Prediction method and device, computer equipment and storage medium

Abstract

The invention discloses a prediction method and device, computer equipment and a storage medium. The method comprises the steps of obtaining feature data of a to-be-predicted business decision scene; wherein the feature data comprises feature sub-data under various combinations, and each combination comprises a subdivision scene and a subdivision dimension involved in the business decision scene; respectively inputting the feature sub-data under the plurality of combinations into the probability distribution models corresponding to the plurality of combinations, and outputting prediction results corresponding to the plurality of combinations; wherein the prediction result is the probability density of a target value, and the target value represents the deviation between the asset prediction value and the asset true value under the combination; and according to the prediction results corresponding to the various combinations, predicting a target floating value of the excess amount of the service decision scene.

Description

Technical Field

[0001] The embodiments of the present specification relate to the field of computer technology, and in particular to a prediction method, apparatus, computer device and storage medium. Background Art

[0002] With the development of computer technology, it is becoming more and more common to achieve capital flow through interbank financing management. The main process of interbank financing management is: to produce future asset forecast results through scientific forecasting models, formulate capital plans based on the forecast results, reserve a certain reserve buffer in the capital plan to cope with unforeseen capital needs, and then use special interbank management tools to assist in the execution of the entire capital management process. However, due to the uncertainty of the forecasting model, the subjective dependence of the reserved reserve buffer, and the high complexity of the model forecasting results, there are many challenges in actual operation. Therefore, how to improve the accuracy of the predicted reserve buffer has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The embodiments of this specification provide a prediction method, apparatus, computer device and storage medium, which can improve the prediction accuracy of the floating value of the reserve amount.

[0004] In a first aspect, a prediction method is provided, comprising: obtaining characteristic data of a business decision scenario to be predicted; wherein the characteristic data comprises: characteristic sub-data under a plurality of combinations, each of the combinations comprising a segmented scenario and a segmented dimension involved in the business decision scenario; inputting the characteristic sub-data under the plurality of combinations into probability distribution models corresponding to the plurality of combinations respectively, and outputting prediction results corresponding to the plurality of combinations respectively; wherein the prediction result is a probability density of a target value, and the target value represents the deviation between the predicted value of the asset under the combination and the actual value of the asset; and predicting the target floating value of the reserve amount of the business decision scenario according to the prediction results corresponding to the plurality of combinations respectively.

[0005] In combination with the first aspect, in some possible implementations, predicting the target floating value of the reserve amount for the business decision scenario based on the prediction results corresponding to each of the multiple combinations includes: converting the prediction results corresponding to each of the multiple combinations to obtain the cumulative distribution functions corresponding to each of the multiple combinations; determining the floating values ​​of the reserve amounts corresponding to each of the combinations under different preset cumulative probability values ​​based on the cumulative distribution functions corresponding to each of the multiple combinations; predicting the target floating values ​​of the reserve amounts for the business decision scenario under multiple parameter combinations based on the floating values ​​of the reserve amounts corresponding to each of the combinations under different preset cumulative probability values; wherein each of the parameter combinations includes: multiple cumulative probability values ​​and each of the cumulative probability values ​​corresponds to one combination.

[0006] In combination with the first aspect, in some possible implementations, the floating value of the reserve amount corresponding to each of the combinations under different preset cumulative probability values ​​is determined based on the cumulative distribution functions corresponding to each of the multiple combinations, including: determining the inverse function of the cumulative distribution function corresponding to each of the combinations; for each of the different cumulative probability values, using the cumulative probability value as an input parameter of the inverse function, and using the output value of the inverse function as the floating value of the reserve amount corresponding to each of the combinations under the cumulative probability value.

[0007] In combination with the first aspect, in some possible implementations, predicting the target floating value of the reserve amount for the business decision scenario under multiple parameter combinations based on the floating values ​​of the reserve amount corresponding to each of the combinations under different preset cumulative probability values ​​includes: determining multiple calculation calibers involved in the business decision scenario; wherein each of the calculation calibers corresponds to a calculation method for the target floating value of the reserve amount; predicting the target floating value of the reserve amount for the business decision scenario under multiple parameter combinations based on the floating values ​​of the reserve amount corresponding to each of the calculation calibers and each of the combinations under different preset cumulative probability values; wherein each of the parameter combinations also includes one of the calculation calibers.

[0008] In combination with the first aspect, in some possible implementations, after predicting the target floating value of the reserve amount under multiple parameter combinations for the business decision scenario based on the floating value of the reserve amount corresponding to each combination under different preset cumulative probability values, the method further includes: determining multiple decision preferences for the business decision scenario; sorting the target floating values ​​of the reserve amount under the multiple parameter combinations according to each decision preference, and outputting the sorted target floating values ​​of the reserve amount under the multiple parameter combinations.

[0009] In combination with the first aspect, in some possible implementations, the target floating values ​​of the reserve amount under the multiple parameter combinations are sorted according to each decision preference, including: determining the backtest evaluation standard corresponding to each decision preference; based on the backtest evaluation standard, counting the backtest evaluation data of the target floating values ​​of the reserve amount under the multiple parameter combinations within a historical time period; based on the backtest evaluation data, sorting the target floating values ​​of the reserve amount under the multiple parameter combinations.

[0010] In combination with the first aspect, in some possible implementations, the probability distribution model corresponding to each of the combinations is trained in the following manner: constructing a data set based on the segmented scenarios and segmented dimensions in the combination; wherein the data set corresponds to historical target values ​​and historical feature sub-data, and the historical target value represents the deviation between the historical asset prediction value and the historical asset actual value under the combination; performing model training according to the data set to obtain the probability distribution model corresponding to the combination.

[0011] In a second aspect, a prediction device is provided, comprising: an acquisition module for acquiring characteristic data of a business decision scenario to be predicted; wherein the characteristic data comprises: characteristic sub-data under multiple combinations, each of the combinations comprising a segmented scenario and a segmented dimension involved in the business decision scenario; an output module for inputting the characteristic sub-data under the multiple combinations into probability distribution models corresponding to the multiple combinations respectively, and outputting prediction results corresponding to the multiple combinations respectively; wherein the prediction result is a probability density of a target value, and the target value represents the deviation between the predicted value of the asset under the combination and the actual value of the asset; a prediction module for predicting the target floating value of the reserve amount of the business decision scenario according to the prediction results corresponding to the multiple combinations respectively.

[0012] In a third aspect, a computer device is provided, comprising one or more processors and one or more memories, wherein at least one computer program is stored in the one or more memories, and the computer program is loaded and executed by the one or more processors to implement the above-mentioned prediction method.

[0013] In a fourth aspect, a computer-readable storage medium is provided, wherein at least one computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the above-mentioned prediction method.

[0014] In a fifth aspect, a computer program product or a computer program is provided, which includes a program code, and the program code is stored in a computer-readable storage medium. A processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device performs the above-mentioned prediction method.

