WO2023273231A1

WO2023273231A1 - Position prediction assessment method and apparatus, device, and storage medium

Info

Publication number: WO2023273231A1
Application number: PCT/CN2021/140613
Authority: WO
Inventors: 高攀; 曾岩; 李晶
Original assignee: 深圳前海微众银行股份有限公司
Priority date: 2021-06-28
Filing date: 2021-12-22
Publication date: 2023-01-05
Also published as: CN113344696A

Abstract

Disclosed is a position prediction assessment method, the method comprising: determining m first actual position amounts of a target to be assessed within a quality assessment period (101), wherein m is an integer greater than or equal to one; determining n second actual position amounts in abnormal states from among the m first actual position amounts (102), wherein n is an integer greater than or equal to zero; and determining a quality assessment result corresponding to the target to be assessed within the quality assessment period on the basis of the n second actual position amounts and the m first actual position amounts (103), wherein the quality assessment result is used for representing a prediction quality of position prediction performed on the target to be assessed within the quality assessment period. Further disclosed are a position prediction assessment apparatus, a device, and a storage medium.

Description

A position forecast evaluation method, device, equipment and storage medium

Cross References to Related Applications

This application is based on a Chinese patent application with application number 202110719946.3 and a filing date of June 28, 2021, and claims the priority of this Chinese patent application. The entire content of this Chinese patent application is hereby incorporated by reference into this application.

technical field

The present application relates to the technical field of computer analysis, and in particular to a position forecast evaluation method, device, equipment and storage medium.

Background technique

With the rapid development of computer technology, more and more technologies are applied in the financial field, and the traditional financial industry is gradually transforming into financial technology (Fintech). However, due to the security and real-time requirements of the financial industry, more and more technical requirements high demands. At present, the commercial banking system usually adopts the method of fund position management in the daily operation process to ensure a reasonable level of positions to deal with all cash outflows, and to prevent the waste of funds caused by excessive retention of positions. In the current fund position management process, the retained amount of the position is mainly realized through position forecasting. In order to determine the accuracy of the position forecast, the commercial banking system usually chooses a period of time such as a week or a month. Scoring the position forecast, that is, by calculating the ratio of the difference between the position forecast amount and the actual position amount to the position forecast amount, the daily forecast reference deviation degree is obtained, and the score corresponding to the forecast reference deviation degree is determined to obtain the daily position forecast score , and then calculate the average score of the position forecast during this period, and use the average score as the quality evaluation value of the position forecast during this period.

However, at present, when evaluating the quality of position forecasts, the relationship between the position forecast amount and the actual position amount is simply considered, and it does not fully consider special scenarios such as depositors' withdrawals or malicious marketing by competitors. The situation of large fluctuations has led to a relatively simple implementation process for the quality assessment of position forecasting, resulting in inaccurate quality assessment of the position forecast process.

Contents of the invention

In order to solve the above technical problems, the embodiment of the present application expects to provide a position forecast evaluation method, device, equipment and storage medium, which solves the problem that the current implementation process of evaluating the quality of position forecast is relatively simple, and realizes a method for assessing the position forecast. The implementation method for evaluating the quality of the forecast enriches the instruction evaluation method for the position forecast and improves the accuracy of evaluating the quality of the position forecast.

The technical scheme of the present application is realized like this:

In the first aspect, a position forecast evaluation method, the method includes:

Determine the actual amount of m first positions of the object to be evaluated within the quality evaluation cycle; where m is an integer greater than or equal to 1;

From the actual actual amount of the m first positions, determine the actual amount of n second positions that are in an abnormal state; wherein, n is an integer greater than or equal to 0;

Based on the n actual occurrences of the second positions and the m actual occurrences of the first positions, determine the corresponding quality evaluation results of the object to be evaluated within the quality evaluation cycle; wherein, the quality evaluation results are used Yu represents the forecast quality of the position forecast for the object to be assessed within the quality assessment period.

In the second aspect, a position forecast and evaluation device, the device includes: a determination unit and a processing unit; wherein:

The determination unit is used to determine the actual amount of m first positions of the object to be assessed within the quality assessment period; wherein, m is an integer greater than or equal to 1;

The processing unit is configured to determine the actual amount of n second positions in an abnormal state from the m actual amount of the first position; wherein, n is an integer greater than or equal to 0;

The processing unit is further configured to determine the corresponding quality evaluation result of the object to be evaluated within the quality evaluation cycle based on the n actual occurrences of the second position and the m actual occurrences of the first position; Wherein, the quality assessment result is used to represent the forecast quality of the position forecast for the object to be assessed within the quality assessment period.

In the third aspect, a position forecast evaluation device, the device includes: a memory, a processor, and a communication bus; wherein:

The memory is used to store executable instructions;

The communication bus is used to realize the communication connection between the processor and the memory;

The processor is configured to execute the position forecast evaluation program stored in the memory, so as to realize the steps of the position forecast evaluation method described in any one of the above.

