确定为成功交易数据，将在预定时间内既不满足买入条件也不满足卖出条件的回测交易数据确定为失败交易数据(失败交易数据数量为N，N为正整数)。其中，双均线金叉主要指股票行情指标的短期线向上穿越长期线的交叉，称之为金叉，反之，行情指标的短期线向下穿越长期线的交叉，称之为死叉。将上述筛选出来的成功交易数据剔除，剩余的N个失败交易数据作为训练集。学习训练过程为：针对回测过程中的N个失败交易数据，制定从买入到卖出时间阶段的股票价格波动曲线并计算相应的股票价格数据特征(例如，均值、波动率、波动最大幅度、波动最小幅度等)；基于上述股票价格数据特征，选择k-means无监督分类算法将上述N个失败交易数据进行分类，分为M种不同类别 的失败交易数据。将上述M种不同类别的失败交易数据作为有监督学习的标签label，并将市场指标相应时刻的各个交易指标作为学习训练的特征因子，然后再将每个类别的失败交易数据输入随机森林模型或者逻辑回归模型中进而构建有监督多分类的量化交易预测模型(每个类别的失败交易数据经过学习训练后都得到一个量化交易预测模型，则量化交易预测模型有M个)。
3)利用预测失败交易行为模型对用户的交易行为进行预测。
获取用户当前的交易行为的市场指标和交易指标作为量化交易预测模型的特征因子，对用户当前的交易行为的交易数据进行预测和判断，确定出用户当前的交易数据属于M种不同类别的失败交易数据的概率。如果上述得到的M种不同类别的失败交易数据的概率都低于一定阈值，则模型认为上述用户当前的交易行为不会导致交易失败，放行该交易行为；反之，则进行相应的智能拦截，建议放弃该交易行为。
进一步的，作为图1方法的具体实现，本申请实施例提供了一种量化交易预测装置，装置包括：获取单元21、更新单元22、预测单元23和确定单元24。获取单元21，用于获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动；更新单元22，用于将当前市场指标和当前交易指标代入量化交易预测模型，对量化交易预测模型中相对应的参数进行更新，进而保证量化交易预测模型进行量化预测的准确性，其中，量化交易预测模型是根据历史失败交易数据进行学习训练获得，量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数；预测单元23，用于获取待测交易行为的待测交易数据，将待测交易数据输入更新后的量化交易预测模型进行预测，确定出待测交易数据属于失败交易数据的失败概率；确定单元24，用于将失败概率与预定概率进行对比，当失败概率大于等于预定概率时，则待测交易数据属于失败交易数据，否则属于成功交易数据。
在具体实施例中，装置还包括：收集单元，用于收集预定时间段内的历史交易行为，并获取各个历史交易行为的历史交易数据；回测单元，用于利用量化策略对历史交易数据进行回测，得到回测交易数据；筛选单元，用于从回测交易数据中筛选出失败交易数据；训练单元，用于将各个失败交易数据输入随机森林模型或逻辑回归模型中进行学习训练，得到量化交易预测模型。
在具体实施例中，历史失败交易数据存在M类，量化交易预测模型有M个，M类历史失败交易数据与M个量化交易预测模型一一对应，其中，M为正整数；则更新单元22具体用于：将当前市场指标和当前交易指标代入M个量化交易预测模型，对M个量化交易预测模型中相对应的参数进行更新；预测单元23具体用于：获取待测交易行为的待测交易 数据，利用M个更新后的量化交易预测模型分别对待测交易数据进行预测；确定出待测交易数据属于各个类型的失败交易数据的失败概率，每个更新后的量化交易模型对应一个失败概率，得到的失败概率的数量为M个；确定单元24具体用于：将M个失败概率与预定概率进行一一对比，当M个失败概率中任一失败概率大于等于预定概率时，则待测交易数据属于失败交易数据，否则属于成功交易数据。
在具体实施例中，训练单元具体包括：分类模块，用于利用无监督分类算法对失败交易数据进行分类，得到M类失败交易数据，其中，M为正整数；代入模块，用于将获取的预定时间段内各个时刻的市场指标和对应的交易指标代入随机森林模型或逻辑回归模型，生成量化交易初始模型，其中，量化交易初始模型中生成分别与市场指标和交易指标相对应的参数；训练模块，用于将M类失败交易数据按照类别对量化交易初始模型进行学习训练，得到对应的M个量化交易预测模型。