[0015] Through the technical solution provided by the embodiment of this specification, by subdividing the business decision scenario into multiple combinations and establishing a corresponding probability distribution model for each combination, the deviation between the predicted value of the asset and the actual value of the asset under each combination can be more accurately predicted. This refined prediction method can improve the accuracy of the overall prediction. Each combination contains a subdivision scenario and a subdivision dimension under the business decision scenario, so that analysis can be performed in multiple dimensions to more comprehensively understand the business situation. Through the probability density of the target value output by the probability density function, the floating value of the reserve amount can be dynamically adjusted, and the market changes and internal demand adjustments can be flexibly responded to, ensuring that the fund management is more flexible and efficient. The prediction results under each combination provide a specific probability density distribution, making risk management more accurate and able to prevent and control risks in a targeted manner. Based on the input of the characteristic sub-data under multiple combinations into the corresponding probability distribution model, the prediction results are automatically output, which simplifies the decision-making process and improves the decision-making efficiency. Through the data-driven prediction model, the decision-making process is made more scientific and rational, reducing the interference of human factors. The characteristic sub-data and the corresponding prediction results under each combination can be recorded and analyzed in detail, enhancing the transparency of the decision-making process. By obtaining the characteristic data of business decision scenarios and inputting them into the probability distribution models corresponding to multiple combinations, the technical solution of finally predicting the target floating value of the reserve amount can significantly improve the accuracy of the prediction, optimize resource allocation, improve decision-making efficiency, and enhance transparency, which enables financial institutions to manage funds more scientifically and improve overall operational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of this specification, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 It is a schematic diagram of an implementation environment of a prediction method provided in an embodiment of this specification;

[0018] Figure 2 is a schematic flow chart of a prediction method provided in an embodiment of this specification;

[0019] Figure 3 This is a block diagram of the implementation principle of a prediction method provided in an embodiment of this specification;

[0020] Figure 4 is a schematic diagram of the structure of a prediction device provided in an embodiment of this specification;

[0021] Figure 5is a schematic diagram of the structure of a terminal provided in an embodiment of this specification;

[0022] Figure 6 It is a structural diagram of a server provided in an embodiment of this specification. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of this specification more clear, the implementation methods of this specification will be further described in detail below with reference to the accompanying drawings.

[0024] In this specification, the terms "first", "second", etc. are used to distinguish identical or similar items with basically the same effects and functions. It should be understood that there is no logical or temporal dependency between "first", "second", and "nth", and there is no limitation on quantity and execution order.

[0025] Interbank financing management refers to the process by which financial institutions meet short-term liquidity needs by borrowing and lending funds from each other. This process involves the raising, allocation and management of funds to ensure that financial institutions can effectively respond to funding needs in daily operations. The main process of interbank financing management is as follows: First, use the algorithm model to produce future asset forecast results. For example, the future asset situation is predicted through the forecast model to obtain the forecast results of the assets, that is, the predicted value of the assets. Then, a funding plan is formulated based on the forecast results. For example, according to the above forecast results, a corresponding fund allocation plan is formulated to ensure the reasonable allocation of funds. In addition, a certain reserve buffer, that is, the floating value of the reserve amount, is reserved in the funding plan to cope with unforeseen funding needs. Then, use interbank management tools to assist in the execution of the entire fund management process.

[0026] Through research, this application has found that in the current interbank financing management process, although a forecasting model is used to predict future asset conditions, the forecasting results of the forecasting model cannot be 100% accurate, and the forecasting errors that exist are uncertain. The reserved reserve buffer relies on expert experience and has a large error. At the same time, due to the rich scenarios, high dimensions and high-frequency refresh of the model prediction results, the human brain has limited computing power, and it is impossible to rely on expert experience to cover all results, and the lag is high, resulting in low accuracy. In other words, due to the uncertainty of the forecasting model, the subjective dependence of the reserved reserve buffer, and the high complexity of the model prediction results, there are many challenges in actual operations, which are specifically manifested in the inability to quantify the accuracy of the forecasting results, the lack of objective basis for the determination of the reserve buffer, and the difficulty of expert experience to cover all forecasting results and the existence of lags. Therefore, how to improve the accuracy of the predicted reserve buffer has become a technical problem that needs to be solved urgently.

[0027] It is understandable that the core data relied on in the main process of interbank financing management is the predicted value, but the predicted value has a certain degree of uncertainty and volatility, which directly affects the cost reduction results. In order to reduce this uncertainty, the embodiment of this specification proposes a prediction method for quantifying uncertainty, which is conducive to improving the accuracy of the prediction results, allowing the business to see the future with more confidence and to reduce the cost of funds with more confidence.

[0028] In order to illustrate the technical solutions provided by the embodiments of this specification, some terms involved in the embodiments of this specification are first introduced.

[0029] Uncertainty Quantification (UQ): There are great uncertainties behind the models of many practical problems in the real world. These uncertainties may come from the complexity of the parameters, experimental measurements and geometric regions in the problem. How to reduce the risks brought by uncertainty by quantifying these uncertainties is the main purpose of uncertainty quantification research.

[0030] Probabilistic model and non-probabilistic model: In supervised regression tasks, the probabilistic model is determined by the conditional probability P(X|Y), and the non-probabilistic model is determined by the decision function Y=f(X).

[0031] Probability Density Function (PDF): describes the probability density of a continuous random variable taking values ​​within a certain interval.

[0032] Cumulative distribution function (CDF): It is the integral of the probability density function PDF, which describes the probability that a random variable is less than or equal to a value, that is, F(x) = P(X <= x).

[0033] Percent Point Function (PPF): Also known as the inverse cumulative distribution function, it is the inverse function of CDF.

[0034] Reserve amount: usually refers to the timely acquisition of sufficient funds at a reasonable cost to meet immediate payment needs, including daily business operations and maturity payments. Financial institutions use position management to monitor their own cash flow and ensure that there are sufficient reserves to cope with short-term funding needs.

[0035] Buffer: The floating value of the reserve amount, or the upper and lower limits of the reserve amount, that is, the amount that needs to be prepared in addition to the reserve amount to meet the instant payment needs. The floating value of the reserve amount is used to make the reserve amount float upward and / or downward, thereby obtaining a range of the reserve amount. The upper limit of the range of the reserve amount can be the reserve amount plus the floating value, and the lower limit of the range of the reserve amount can be the reserve amount minus the floating value.

[0036] Forecast value: The amount of funds that will flow in or out in the future preset period. Usually, the reserve amount is greater than the forecast value. In the case of fund inflow, the forecast value is the amount of funds that will flow in in the future preset period. In the case of fund outflow, the forecast value is the amount of funds that will flow out in the future preset period.

[0037] Actual value: the amount of funds that actually flow in or out during the current time period. In the case of funds inflow, the actual value is the amount of funds that actually flow in during the current time period. In the case of funds outflow, the actual value is the amount of funds that actually flow out during the current time period. For example, if the current time period and the preset time period are both daily time periods, and the predicted value is the amount of funds that will flow in or out on September 2 as predicted on September 1, the actual value can be understood as the amount of funds that actually flow in or out on September 2.

[0038] To facilitate understanding of the above reserve amount, reserve buffer and forecast value, a specific example is given below:

[0039] Suppose, in a certain scenario, it is predicted that 3 billion will flow out tomorrow, and the predicted value is 3 billion. Assuming that the lowest water level from a safety perspective is 1 billion, then if the prediction is 100% accurate, theoretically only 1 billion + 3 billion reserve funds need to be prepared, that is, the reserve amount is 4 billion. Assuming that the equipment payment buffer is 200 million, that is, the floating value of the reserve amount is 200 million, the reserve amount floats down 200 million to 3.8 billion, and the reserve amount floats up 200 million to 4.2 billion, then the final reserve amount is between 3.8 billion and 4.2 billion, and the predicted value is between 2.8 billion and 3.2 billion.

[0040] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this manual are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions. For example, the asset information involved in this manual is obtained with full authorization.

[0041] After introducing some terms involved in the embodiments of this specification, the implementation environment of the embodiments of this specification is described below.