In a fourth aspect, a storage medium stores a position forecast evaluation program, and when the position forecast evaluation program is executed by a processor, the steps of the position forecast evaluation method described in any one of the above are implemented.

In the embodiment of this application, after determining the actual amount of m first positions of the object to be evaluated within the quality evaluation period, from the actual amount of m first positions, determine the actual amount of n second positions that are in an abnormal state amount, and based on the actual amount of n second positions and the actual amount of m first positions, to determine the corresponding quality evaluation results of the object to be evaluated in the quality evaluation cycle, so that m of the object to be evaluated in the quality evaluation cycle Analyze the actual amount of n second positions and the actual amount of m first positions that are in an abnormal state in the actual amount of the first position to determine the quality evaluation results for the objects to be evaluated in the quality evaluation cycle, and solve the current problem. The implementation process of evaluating the quality of position forecast is relatively simple, and a realization method of evaluating the quality of position forecast is realized, which enriches the instruction evaluation method of position forecast and improves the quality of position forecast evaluation. accuracy.

Description of drawings

Fig. 1 is a schematic flow chart of a position forecast evaluation method provided by the embodiment of the present application;

Fig. 2 is a schematic flow chart of another position forecast evaluation method provided by the embodiment of the present application;

Fig. 3 is a schematic flow chart of another position forecast evaluation method provided by the embodiment of the present application;

Fig. 4 is a schematic diagram of the implementation flow of a position forecast evaluation method provided by the embodiment of the present application;

FIG. 5 is a schematic flow chart of data processing provided by an embodiment of the present application;

Fig. 6 is a schematic flow chart of applying position forecast scoring provided by the embodiment of the present application;

Fig. 7 is a schematic flow chart of a position forecast evaluation method provided by another embodiment of the present application;

Fig. 8 is a schematic flow chart of another position forecast evaluation method provided by another embodiment of the present application;

FIG. 9 is a schematic structural diagram of a position forecasting and evaluating device provided in an embodiment of the present application;

FIG. 10 is a schematic structural diagram of a position forecasting and evaluating device provided in an embodiment of the present application.

detailed description

In order to make the purpose, technical solutions and advantages of the application clearer, the application will be further described in detail below in conjunction with the accompanying drawings. All other embodiments obtained under the premise of creative labor belong to the scope of protection of this application.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.

The embodiment of the present application provides a position forecast evaluation method, as shown in FIG. 1, the method is applied to a position forecast evaluation device, and the method includes the following steps:

Step 101. Determine the actual amount of m first positions of the object to be evaluated within the quality evaluation cycle.

Wherein, m is an integer greater than or equal to 1.

In the embodiment of the present application, the object to be evaluated may be any business product in the commercial banking system that requires position forecasting. The quality evaluation cycle is a cycle for evaluating the quality of the forecast result of the position forecast of the object to be evaluated, for example, it may be one week, 15 days, or one month. The actual amount of the first position is the actual position generated during the actual transaction process of the object to be evaluated. It should be noted that the actual amount of the m first positions of the object to be evaluated in the quality evaluation cycle can be obtained by other forecasting equipment during the actual operation of the object to be evaluated during the actual operation process, wherein, Forecasting the object to be evaluated can be multiple collections at multiple fixed times every day. The position forecasting and evaluating device is usually a management device in a commercial banking system, such as a server.

Step 102. From the actual actual amount of m first positions, determine the actual amount of n second positions in an abnormal state.

Wherein, n is an integer greater than or equal to 0.

In the embodiment of the present application, the m actual occurrences of the first positions are analyzed, and the n actual occurrences of the second positions that are obviously abnormal, that is, in an abnormal state, are determined from the m actual occurrences of the first positions. In some application scenarios, the actual amount of the second position in an abnormal state may not exist among the m actual amount of the first position, that is, n is zero.

Step 103: Based on the actual amount of n second positions and the actual amount of m first positions, determine the corresponding quality evaluation result of the object to be evaluated within the quality evaluation cycle.

Wherein, the quality evaluation result is used to represent the forecast quality of the position forecast for the object to be evaluated within the quality evaluation period.

In the embodiment of the present application, the determined n second position actual occurrences and m first position actual occurrences are analyzed to determine the quality of the forecast quality used to evaluate the position forecast of the object to be evaluated within the quality evaluation period evaluation result. The obtained quality assessment results can be stored, so that downstream data users can obtain the corresponding quality assessment results and further process the quality assessment results, such as large-screen display, performance assessment and/or position management, etc.