在具体实施例中，分类模块具体包括：曲线制定模块，用于为各个失败交易数据制定从买入到卖出时间段内的股票价格波动曲线；计算模块，用于计算各个股票价格波动曲线的股票价格数据特征；计算模块，还用于基于股票价格数据特征，利用K-means无监督分类算法对为失败交易数据进行分类，得到M类失败交易数据。
在具体实施例中，筛选单元具体用于：获取回测交易数据中的盈利数据，并将在预定时间段内盈利数据为负值的回测交易数据确定为失败交易数据；或者，为各个回测交易数据制定双均线，将双均线中在预定时间段内既不满足买入条件又不满足卖出条件的回测交易数据确定为失败交易数据，其中，当双均线出现金叉时回测交易数据满足买入条件，当双均线出现死叉时回测交易数据满足卖出条件。
在具体实施例中，回测单元具体用于：确定历史交易数据对应的回测时间段；在回测时间段内，利用股票策略、和/或宏观策略、和/或套利策略对历史交易进行模拟买入和卖出回测，得到回测时间段内的回测交易数据。
基于上述如图1所示方法，相应的，本申请实施例还提供了一种非易失性可读存储介质，其上存储有计算机可读指令，该计算机可读指令被处理器执行时实现上述如图1所示的量化交易预测方法。
基于这样的理解，本申请的技术方案可以以软件产品的形式体现出来，该软件产品可以存储在一个非易失性存储介质(可以是CD-ROM，U盘，移动硬盘等)中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施场景所述的方法。
基于上述图1所示方法和图2所示装置的实施例，为了实现上述目的，本申请实施例 还提供了一种计算机设备，如图3所示，包括非易失性可读存储介质32和处理器31，其中非易失性可读存储介质32和处理器31均设置在总线33上。非易失性可读存储介质32，用于存储计算机可读指令；处理器21，用于执行该计算机可读指令以实现上述如图1所示的量化交易预测方法。
可选地，该设备还可以连接用户接口、网络接口、摄像头、射频(Radio Frequency，RF)电路，传感器、音频电路、WI-FI模块等等。用户接口可以包括显示屏(Display)、输入单元比如键盘(Keyboard)等，可选用户接口还可以包括USB接口、读卡器接口等。网络接口可选的可以包括标准的有线接口、无线接口(如蓝牙接口、WI-FI接口)等。
本领域技术人员可以理解，本实施例提供的一种计算机设备的结构并不构成对该实体设备的限定，可以包括更多或更少的部件，或者组合某些部件，或者不同的部件布置。
非易失性可读存储介质中还可以包括操作系统、网络通信模块。操作系统是管理计算机设备硬件和软件资源的程序，支持信息处理计算机可读指令以及其它软件和/或计算机可读指令的运行。网络通信模块用于实现非易失性可读存储介质内部各组件之间的通信，以及与计算机设备中其它硬件和软件之间通信。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到本申请可以借助软件加必要的通用硬件平台的方式来实现，也可以通过硬件实现。
通过应用本申请的技术方案，能够利用经过历史失败交易行为进行学习训练得到的量化交易预测模型，对用户的当前交易行为进行预测分析，这样可以直接判断出用户当前的交易行为是否是失败的交易，这种预测方式无需人工处理，有效提高交易行为预测判断的速度，并且量化交易预测模型具有学习训练的能力，能够随着市场交易的变化及时更新，进而有效提高交易行为预测的准确率。
本领域技术人员可以理解附图只是一个优选实施场景的示意图，附图中的模块或流程并不一定是实施本申请所必须的。本领域技术人员可以理解实施场景中的装置中的模块可以按照实施场景描述进行分布于实施场景的装置中，也可以进行相应变化位于不同于本实施场景的一个或多个装置中。上述实施场景的模块可以合并为一个模块，也可以进一步拆分成多个子模块。
上述本申请序号仅仅为了描述，不代表实施场景的优劣。以上公开的仅为本申请的几个具体实施场景，但是，本申请并非局限于此，任何本领域的技术人员能思之的变化都应落入本申请的保护范围。