[0042] Figure 1 This is a schematic diagram of an implementation environment of a prediction method provided in an embodiment of this specification, see Figure 1 , the implementation environment may include a prediction platform 110. The prediction platform 110 is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, distribution networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0043] The technical solution provided in the embodiments of this specification can be applied to interbank financing scenarios in the financial field. Interbank financing scenarios refer to a financial activity scenario in which financial institutions meet liquidity needs through short-term loans. In interbank financing scenarios, there may be different decision preferences for reserve buffers in different scenarios. At the same time, there are obvious differences in the accuracy performance of upstream asset prediction models in different dimensions and scenarios. The embodiments of this specification mainly solve the problem of how to quantitatively express the uncertainty of prediction results under multiple dimensions, multiple scenarios, and multiple decision preferences, and improve the prediction accuracy of reserve buffers.

[0044] Figure 2 is a schematic flow chart of a prediction method provided in the embodiments of this specification, see Figure 2 , taking the execution subject as a prediction platform as an example, the method includes the following steps.

[0045] Step 202: Acquire characteristic data of the business decision scenario to be predicted; wherein the characteristic data includes: characteristic sub-data in multiple combinations, each combination including a subdivision scenario and a subdivision dimension involved in the above business decision scenario.

[0046] Step 204: Input the characteristic sub-data under the various combinations into the probability distribution models corresponding to the various combinations respectively, and output the prediction results corresponding to the various combinations respectively; wherein the prediction result is the probability density of the target value, and the target value represents the deviation between the predicted value of the asset under the combination and the actual value of the asset.

[0047] Step 206: predicting the target floating value of the reserve amount for the above business decision scenario based on the prediction results corresponding to the various combinations.

[0048] exist Figure 2In the illustrated embodiment, by subdividing the business decision scenario into multiple combinations and establishing a corresponding probability distribution model for each combination, the deviation between the predicted value of the asset and the actual value of the asset under each combination can be predicted more accurately. This refined prediction method can improve the accuracy of the overall prediction. Each combination contains a subdivision scenario and a subdivision dimension under the business decision scenario, so that analysis can be performed in multiple dimensions to more comprehensively understand the business situation. Through the probability density of the target value output by the probability density function, the floating value of the reserve amount can be dynamically adjusted, and the market changes and internal demand adjustments can be flexibly responded to, ensuring that the fund management is more flexible and efficient. The prediction results under each combination provide a specific probability density distribution, making risk management more accurate and able to prevent and control risks in a targeted manner. Based on the input of the characteristic sub-data under multiple combinations into the corresponding probability distribution model, the prediction results are automatically output, which simplifies the decision-making process and improves the decision-making efficiency. Through the data-driven prediction model, the decision-making process is made more scientific and rational, and the interference of human factors is reduced. The characteristic sub-data and the corresponding prediction results under each combination can be recorded and analyzed in detail, enhancing the transparency of the decision-making process. By obtaining the characteristic data of business decision scenarios and inputting them into the probability distribution models corresponding to multiple combinations, the technical solution of finally predicting the target floating value of the reserve amount can significantly improve the accuracy of the prediction, optimize resource allocation, improve decision-making efficiency, and enhance transparency, which enables financial institutions to manage funds more scientifically and improve overall operational efficiency.

[0049] Below Figure 2 The specific implementation method of each step in the embodiment shown is described as follows:

[0050] In step 202, the prediction platform can obtain characteristic data of the business decision scenario to be predicted. Among them, the business decision scenario to be predicted can be understood as: a decision scenario that requires a floating value prediction of the reserve amount, such as: a large market position, a pure on-balance sheet position and other decision scenarios. In the financial field, large market positions refer to: these positions are directly or indirectly owned or assumed by financial institutions, including those indirectly held through off-balance sheet instruments or structured financing arrangements. Pure on-balance sheet positions refer to: assets or liabilities held by financial institutions, which are directly recorded on their balance sheets and reflect the actual financial status of the financial institutions.

[0051] The characteristic data includes: characteristic sub-data in multiple combinations, each combination includes a segmentation scenario and a segmentation dimension involved in the above business decision scenario. Each combination can be expressed as: segmentation dimension i*segmentation scenario j, where segmentation dimension i is any segmentation dimension involved in the above business decision scenario, and segmentation scenario j is any segmentation scenario involved in the above business decision scenario.

[0052] Among them, the characteristic sub-data mainly include: date characteristics and business characteristics. The business characteristics are mainly the statistical characteristics of the historical target value Y1 in recent days, such as the mean, maximum and minimum values, and variance of the historical target value Y1 in the past 7 days. The historical target value Y1 is the deviation between the predicted value and the real value of the historical asset. According to different dimensions, the historical target value Y1 is defined as ae = abs (prediction_value-real_value), or ape = abs (prediction_value-real_value) / real_value. Among them, prediction_value is the predicted value, real_value is the real value, ae represents the absolute value of the deviation between the predicted value and the real value, and ape represents the ratio of the absolute value of the deviation between the predicted value and the real value to the real value.

[0053] The subdivision scenarios are mainly divided based on the scenario differentiation requirements of business decisions. The following takes the two business decision scenarios of large-cap positions and pure on-balance sheet positions as examples to explain the subdivision scenarios involved:

[0054] The sub-scenarios involved in the business decision scenario of market position include: daily consumer credit scenario, hourly consumer credit scenario, daily borrowing scenario and hourly borrowing scenario. For example, in the consumer credit scenario, users are allowed to consume first and then repay on the payment platform. This consumer credit scenario can be understood as the Huabei scenario, where users can use Huabei to shop at merchants that support the service, enjoy a certain credit limit, and pay off the outstanding balance on a fixed monthly repayment date. In the borrowing scenario, users are allowed to apply for short-term loans for various consumer needs. This borrowing scenario can be understood as the Jiebei scenario, where users can apply online through Jiebei, obtain loans after approval, and repay the principal and interest according to the agreed period.

[0055] The sub-scenarios involved in the pure on-balance sheet position business decision scenario include: daily consumer credit scenario, hourly consumer credit scenario, daily borrowing scenario, hourly borrowing scenario, daily fund transfer scenario and hourly fund transfer scenario. Among them, the consumer credit scenario and borrowing scenario can refer to the above description, and the fund transfer scenario can include the bank-registered credit revolving pool scenario and the bank-registered credit suspension pool scenario. In the bank-registered credit revolving pool scenario, financial institutions package non-performing assets into a revolving pool through a specific asset management plan, and manage and dispose of these assets through cyclical purchases and sales. This method can help banks better manage and dispose of non-performing assets and optimize their balance sheets. In the bank-registered credit suspension pool scenario, financial institutions put some non-performing assets that cannot be disposed of temporarily into a special pool, suspend the disposal of these assets, and wait for the right time to deal with them. This method can help banks temporarily alleviate the pressure of non-performing assets and provide more flexibility for future disposal.

[0056] The above-mentioned hourly level refers to the prediction made within each hour's time period, and the above-mentioned daily level refers to the prediction made within the daily time period.

[0057] The subdivision dimensions are mainly divided based on the asset forecasting stage. The subdivision dimensions may include: loan dimension, repayment dimension, bank registration transfer dimension, and pure on-balance sheet repayment dimension. Specifically, the subdivision dimensions involved in the business decision scenario of the large market position include the loan dimension and the repayment dimension. The subdivision dimensions involved in the business decision scenario of the pure on-balance sheet position include: loan dimension, repayment dimension, bank registration transfer dimension, and pure on-balance sheet repayment dimension.

[0058] In step 204, each combination of segmentation dimension i*segmentation scenario j corresponds to its own probability distribution model, so that the characteristic sub-data under various combinations are respectively input into the probability distribution models corresponding to the various combinations, and the prediction results corresponding to the various combinations are output. The prediction result is the probability density of the target value, and the target value represents the deviation between the predicted value of the asset under the above combination and the actual value of the asset. The target value is the absolute value of the deviation between the predicted value of the asset under the above combination and the actual value of the asset, or the target value is the ratio of the absolute value of the deviation between the predicted value of the asset under the above combination and the actual value of the asset to the actual value of the asset. Among them, the probability distribution model of each combination can be trained based on the data set of each combination.

[0059] Exemplarily, the probability distribution model corresponding to each combination is trained in the following manner: constructing a data set based on the segmented scenarios and segmented dimensions in the above combination; performing model training according to the data set to obtain the probability distribution model corresponding to the combination; wherein the data set corresponds to historical target values ​​and historical feature sub-data, and the historical target value represents the deviation between the historical asset prediction value and the historical asset actual value under the combination.

[0060] Specifically, a data set can be constructed based on the combination of each segmentation dimension and segmentation scenario. The data set includes a data set of combination 1, a data set of combination 2, and a data set of combination n. Combination 1 is a combination of segmentation dimension 1 and segmentation scenario 1, combination 2 is a combination of segmentation dimension 2 and segmentation scenario 2, and combination n is a combination of segmentation dimension i and segmentation scenario j. For example, see Figure 3 , the data set of combination 1 includes the training set Training data and the test set Test data of combination 1, and the data set of combination n includes the training set Training data and the test set Test data of combination n. The historical target value and historical feature sub-data corresponding to each data set are: the historical target value corresponding to the training set in each data set ( Figure 3 The schematic diagram shows the expected deviation ae / ape) and historical characteristic sub-data ( Figure 3The historical feature sub-data mainly includes the date features and business features of the training set. Model training is performed based on the training set of each combination to obtain the probability distribution model corresponding to each combination. Model testing is performed based on the test set of each combination to obtain the probability distribution model that passes the test corresponding to each combination.

[0061] In the embodiment of the present application, data sets are constructed for different combinations of segmentation scenarios and segmentation dimensions, and the historical target value of the data set represents the deviation between the historical asset prediction value and the historical asset actual value. The feature set corresponding to the data set mainly includes date features and business features. By distinguishing different segmentation scenarios and segmentation dimensions and performing model training, the probability distribution model for prediction under different segmentation scenarios and segmentation dimensions can be accurately obtained. When performing model training, the current model uses NGBooster (Natural Gradient Boosting), which uses natural gradients to integrate base learners, probability distribution of parameters, and scoring rules.

[0062] When making predictions, we combine Figure 3 , the feature sub-data under combination 1 consisting of segmentation dimension 1 and segmentation scenario 1 can be input into probability distribution model 1 corresponding to combination 1, and prediction result 1 corresponding to the combination of segmentation dimension 1 and segmentation scenario 1 can be output. The feature sub-data under combination 2 consisting of segmentation dimension 2 and segmentation scenario 2 can be input into probability distribution model 2 corresponding to combination 2, and prediction result 2 corresponding to the combination of segmentation dimension 2 and segmentation scenario 2 can be output. The feature sub-data under combination n consisting of segmentation dimension i and segmentation scenario j can be input into probability distribution model n corresponding to combination n, and prediction result n corresponding to the combination of segmentation dimension i and segmentation scenario j can be output. Since it is a probability distribution model, the prediction result is the probability density P of the target value. θ (XY), θ is the parameter of the distribution.

[0063] In step 206, the prediction platform predicts the target floating value of the reserve amount of the business decision scenario based on the prediction results corresponding to each of the multiple combinations. According to the actual prediction needs, the target floating value of the reserve amount of the business decision scenario at a certain point in the future can be predicted, for example, the target floating value of the reserve amount after 1 hour in the future, the target floating value of the reserve amount after one day in the future, the target floating value of the reserve amount after one month in the future, etc. However, the embodiments of the present application do not make specific limitations on this. The target floating value of the reserve amount, that is, the target reserve buffer, refers to the floating value of the reserve amount under the business decision scenario. For example, predicting the target reserve buffer after one day in the future is to estimate how much reserve buffer the financial institution needs to prepare tomorrow for the business decision scenario through data analysis and prediction models to ensure that the financial institution has sufficient liquidity to cope with potential funding needs. This prediction helps financial institutions prepare in advance to ensure the safety and efficient use of funds.

[0064] It can be understood that the various combinations are composed of segmented scenarios and segmented dimensions involved in the business decision scenario to be predicted. Therefore, based on the prediction results corresponding to the various combinations, the target reserve buffer of the business decision scenario can be predicted.

[0065] Exemplarily, the implementation method of predicting the target floating value of the reserve amount of the above business decision scenario according to the prediction results corresponding to each of the multiple combinations includes the following S12 to S16:

[0066] S12: transforming the prediction results corresponding to the various combinations to obtain the cumulative distribution functions corresponding to the various combinations.

[0067] Specifically, the prediction results output by the probability distribution model corresponding to each of the multiple combinations can be converted into a cumulative distribution function of the predicted value as the target value, thereby obtaining the cumulative distribution function corresponding to each of the multiple combinations. In the embodiments of this specification, the cumulative distribution function CDF corresponding to each combination is used to describe the probability that the prediction result corresponding to the combination, i.e., the target value, is less than or equal to a certain specific value.

[0068] S14: According to the cumulative distribution functions corresponding to the various combinations, the floating values ​​of the reserve amounts corresponding to each combination under different preset cumulative probability values ​​are determined.

[0069] Among them, the preset different cumulative probability values ​​can be pre-set according to business needs, and the different cumulative probability values ​​set for different combinations can be the same or different. For example, the different cumulative probability values ​​set for combination 1 include: 50%, 60%, 70%, and the different cumulative probability values ​​set for combination 2 include: 55%, 65%, 75%. In this case, the cumulative probability values ​​set for each combination are different. Alternatively, the different cumulative probability values ​​set for all combinations include: 50%, 60%, 70%. In this case, the cumulative probability values ​​set for each combination are the same. It should be noted that the above example only takes the number of cumulative probability values ​​preset for each combination as 3 as an example. In the specific implementation, the number of cumulative probability values ​​preset for each combination is not limited to this.

[0070] Exemplarily, the above-mentioned cumulative distribution functions corresponding to each of the multiple combinations are used to determine the floating value of the reserve amount corresponding to each combination under different preset cumulative probability values, including: determining the inverse function of the cumulative distribution function corresponding to each combination; for each cumulative probability value among the different cumulative probability values, using the cumulative probability value as the input parameter of the inverse function, and using the output value of the inverse function as the floating value of the reserve amount corresponding to each combination under the cumulative probability value.

[0071] Among them, the inverse function of the cumulative distribution function is the percentile function PPF. After a cumulative probability value Xn is input into the PPF function, the PPF function can output PPF(Xn), which can be used as a floating value of the corresponding reserve amount under the cumulative probability value Xn. In other words, after each cumulative probability value Xn is input into the PPF function, the PPF function can output a specific value PPF(Xn), so that the probability value that the target value is less than or equal to the PPF(Xn) is equal to the cumulative probability value Xn, and the specific value is the floating value of the corresponding reserve amount under the cumulative probability value.

[0072] It can be understood that each combination corresponds to its own cumulative distribution function, which also corresponds to its own inverse function, namely the PPF function. Based on the PPF function corresponding to each combination and the multiple cumulative probability values ​​set for each combination, the floating value of the reserve amount corresponding to each combination under multiple cumulative probability values ​​can be obtained.