Claims (20)

  1. 一种量化交易预测方法，其特征在于，所述方法包括：
    获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动；
    将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，进而保证所述量化交易预测模型进行量化预测的准确性，其中，所述量化交易预测模型是根据历史失败交易数据进行学习训练获得，所述量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数；
    获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率；
    将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
  2. 根据权利要求1所述的方法，其特征在于，在将所述当前市场指标和所述当前交易指标代入量化交易预测模型之前，所述方法还包括：
    收集预定时间段内的历史交易行为，并获取各个历史交易行为的历史交易数据；
    利用量化策略对所述历史交易数据进行回测，得到回测交易数据；
    从所述回测交易数据中筛选出失败交易数据；
    将各个所述失败交易数据输入随机森林模型或逻辑回归模型中进行学习训练，得到量化交易预测模型。
  3. 根据权利要求1或2所述的方法，其特征在于，所述历史失败交易数据存在M类，所述量化交易预测模型有M个，所述M类历史失败交易数据与所述M个量化交易预测模型一一对应，其中，M为正整数；
    所述将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，具体包括：
    将所述当前市场指标和所述当前交易指标代入M个量化交易预测模型，对所述M个量化交易预测模型中相对应的参数进行更新；
    所述获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率，具体包括：
    获取待测交易行为的待测交易数据，利用M个更新后的量化交易预测模型分别对所述待测交易数据进行预测；
    确定出所述待测交易数据属于各个类型的失败交易数据的失败概率，每个更新后的量化交易模型对应一个失败概率，得到的失败概率的数量为M个；
    将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，对所述待测交易行为进行拦截，否则属于成功交易数据，对所述待测交易行为进行放行，具体包括：
    将所述M个失败概率与预定概率进行一一对比，当所述M个失败概率中任一失败概率大于等于预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
  4. 根据权利要求2所述的方法，其特征在于，将各个所述失败交易数据输入随机森林模型或逻辑回归模型中进行学习训练，得到量化交易预测模型，具体包括：
    利用无监督分类算法对所述失败交易数据进行分类，得到M类失败交易数据；
    将获取的预定时间段内各个时刻的市场指标和对应的交易指标代入随机森林模型或逻辑回归模型，生成量化交易初始模型，其中，所述量化交易初始模型中生成分别与市场指标和交易指标相对应的参数；
    将所述M类失败交易数据按照类别对所述量化交易初始模型进行学习训练，得到对应的M个量化交易预测模型。
  5. 根据权利要求4所述的方法，其特征在于，所述利用无监督分类算法对所述失败交易数据进行分类，得到M类失败交易数据，具体包括：
    为各个失败交易数据制定从买入到卖出时间段内的股票价格波动曲线；
    计算各个股票价格波动曲线的股票价格数据特征；
    基于所述股票价格数据特征，利用K-means无监督分类算法对为所述失败交易数据进行分类，得到M类失败交易数据。
  6. 根据权利要求2所述的方法，其特征在于，从所述回测交易数据中筛选出失败交易数据，具体包括：
    获取所述回测交易数据中的盈利数据，并将在所述预定时间段内盈利数据为负值的回测交易数据确定为失败交易数据；或者，
    为各个所述回测交易数据制定双均线，将所述双均线中在所述预定时间段内既不满足买入条件又不满足卖出条件的回测交易数据确定为失败交易数据，其中，当所述双均线出现金叉时所述回测交易数据满足买入条件，当所述双均线出现死叉时所述回测交易数据满足卖出条件。
  7. 根据权利要求2所述的方法，其特征在于，所述利用量化策略对所述历史交易数据进行回测，得到回测交易数据，具体包括：
    确定历史交易数据对应的回测时间段；
    在所述回测时间段内，利用股票策略、和/或宏观策略、和/或套利策略对所述历史交易进行模拟买入和卖出回测，得到所述回测时间段内的回测交易数据。
  8. 一种量化交易预测装置，其特征在于，所述装置包括：
    获取单元，用于获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动；