[0073] See also Figure 3 , transform the prediction results corresponding to combination 1, combination 2, ... combination n, and obtain the cumulative distribution function CDF1 corresponding to combination 1, CDF2 corresponding to combination 2, ... CDFn corresponding to combination n. Then, determine the percentile function PPF1 corresponding to CDF1, PPF2 corresponding to CDF2, ... PPFn corresponding to CDFn.

[0074] Next, based on PPF1, the floating value of the reserve amount corresponding to combination 1 under multiple cumulative probability values ​​is obtained, that is, reserve buffer 1. Reserve buffer 1 includes multiple reserve buffers and each reserve buffer corresponds to a cumulative probability value. Each reserve buffer in reserve buffer 1 is calculated through PPF1 and a cumulative probability value.

[0075] Based on PPF2, the floating value of the reserve amount corresponding to combination 2 under multiple cumulative probability values ​​is obtained, that is, reserve buffer 2. Reserve buffer 2 includes multiple reserve buffers and each reserve buffer corresponds to a cumulative probability value. Each reserve buffer in reserve buffer 2 is calculated by PPF2 and a cumulative probability value.

[0076] Based on PPFn, the floating value of the reserve amount corresponding to combination n under multiple cumulative probability values ​​is obtained, namely, reserve buffern. The reserve buffern includes multiple reserve buffers and each reserve buffer corresponds to a cumulative probability value. Each reserve buffer in the reserve buffern is calculated by PPFn and a cumulative probability value.

[0077] S16: According to the reserve amounts corresponding to each combination under different preset cumulative probability values, predict the target floating value of the reserve amount under multiple parameter combinations of the above business decision scenario. Each parameter combination includes: multiple cumulative probability values ​​and each cumulative probability value corresponds to one combination.

[0078] Among them, each business decision scenario corresponds to its own calculation formula for the target floating value of the reserve amount, so that the target floating value of the reserve amount for each business decision scenario under each parameter combination can be calculated based on the reserve amount corresponding to each combination under different preset cumulative probability values ​​and the calculation formula corresponding to the business decision scenario. One parameter combination corresponds to a target floating value of the reserve amount. In the embodiments of this specification, not only the target floating value of the reserve amount will be output, but also the parameter combination corresponding to the target floating value of the reserve amount will be output, so that relevant business personnel can know not only the target floating value of the reserve amount, but also the calculation logic inside the target floating value of the reserve amount, that is, based on what parameters it is calculated.

[0079] Exemplarily, the above-mentioned method predicts the target floating value of the reserve amount under multiple parameter combinations for business decision scenarios based on the reserve amounts corresponding to the different preset cumulative probability values ​​of each combination, including: determining multiple calculation calibers involved in the business decision scenario; wherein each calculation caliber corresponds to a method for calculating the target floating value of the reserve amount; predicting the target floating value of the reserve amount under multiple parameter combinations for business decision scenarios based on the reserve amounts corresponding to the different preset cumulative probability values ​​of each combination; wherein each parameter combination also includes a calculation caliber.

[0080] Specifically, the calculation formula for the target floating value of the reserve amount corresponding to each business decision scenario includes calculation methods of multiple calculation calibers involved in the business scenario, and each calculation caliber corresponds to a calculation method of the target floating value of the reserve amount. The multiple calculation calibers may specifically include the following four calculation calibers:

[0081] Calculation caliber 1: upper limit of capital outflow - lower limit of capital inflow. Under this calculation caliber, the calculation formula for the target floating value of the reserve amount is: (the forecast value corresponding to the capital outflow combination + the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1 of 1) - (the forecast value corresponding to the capital inflow combination - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2 of 1) - (the forecast value corresponding to the capital outflow combination - the forecast value corresponding to the capital inflow combination) = the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1 of 1 + the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2 of 1. Among them, Xn1 and Xn2 are the cumulative probability values ​​in the preset different cumulative probability values. For calculation caliber 1, each time Xn1 and Xn2 are selected, a target floating value of the reserve amount can be calculated by the above formula.

[0082] Calculation caliber 2: upper limit of capital outflow - upper limit of capital inflow. Under this calculation caliber, the calculation formula for the target floating value of the reserve amount is: (the forecast value corresponding to the capital outflow combination + the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1 of 1) - (the forecast value corresponding to the capital inflow combination + the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2 of 1) - (the forecast value corresponding to the capital outflow combination - the forecast value corresponding to the capital inflow combination) = the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1 of 1 - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2 of 1. Among them, Xn1 and Xn2 are the cumulative probability values ​​in the preset different cumulative probability values. For calculation caliber 2, each time Xn1 and Xn2 are selected, a target floating value of the reserve amount can be calculated by the above formula.

[0083] Calculation caliber 3: lower limit of capital outflow - upper limit of capital inflow. Under this calculation caliber, the calculation formula for the target floating value of the reserve amount is: (the forecast value corresponding to the capital outflow combination - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1) - (the forecast value corresponding to the capital inflow combination + the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2) - (the forecast value corresponding to the capital outflow combination - the forecast value corresponding to the capital inflow combination) = - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1 - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2. Among them, Xn1 and Xn2 are the cumulative probability values ​​in the preset different cumulative probability values. For calculation caliber 3, each time Xn1 and Xn2 are selected, a target floating value of the reserve amount can be calculated by the above formula.

[0084] Calculation caliber 4: lower limit of capital outflow - lower limit of capital inflow. Under this calculation caliber, the calculation formula for the target floating value of the reserve amount is: (the forecast value corresponding to the capital outflow combination - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1 of 1) - (the forecast value corresponding to the capital inflow combination - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2 of 1) - (the forecast value corresponding to the capital outflow combination - the forecast value corresponding to the capital inflow combination) = - the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn1 of 1 + the floating value of the reserve amount corresponding to the capital outflow combination under the cumulative probability value Xn2 of 1. Among them, Xn1 and Xn2 are the cumulative probability values ​​in the preset different cumulative probability values. For calculation caliber 4, each time Xn1 and Xn2 are selected, a target floating value of the reserve amount can be calculated by the above formula.

[0085] Among them, the fund outflow combination may include: the combination formed by Huabei scenario and loan dimension, the combination formed by Jiebei scenario and loan dimension, etc., which represent fund outflow. The fund inflow combination may include: the combination formed by Huabei scenario and repayment dimension, the combination formed by Jiebei scenario and repayment dimension, etc., which represent fund inflow. The calculation caliber involving the lower limit is expressed in the calculation formula as: predicted value - floating value of the reserve amount, and the calculation caliber involving the upper limit is expressed in the calculation formula as: predicted value + floating value of the reserve amount.

[0086] To facilitate the understanding of the above four calculation methods, the following is an explanation of the calculation method of the target floating value of the reserve amount corresponding to the above calculation method 2, with the help of specific examples:

[0087] Assuming that the business decision scenario is a large market position, the sub-scenarios involved in the large market position include: Huabei scenario and Jiebei scenario, and the sub-dimensions involved include: loan dimension and repayment dimension. The capital outflow combinations involved in the large market position business decision scenario may include: the combination formed by the Huabei scenario and the loan dimension, and the combination formed by the Jiebei scenario and the loan dimension. The capital inflow combinations involved in the large market position business decision scenario may include: the combination formed by the Huabei scenario and the repayment dimension, and the combination formed by the Jiebei scenario and the repayment dimension. Assuming that the predicted value corresponding to the combination formed by the Huabei scenario and the loan dimension (referred to as Huabei's loan) is 1 billion, the predicted value corresponding to the combination formed by the Huabei scenario and the repayment dimension (referred to as Huabei's repayment) is 2 billion, the predicted value corresponding to the combination formed by the Jiebei scenario and the loan dimension (referred to as Jiebei's loan) is 2 billion, and the predicted value corresponding to the combination formed by the Jiebei scenario and the repayment dimension (referred to as Jiebei's repayment) is 3 billion. The forecast value for the fund outflow combination is: the forecast value for Jiebei's loan (2 billion) + the forecast value for Huabei's loan (1 billion) = 3 billion. The forecast value for the fund inflow combination is: the forecast value for Huabei's repayment (2 billion) + the forecast value for Jiebei's repayment (3 billion) = 5 billion.

[0088] For the four combinations of Huabei's lending, Huabei's repayment, Jiebei's lending and Jiebei's repayment, assuming that the respectively preset cumulative probability values ​​are: 60%, 50%, 50%, 60%, then a parameter combination may include: [60%, 50%, 50%, 60%, calculation caliber 2].

[0089] Assume that the corresponding reserve amount for Huabei's loan combination under the cumulative probability value of 60% is 200 million, the corresponding floating value of the reserve amount for Huabei's repayment combination under the cumulative probability value of 50% is 100 million, the corresponding floating value of the reserve amount for Jiebei's loan combination under the cumulative probability value of 50% is 300 million, and the corresponding floating value of the reserve amount for Jiebei's repayment combination under the cumulative probability value of 60% is 100 million.

[0090] The floating value of the reserve amount corresponding to the fund outflow combination under the cumulative probability value Xn1 is: the floating value of the reserve amount corresponding to the Huabei loan combination under the cumulative probability value of 60% (200 million) + the floating value of the reserve amount corresponding to the Jiebei loan combination under the cumulative probability value of 50% (300 million) = 500 million. The floating value of the reserve amount corresponding to the fund inflow combination under the cumulative probability value Xn2 is: the floating value of the reserve amount corresponding to the Huabei repayment combination under the cumulative probability value of 50% (100 million) + the floating value of the reserve amount corresponding to the Jiebei repayment combination under the cumulative probability value of 60% (100 million) = 200 million.

[0091] The target floating value of the reserve amount for the business decision scenario of the large-cap position under the parameter combination of [60%, 50%, 50%, 60%, calculation basis 2] is: [(10+2)+(20+3)]-[(20+1)-(30+1)]-[(10+20)-(20+30)]=-17-(-20)=300 million.

[0092] By adopting the above method of calculating the target floating value of the reserve amount, the target floating value of the reserve amount for the business decision scenario of the large-cap position under other parameter combinations can be calculated. Each parameter combination includes a cumulative probability value corresponding to the above four combinations and any one of the four calculation methods.

[0093] Exemplarily, after predicting the target floating value of the reserve amount under multiple parameter combinations for the above-mentioned business decision scenarios based on the reserve amounts corresponding to the respective preset different cumulative probability values ​​for each combination, it also includes: determining multiple decision preferences for the above-mentioned business decision scenarios; sorting the target floating values ​​of the reserve amount under multiple parameter combinations according to each decision preference, and outputting the sorted target floating values ​​of the reserve amount under the multiple parameter combinations.

[0094] It is understandable that, since there is no certain order of arrangement between the target floating values ​​of the reserve amount corresponding to each of the predicted multiple parameter combinations, it is not convenient for the subsequent relevant business personnel to finally select the target floating value of the reserve amount corresponding to a specific parameter combination. Based on this, the present description embodiment can determine multiple decision preferences for the above-mentioned business decision scenarios; according to each decision preference, the target floating values ​​of the reserve amount under the multiple parameter combinations are sorted, and the target floating values ​​of the reserve amount under the sorted multiple parameter combinations are output, which is convenient for the relevant business personnel to make targeted choices based on the target floating values ​​of the reserve amount under the sorted multiple parameter combinations, and obtain the appropriate target floating value of the reserve amount. In a specific implementation, the target floating value of the reserve amount under the sorted multiple parameter combinations can be output to the display device of the relevant business personnel, and the reserve amounts under the sorted multiple parameter combinations are displayed in order on the display device, which is convenient for the relevant business personnel to view.

[0095] In the financial field, decision preferences are used to describe the behavioral characteristics and psychological tendencies of investors when facing investment decisions. Among them, various decision preferences may include: risk aversion, return preference, etc. Risk aversion refers to the preference for zero risk and would rather lose some returns. Return preference refers to the preference for higher returns. For different decision preferences, the target floating values ​​of the reserve amount under the sorted multiple parameter combinations are different. Under each decision preference, the target floating values ​​of the reserve amount under the sorted multiple parameter combinations obtained, the target floating values ​​of the reserve amount corresponding to the parameter combinations with higher rankings have a higher degree of match with the decision preference, indicating that under this decision preference, it is more appropriate to select the target floating value of the reserve amount corresponding to the parameter combination with a higher degree of match, which is conducive to the intelligent recommendation of the target floating value of the reserve amount.

[0096] Exemplarily, the above-mentioned sorting of the target floating values ​​of the reserve amount under multiple parameter combinations according to each decision preference includes: determining the backtest evaluation standard corresponding to each decision preference; based on the backtest evaluation standard, counting the backtest evaluation data of the target floating values ​​of the reserve amount under multiple parameter combinations within the historical time period; and sorting the target floating values ​​of the reserve amount under multiple parameter combinations based on the backtest evaluation data.

[0097] Among them, the backtest evaluation criteria are formulated according to decision preferences. For example, the risk-averse backtest evaluation criteria include: the possibility that the capital supply is still insufficient after the target floating value of the reserve amount is evaluated (mainly considering the positive and negative signs of the error between the historical forecast value plus the target floating value of the reserve amount and the historical true value). Breakdown means: the historical forecast value plus the predicted target floating value of the reserve amount is less than the historical true value, so there is not enough money, which means that the forecast value is broken by the true value. The return preference backtest evaluation criteria include: the evaluation of the situation where the target floating value of the reserve amount is close to the historical true value after adding the reserve amount (mainly considering the error between the historical forecast value plus the target floating value of the reserve amount and the historical true value).

[0098] The historical time period can be pre-calibrated, for example, it can be the historical time period of the past 1 month, the past 2 months, the past half year, etc. For example, for risk aversion, the backtest performance of the target floating value of the reserve amount under various parameter combinations in the past 2 months can be counted according to the backtest evaluation standard, that is, the backtest evaluation data. The backtest evaluation data can represent the probability of being penetrated after the historical prediction value is added to the target floating value of the reserve amount, so that the target floating value of the reserve amount under each parameter combination will correspond to a penetration probability. According to the penetration probability, the target floating values ​​of the reserve amount under various parameter combinations are sorted, and the target floating value of the reserve amount under the parameter combination with the smaller penetration probability is ranked higher, or the target floating value of the reserve amount under the parameter combination with the smaller penetration probability and the smaller the difference between the historical prediction value plus the target floating value and the historical true value is ranked higher.

[0099] For the yield preference type, the backtest performance of the target floating value of the reserve amount under various parameter combinations in the past two months can be counted according to the backtest evaluation standard, that is, the backtest evaluation data. The backtest evaluation data can represent the historical difference between the historical forecast value plus the target floating value of the reserve amount and the historical actual value, so that the target floating value of the reserve amount under each parameter combination will correspond to a historical difference. According to the historical difference, the target floating values ​​of the reserve amount under various parameter combinations are sorted, and the target floating values ​​of the reserve amount under the parameter combination with the smaller historical difference are ranked higher.