    更新单元，用于将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，进而保证所述相对应的参数的准确性，所述量化交易预测模型是根据历史失败交易数据进行学习训练获得，所述量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数；
    预测单元，用于获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率；
    确定单元，用于将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
  9. 根据权利要求8所述的装置，其特征在于，还包括：
    收集单元，用于收集预定时间段内的历史交易行为，并获取各个历史交易行为的历史交易数据；
    回测单元，用于利用量化策略对所述历史交易数据进行回测，得到回测交易数据；
    筛选单元，用于从所述回测交易数据中筛选出失败交易数据；
    训练单元，用于将各个所述失败交易数据输入随机森林模型或逻辑回归模型中进行学习训练，得到量化交易预测模型。
  10. 根据权利要求8或9所述的装置，其特征在于，所述历史失败交易数据存在M类，所述量化交易预测模型有M个，所述M类历史失败交易数据与所述M个量化交易预测模型一一对应，其中，M为正整数；
    则所述更新单元具体用于：
    将所述当前市场指标和所述当前交易指标代入M个量化交易预测模型，对所述M个量化交易预测模型中相对应的参数进行更新；
    所述预测单元具体用于：
    所述获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率，具体包括：
    获取待测交易行为的待测交易数据，利用M个更新后的量化交易预测模型分别对所述 待测交易数据进行预测；
    确定出所述待测交易数据属于各个类型的失败交易数据的失败概率，每个更新后的量化交易模型对应一个失败概率，得到的失败概率的数量为M个；
    确定单元24具体用于：
    将所述M个失败概率与预定概率进行一一对比，当所述M个失败概率中任一失败概率大于等于预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
  11. 根据权利要求9所述的装置，其特征在于，所述训练单元具体包括：
    分类模块，用于利用无监督分类算法对所述失败交易数据进行分类，得到M类失败交易数据；
    代入模块，用于将获取的预定时间段内各个时刻的市场指标和对应的交易指标代入随机森林模型或逻辑回归模型，生成量化交易初始模型，其中，所述量化交易初始模型中生成分别与市场指标和交易指标相对应的参数；
    训练模块，用于将所述M类失败交易数据按照类别对所述量化交易初始模型进行学习训练，得到对应的M个量化交易预测模型。
  12. 根据权利要求11所述的装置，其特征在于，所述分类模块具体包括：
    曲线制定模块，用于为各个失败交易数据制定从买入到卖出时间段内的股票价格波动曲线；
    计算模块，用于计算各个股票价格波动曲线的股票价格数据特征；
    所述计算模块，还用于基于所述股票价格数据特征，利用K-means无监督分类算法对为所述失败交易数据进行分类，得到M类失败交易数据。
  13. 根据权利要求9所述的装置，其特征在于，所述筛选单元具体用于：
    获取所述回测交易数据中的盈利数据，并将在所述预定时间段内盈利数据为负值的回测交易数据确定为失败交易数据；或者，
    为各个所述回测交易数据制定双均线，将所述双均线中在所述预定时间段内既不满足买入条件又不满足卖出条件的回测交易数据确定为失败交易数据，其中，当所述双均线出现金叉时所述回测交易数据满足买入条件，当所述双均线出现死叉时所述回测交易数据满足卖出条件。
  14. 根据权利要求9所述的装置，其特征在于，所述回测单元具体用于：
    确定历史交易数据对应的回测时间段；在所述回测时间段内，利用股票策略、和/或宏观策略、和/或套利策略对所述历史交易进行模拟买入和卖出回测，得到所述回测时间段内的回测交易数据。
  15. 一种非易失性可读存储介质，其上存储有计算机可读指令，其特征在于，所述计算机可读指令被处理器执行时实现量化交易预测方法，包括：
    获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动；将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，进而保证所述量化交易预测模型进行量化预测的准确性，其中，所述量化交易预测模型是根据历史失败交易数据进行学习训练获得，所述量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数；获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率；将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
  16. 根据权利要求15所述的非易失性可读存储介质，其特征在于，所述处理器执行所述计算机可读指令时实现所述方法还包括：