[0100] Based on the historical backtest performance, that is, the backtest evaluation data, it is easy to understand the performance of the target floating value of the reserve amount under various parameter combinations under actual historical circumstances, and then output the target floating value of the reserve amount under various parameter combinations with the best ranking under different decision preferences in each business decision scenario, so that relevant business personnel can select the target floating value of the reserve amount under a specific parameter combination suitable for a certain decision preference based on the performance of the target floating value of the reserve amount under various parameter combinations under actual historical circumstances.

[0101] In the embodiments of this specification, the deviation between the predicted value and the true value is directly modeled, the result is explainable, and it is coupled with the predicted value obtained by the upstream prediction link prediction, which is more flexible. The probability density of the deviation between the future predicted value and the true value is predicted by a probability distribution model, which is more in line with the actual business needs than the traditional non-probability model (point prediction, only a scalar value is given). Through the PPF function and historical backtesting performance, the abstract probability density function is converted into a reserve amount that actually has business significance, which can be flexibly adapted to various business decision preferences.

[0102] All the optional implementation modes described above can be arbitrarily combined to form optional embodiments of this specification, and will not be described one by one here.

[0103] Figure 4 is a schematic diagram of a prediction device provided in the embodiments of this specification, see Figure 4 , the prediction device 400 comprises:

[0104] The acquisition module 401 is used to acquire characteristic data of the business decision scenario to be predicted; wherein the characteristic data includes: characteristic sub-data in multiple combinations, each of which includes a subdivision scenario and a subdivision dimension involved in the business decision scenario;

[0105] Output module 402, used to input the characteristic sub-data under the multiple combinations into the probability distribution models corresponding to the multiple combinations respectively, and output the prediction results corresponding to the multiple combinations respectively; wherein the prediction results are the probability density of the target value, and the target value represents the deviation between the predicted value of the asset under the combination and the actual value of the asset;

[0106] The prediction module 403 is used to predict the target floating value of the reserve amount of the business decision scenario according to the prediction results corresponding to the multiple combinations.

[0107] In one possible implementation, the prediction module 403 is specifically used to: transform the prediction results corresponding to each of the multiple combinations to obtain the cumulative distribution functions corresponding to each of the multiple combinations; determine the floating values ​​of the reserve amounts corresponding to each of the combinations under different preset cumulative probability values ​​based on the cumulative distribution functions corresponding to each of the multiple combinations; predict the target floating values ​​of the reserve amounts for the business decision scenarios under multiple parameter combinations based on the floating values ​​of the reserve amounts corresponding to each of the combinations under different preset cumulative probability values; wherein each of the parameter combinations includes: multiple cumulative probability values ​​and each of the cumulative probability values ​​corresponds to one combination.

[0108] In one possible implementation, the prediction module 403 is specifically used to: determine the inverse function of the cumulative distribution function corresponding to each of the combinations; for each of the different cumulative probability values, use the cumulative probability value as an input parameter of the inverse function, and use the output value of the inverse function as a floating value of the reserve amount corresponding to each of the combinations under the cumulative probability value.

[0109] In one possible implementation, the prediction module 403 is specifically used to: determine multiple calculation calibers involved in the business decision scenario; wherein each of the calculation calibers corresponds to a calculation method for the target floating value of the reserve amount; based on the reserve amounts corresponding to each of the calculation calibers and each of the combinations under different preset cumulative probability values, predict the target floating value of the reserve amount under multiple parameter combinations for the business decision scenario; wherein each of the parameter combinations also includes one of the calculation calibers.

[0110] In a possible implementation, the prediction device 400 also includes a sorting module, which is used to determine multiple decision preferences for the business decision scenario after predicting the target floating value of the reserve amount under multiple parameter combinations according to the floating value of the reserve amount corresponding to each combination under different preset cumulative probability values; sort the target floating values ​​of the reserve amount under the multiple parameter combinations according to each decision preference, and output the sorted target floating values ​​of the reserve amount under the multiple parameter combinations.

[0111] In one possible implementation, the sorting module is specifically used to: determine the backtest evaluation criteria corresponding to each of the decision preferences; based on the backtest evaluation criteria, collect backtest evaluation data of the target floating values ​​of the reserve amount under multiple parameter combinations within a historical time period; and sort the target floating values ​​of the reserve amount under the multiple parameter combinations based on the backtest evaluation data.

[0112] In one possible implementation, the prediction device 400 also includes: a training module, used to construct a data set based on the segmented scenarios and segmented dimensions in the combination; wherein the data set corresponds to historical target values ​​and historical feature sub-data, and the historical target value represents the deviation between the historical asset prediction value and the historical asset actual value under the combination; model training is performed according to the data set to obtain a probability distribution model corresponding to the combination.

[0113] It should be noted that: the prediction device provided in the above embodiment only uses the division of the above functional modules as an example for explanation during prediction. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the functions described above. In addition, the prediction device provided in the above embodiment belongs to the same concept as the prediction method embodiment. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0114] The embodiments of this specification provide a computer device for executing the above method. The computer device can be implemented as a terminal or a server. The structure of the terminal is first introduced below:

[0115] Figure 5 5 is a schematic diagram of a terminal structure provided in an embodiment of the present specification. The terminal 500 may be: a smart phone, a tablet computer, a laptop computer or a desktop computer. The terminal 500 may also be called a user equipment, a portable terminal, a laptop terminal, a desktop terminal or other names.

[0116] Typically, the terminal 500 includes: one or more processors 501 and one or more memories 502 .

[0117] The processor 501 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 501 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 501 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 501 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen.

[0118] The memory 502 may include one or more computer-readable storage media, which may be non-transitory. The memory 502 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 502 is used to store at least one computer program, which is used to be executed by the processor 501 to implement the prediction method provided in the method embodiment of this specification.

[0119] In some embodiments, the terminal 500 may further optionally include: a peripheral device interface 503 and at least one peripheral device. The processor 501, the memory 502 and the peripheral device interface 503 may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface 503 via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit 504, a display screen 505, a camera assembly 506, an audio circuit 507 and a power supply 508.

[0120] The peripheral device interface 503 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 501 and the memory 502. In some embodiments, the processor 501, the memory 502, and the peripheral device interface 503 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 501, the memory 502, and the peripheral device interface 503 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0121] The radio frequency circuit 504 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 504 communicates with a communication network and other communication devices via electromagnetic signals. The radio frequency circuit 504 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 504 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, and the like.

[0122] The display screen 505 is used to display a UI (User Interface). The UI may include graphics, text, icons, videos, and any combination thereof. When the display screen 505 is a touch display screen, the display screen 505 also has the ability to collect touch signals on the surface or above the surface of the display screen 505. The touch signal may be input as a control signal to the processor 501 for processing. At this time, the display screen 505 may also be used to provide virtual buttons and / or virtual keyboards, also known as soft buttons and / or soft keyboards.

[0123] The camera assembly 506 is used to collect images or videos. Optionally, the camera assembly 506 includes a front camera and a rear camera. Usually, the front camera is set on the front panel of the terminal, and the rear camera is set on the back of the terminal.

[0124] The audio circuit 507 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, and convert the sound waves into electrical signals and input them to the processor 501 for processing, or input them to the radio frequency circuit 504 to achieve voice communication.

[0125] The power supply 508 is used to supply power to various components in the terminal 500. The power supply 508 may be alternating current, direct current, a disposable battery, or a rechargeable battery.

[0126] In some embodiments, the terminal 500 further includes one or more sensors 509 , including but not limited to: an acceleration sensor 510 , a gyroscope sensor 511 , a pressure sensor 512 , an optical sensor 513 , and a proximity sensor 514 .

[0127] The acceleration sensor 510 can detect the magnitude of acceleration on three coordinate axes of the coordinate system established with the terminal 500 .

[0128] The gyro sensor 511 can detect the body direction and rotation angle of the terminal 500 . The gyro sensor 511 can cooperate with the acceleration sensor 510 to collect the user's 3D actions on the terminal 500 .

[0129] The pressure sensor 512 may be provided at the side frame of the terminal 500 and / or the lower layer of the display screen 505. When the pressure sensor 512 is provided at the side frame of the terminal 500, it may detect the user's holding signal of the terminal 500, and the processor 501 may perform left and right hand recognition or shortcut operation according to the holding signal collected by the pressure sensor 512. When the pressure sensor 512 is provided at the lower layer of the display screen 505, the processor 501 may control the operability controls on the UI interface according to the user's pressure operation on the display screen 505.

[0130] The optical sensor 513 is used to collect the ambient light intensity. In one embodiment, the processor 501 can control the display brightness of the display screen 505 according to the ambient light intensity collected by the optical sensor 513.

[0131] The proximity sensor 514 is used to collect the distance between the user and the front of the terminal 500 .

[0132] Those skilled in the art will understand that Figure 5 The structure shown in the figure does not constitute a limitation on the terminal 500, and the terminal 500 may include more or less components than those shown in the figure, or combine some components, or adopt a different component arrangement.

[0133] The above-mentioned computer device can also be implemented as a server. The structure of the server is introduced below:

[0134] Figure 6 This is a schematic diagram of the structure of a server provided in an embodiment of this specification. The server 600 may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) 601 and one or more memories 602, wherein the one or more memories 602 store at least one computer program, and the at least one computer program is loaded and executed by the one or more processors 601 to implement the methods provided in the above-mentioned various method embodiments. Of course, the server 600 may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output. The server 600 may also include other components for implementing device functions, which will not be described in detail here.

[0135] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including a computer program, and the computer program can be executed by a processor to perform the prediction method in the above embodiment. For example, the computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0136] In an exemplary embodiment, a computer program product or a computer program is also provided, which includes a program code, and the program code is stored in a computer-readable storage medium. A processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device performs the above-mentioned prediction method.

[0137] In some embodiments, the computer program involved in the embodiments of this specification may be deployed and executed on a computer device, or on multiple computer devices located at one location, or on multiple computer devices distributed at multiple locations and interconnected by a communication network. Multiple computer devices distributed at multiple locations and interconnected by a communication network may constitute a blockchain system.

[0138] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0139] The above are only optional embodiments of this specification and are not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this specification should be included in the protection scope of this specification.

Claims

1. A prediction method, comprising: Acquire characteristic data of the business decision scenario to be predicted; wherein the characteristic data includes: characteristic sub-data in multiple combinations, each of the combinations including a subdivision scenario and a subdivision dimension involved in the business decision scenario; Input the characteristic sub-data under the multiple combinations into the probability distribution models corresponding to the multiple combinations respectively, and output the prediction results corresponding to the multiple combinations respectively; wherein the prediction results are the probability density of the target value, and the target value represents the deviation between the predicted value of the asset under the combination and the actual value of the asset; According to the prediction results corresponding to each of the multiple combinations, a target floating value of the reserve amount of the business decision scenario is predicted.

2. The method according to claim 1, wherein: The predicting the target floating value of the reserve amount of the business decision scenario according to the prediction results corresponding to each of the multiple combinations includes: Transforming the prediction results corresponding to each of the multiple combinations to obtain the cumulative distribution functions corresponding to each of the multiple combinations; According to the cumulative distribution functions corresponding to the plurality of combinations, the floating values ​​of the reserve amounts corresponding to each of the combinations under different preset cumulative probability values ​​are determined; According to the floating values ​​of the reserve amount corresponding to each of the combinations under the preset different cumulative probability values, the target floating value of the reserve amount under multiple parameter combinations of the business decision scenario is predicted; wherein each of the parameter combinations includes: multiple cumulative probability values ​​and each of the cumulative probability values ​​corresponds to one combination.

3. The method according to claim 2, wherein: The step of determining the floating value of the reserve amount corresponding to each of the combinations under different preset cumulative probability values ​​according to the cumulative distribution functions corresponding to the multiple combinations respectively includes: Determine the inverse function of the cumulative distribution function corresponding to each of the combinations; For each of the different cumulative probability values, the cumulative probability value is used as an input parameter of the inverse function, and the output value of the inverse function is used as a floating value of the reserve amount corresponding to each combination under the cumulative probability value.

4. The method according to claim 2, wherein: The method of predicting the target floating value of the reserve amount under multiple parameter combinations of the business decision scenario according to the floating value of the reserve amount corresponding to each combination under different preset cumulative probability values ​​includes: Determine multiple calculation calibers involved in the business decision scenario; wherein each calculation caliber corresponds to a calculation method of the target floating value of the reserve amount; According to the reserve amounts corresponding to each of the calculation calibers and each of the combinations under different preset cumulative probability values, the target floating value of the reserve amount under multiple parameter combinations of the business decision scenario is predicted; wherein each of the parameter combinations also includes one of the calculation calibers.

5. The method according to claim 2, wherein: After predicting the target floating value of the reserve amount under multiple parameter combinations of the business decision scenario according to the floating value of the reserve amount corresponding to each combination under different preset cumulative probability values, the method further includes: determining multiple decision preferences for the business decision scenario; According to each of the decision preferences, the target floating values ​​of the reserve amounts under the multiple parameter combinations are sorted, and the sorted target floating values ​​of the reserve amounts under the multiple parameter combinations are output.

6. The method according to claim 5, wherein: The step of ranking the target floating values ​​of the reserve amounts under the multiple parameter combinations according to each of the decision preferences includes: Determine the backtest evaluation criteria corresponding to each of the decision preferences; According to the backtest evaluation criteria, the backtest evaluation data of the target floating value of the reserve amount under various parameter combinations within the historical time period are collected; According to the backtest evaluation data, the target floating values ​​of the reserve amounts under the multiple parameter combinations are sorted.

7. The method according to claim 1, wherein: The probability distribution model corresponding to each combination is trained in the following way: A data set is constructed based on the segmented scenarios and segmented dimensions in the combination; wherein the data set corresponds to a historical target value and historical characteristic sub-data, and the historical target value represents the deviation between the historical asset forecast value and the historical asset real value under the combination; Model training is performed according to the data set to obtain a probability distribution model corresponding to the combination.

8. A prediction device, comprising: An acquisition module is used to acquire characteristic data of a business decision scenario to be predicted; wherein the characteristic data includes: characteristic sub-data in multiple combinations, each of which includes a subdivision scenario and a subdivision dimension involved in the business decision scenario; An output module, used to input the characteristic sub-data under the multiple combinations into the probability distribution models corresponding to the multiple combinations respectively, and output the prediction results corresponding to the multiple combinations respectively; wherein the prediction result is the probability density of the target value, and the target value represents the deviation between the predicted value of the asset under the combination and the actual value of the asset; The prediction module is used to predict the target floating value of the reserve amount of the business decision scenario according to the prediction results corresponding to each of the multiple combinations.

9. A computer device, comprising one or more processors and one or more memories, wherein at least one computer program is stored in the one or more memories, and the computer program is loaded and executed by the one or more processors to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, wherein at least one computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the method according to any one of claims 1 to 7.

Images