    收集预定时间段内的历史交易行为，并获取各个历史交易行为的历史交易数据；利用量化策略对所述历史交易数据进行回测，得到回测交易数据；从所述回测交易数据中筛选出失败交易数据；将各个所述失败交易数据输入随机森林模型或逻辑回归模型中进行学习训练，得到量化交易预测模型。
  17. 根据权利要求15或16所述的非易失性可读存储介质，其特征在于，所述处理器执行所述计算机可读指令时，所述历史失败交易数据存在M类，所述量化交易预测模型有M个，所述M类历史失败交易数据与所述M个量化交易预测模型一一对应，其中，M为正整数；
    所述将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，具体包括：
    将所述当前市场指标和所述当前交易指标代入M个量化交易预测模型，对所述M个量化交易预测模型中相对应的参数进行更新；
    所述获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率，具体包括：
    获取待测交易行为的待测交易数据，利用M个更新后的量化交易预测模型分别对所述待测交易数据进行预测；
    确定出所述待测交易数据属于各个类型的失败交易数据的失败概率，每个更新后的量化交易模型对应一个失败概率，得到的失败概率的数量为M个；
    将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，对所述待测交易行为进行拦截，否则属于成功交易数据，对所述待测交易行为进行放行，具体包括：
    将所述M个失败概率与预定概率进行一一对比，当所述M个失败概率中任一失败概率大于等于预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
  18. 一种计算机设备，其特征在于，所述计算机设备包括非易失性可读存储介质和处理器，所述非易失性可读存储介质，用于存储计算机可读指令；
    所述处理器，执行所述计算机可读指令以实现量化交易预测方法时，包括：
    获取当前时刻交易市场的当前市场指标和当前交易指标，其中，当前市场指标和当前交易指标均会随着时间的变化而不断变动；将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，进而保证所述量化交易预测模型进行量化预测的准确性，其中，所述量化交易预测模型是根据历史失败交易数据进行学习训练获得，所述量化交易预测模型中设有分别与当前市场指标和当前交易指标相对应的参数；获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率；将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
  19. 根据权利要求18所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时实现所述方法还包括：
    收集预定时间段内的历史交易行为，并获取各个历史交易行为的历史交易数据；利用量化策略对所述历史交易数据进行回测，得到回测交易数据；从所述回测交易数据中筛选出失败交易数据；将各个所述失败交易数据输入随机森林模型或逻辑回归模型中进行学习训练，得到量化交易预测模型。
  20. 根据权利要求18或19所述的计算机设备，其特征在于，
    所述处理器执行所述计算机可读指令时，所述历史失败交易数据存在M类，所述量化交易预测模型有M个，所述M类历史失败交易数据与所述M个量化交易预测模型一一对应，其中，M为正整数；
    所述将所述当前市场指标和所述当前交易指标代入量化交易预测模型，对所述量化交易预测模型中相对应的参数进行更新，具体包括：
    将所述当前市场指标和所述当前交易指标代入M个量化交易预测模型，对所述M个量化交易预测模型中相对应的参数进行更新；
    所述获取待测交易行为的待测交易数据，将所述待测交易数据输入更新后的量化交易预测模型进行预测，确定出所述待测交易数据属于失败交易数据的失败概率，具体包括：
    获取待测交易行为的待测交易数据，利用M个更新后的量化交易预测模型分别对所述待测交易数据进行预测；
    确定出所述待测交易数据属于各个类型的失败交易数据的失败概率，每个更新后的量化交易模型对应一个失败概率，得到的失败概率的数量为M个；
    将所述失败概率与预定概率进行对比，当所述失败概率大于等于所述预定概率时，则所述待测交易数据属于失败交易数据，对所述待测交易行为进行拦截，否则属于成功交易数据，对所述待测交易行为进行放行，具体包括：
    将所述M个失败概率与预定概率进行一一对比，当所述M个失败概率中任一失败概率大于等于预定概率时，则所述待测交易数据属于失败交易数据，否则属于成功交易数据。
PCT/CN2018/111689 2018-08-01 2018-10-24 一种量化交易预测方法、装置及设备 WO2020024456A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810864202.9 2018-08-01
CN201810864202.9A CN109345050A (zh) 2018-08-01 2018-08-01 一种量化交易预测方法、装置及设备

Publications (1)

Publication Number Publication Date
WO2020024456A1 true WO2020024456A1 (zh) 2020-02-06

Family

ID=65291450

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/111689 WO2020024456A1 (zh) 2018-08-01 2018-10-24 一种量化交易预测方法、装置及设备

Country Status (2)

Country Link
CN (1) CN109345050A (zh)
WO (1) WO2020024456A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230068032A1 (en) * 2021-09-02 2023-03-02 Amazon Technologies, Inc. Techniques for detecting outages of an external system

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110443707A (zh) * 2019-07-26 2019-11-12 上海量锐信息科技有限公司 一种高频量化交易策略回测验证方法及系统
CN110490671B (zh) * 2019-08-28 2022-08-12 中国银行股份有限公司 量化报价策略模型的测试方法、系统及装置
CN111127084B (zh) * 2019-12-11 2024-06-11 广州优策科技有限公司 交易量预测的方法、装置、可读存储介质及电子设备
CN111427935B (zh) * 2020-02-28 2023-05-30 中信建投证券股份有限公司 量化交易指标的预测和显示方法、电子设备和介质
CN112488831B (zh) * 2020-11-20 2024-08-23 东软集团股份有限公司 区块链网络交易方法、装置、存储介质及电子设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102890803A (zh) * 2011-07-21 2013-01-23 阿里巴巴集团控股有限公司 电子商品异常交易过程的确定方法及其装置
CN105046566A (zh) * 2015-08-28 2015-11-11 苗青 基于风控的量化趋势交易决策系统及方法
CN105761000A (zh) * 2016-02-17 2016-07-13 中国工商银行股份有限公司 交易数据处理及风险预警系统和方法
US20170103459A1 (en) * 2015-10-08 2017-04-13 Sung Hun Kim Financial trading system based on real-time prediction using virtual cash
CN107798609A (zh) * 2017-11-08 2018-03-13 上海宽全智能科技有限公司 量化交易策略生成方法与装置、设备及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102890803A (zh) * 2011-07-21 2013-01-23 阿里巴巴集团控股有限公司 电子商品异常交易过程的确定方法及其装置
CN105046566A (zh) * 2015-08-28 2015-11-11 苗青 基于风控的量化趋势交易决策系统及方法
US20170103459A1 (en) * 2015-10-08 2017-04-13 Sung Hun Kim Financial trading system based on real-time prediction using virtual cash
CN105761000A (zh) * 2016-02-17 2016-07-13 中国工商银行股份有限公司 交易数据处理及风险预警系统和方法
CN107798609A (zh) * 2017-11-08 2018-03-13 上海宽全智能科技有限公司 量化交易策略生成方法与装置、设备及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230068032A1 (en) * 2021-09-02 2023-03-02 Amazon Technologies, Inc. Techniques for detecting outages of an external system
US12045119B2 (en) * 2021-09-02 2024-07-23 Amazon Technologies, Inc. Techniques for detecting outages of an external system

Also Published As

Publication number Publication date
CN109345050A (zh) 2019-02-15

Similar Documents

Publication Publication Date Title
WO2020024456A1 (zh) 一种量化交易预测方法、装置及设备
Huang et al. Enterprise credit risk evaluation based on neural network algorithm
Hsu et al. A Comprehensive performance evaluation and ranking methodology under a sustainable development perspective
CN110009503A (zh) 理财产品推荐方法、装置、计算机设备及存储介质
AU2444101A (en) Methods and systems for finding value and reducing risk
Ereiz Predicting default loans using machine learning (OptiML)
CN111882420A (zh) 响应率的生成方法、营销方法、模型训练方法及装置
CN113642923A (zh) 基于历史催收数据的不良资产包价值评估方法
CN111931992A (zh) 一种电力负荷预测指标选取方法及装置
CN111626855A (zh) 债券信用利差预测方法及系统
CN109978300B (zh) 客户风险承受力量化方法及系统、资产配置方法及系统
CN114493208A (zh) 工程项目全寿命周期的评价方法、装置、电子设备及介质
CN113159634A (zh) 一种金融产品管理方法、装置及电子设备
KR101292130B1 (ko) 기업별 실시간 상대가치 평가 방법
CN117132383A (zh) 一种信贷数据处理方法、装置、设备及可读存储介质
CN112116197A (zh) 一种基于供应商评价体系的不良行为预警方法及系统
CN111489134B (zh) 数据模型的构建方法、装置、设备及计算机可读存储介质
CN114626940A (zh) 数据分析方法、装置及电子设备
TW202213239A (zh) 類神經網路分析多技術指標之股價趨勢預測及交易決策輔助系統
CN113763181A (zh) 一种风险压力测试系统
CN111612626A (zh) 一种债券评估数据预处理方法和装置
Nur Comparing the Accuracy of Multiple Discriminant Analyisis, Logistic Regression, and Neural Network to estimate pay and not to pay Dividend
Mauri et al. LGD Stress Testing: a Bayesian Averaging Modeling Approach
CN117853225A (zh) 一种发债主体信用评价方法
Elwakil Functional expert-based performance assessment models at organizational level

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18928672

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18928672

Country of ref document: EP

Kind code of ref document: A1