Based on the foregoing embodiments, the embodiments of the present application provide a position forecast evaluation method, as shown in FIG. 2 , the method is applied to a position forecast evaluation device, and the method includes the following steps:

Step 201. Determine the actual amount of m first positions of the object to be evaluated within the quality evaluation period.

Wherein, m is an integer greater than or equal to 1.

In the embodiment of this application, the quality evaluation period is one week, and the object to be evaluated is product A as an example. The actual amount of all positions collected during the position forecast of product A within one week, that is, the mth The actual amount of a position.

Step 202: Use p kinds of abnormal value detection methods to detect the actual amount of positions in an abnormal state among the m actual amount of the first position, and obtain the actual amount of the p group of q third positions.

Wherein, p is an integer greater than or equal to 1, and q is an integer greater than or equal to 0.

In this embodiment of the application, after determining the P outlier detection methods, each outlier detection method is used to analyze and process the actual occurrences of the m first positions to obtain the actual occurrences of the P group of q third positions, Wherein, the q values in the actual amount of the q third positions in each group may be the same or different, and the q values corresponding to each group are specifically determined by each outlier detection method. For example, assume that two outlier detection methods are determined, including method 1 and method 2. Correspondingly, method 1 is used to analyze and process the actual amount of the m first positions of product A, and obtain q1 actual amount of the third position For the actual amount of occurrence, method 2 is used to analyze and process the actual amount of m first positions of product A to obtain the actual amount of q2 third positions.

Step 203: Determine the actual amount of positions included in the actual amount of q third positions in p groups, and obtain n actual amounts of second positions.

Wherein, n is less than or equal to q, and n is an integer greater than or equal to 0.

In the embodiment of the present application, the actual amount of positions included in groups P is determined from the actual amount of q third positions in groups p, and n actual amounts of second positions are obtained. Exemplarily, assuming that the actual amount of q1 third positions is 4 actual amounts of third positions c1, c2, c3 and c4, and the actual amount of q2 third positions is 6 actual amounts of third positions c1 and c3 , c4, c6, c7 and c8, therefore, determined from the 4 third position actual amounts c1, c2, c3 and c4, and the 6 third position actual amounts c1, c3, c4, c6, c7 and c8 It is obtained that the actual amount of the third position included in the actual amount of the third position of these two groups is c1, c3 and c4, that is, the actual amount of the n second positions of product A is 3 actual amounts of the second position c1, c3 and c4.

Step 204, based on the n actual amounts of the second positions and the m actual amounts of the first positions, determine the quality evaluation results corresponding to the objects to be evaluated within the quality evaluation cycle.

In this embodiment of the application, the forecast quality analysis is performed on the actual occurrences of the n second positions and the m actual occurrences of the first positions to obtain the corresponding quality evaluation results of the objects to be evaluated within the quality evaluation period. Exemplarily, assuming that the actual amount of m first positions is the actual amount of 10 first positions c1, c2, c3, c4, c5, c6, c7, c8, c9 and c10, in this way, based on 3 second positions By analyzing the actual amount c1, c3, and c4 and the actual amount c1, c2, c3, c4, c5, c6, c7, c8, c9, and c10 of the 10 first positions, it can be determined that the corresponding amount of product A in the week Quality assessment results.

Based on the foregoing embodiments, in other embodiments of the present application, step 204 may be implemented by steps 204a-204d:

Step 204a. Determine a first ratio of n to m.

In the embodiment of the present application, the first ratio=n/m.

Step 204b. Obtain m first position forecast sample data corresponding to m first position actual occurrence amounts.

In the embodiment of the present application, the first position forecast sample data is the position forecast sample data obtained from a certain position forecast of the object to be evaluated. It should be noted that the m first position forecast sample data of the object to be evaluated in the quality evaluation period can be obtained by other forecasting equipment during the quality evaluation period to predict the object to be evaluated, or it can be obtained by the position forecast evaluation device during the quality evaluation It is obtained by forecasting the object to be evaluated within the period. When performing position forecasting, it is assumed that the position forecast data obtained from the forecast from the current period to the next period is the first position forecast sample data, and in the actual operation process, the position actually occurred between the current period and the next period It is recorded as the actual amount of the position corresponding to the first position forecast sample data, that is, one first position forecast sample data corresponds to one actual first position amount. In this way, for the m actual occurrences of the first positions, the first position forecast sample data corresponding to each first position actual occurrence is determined to obtain m first position forecast sample data.

Exemplarily, for the 10 actual first position amounts c1, c2, c3, c4, c5, c6, c7, c8, c9 and c10 of product A, the first position forecast sample data d1 corresponding to c1 is determined sequentially , c2 corresponding to the first position forecast sample data d2, c3 corresponding to the first position forecast sample data d3, c4 corresponding to the first position forecast sample data d4, c5 corresponding to the first position forecast sample data d5, c6 corresponding to the first The position forecast sample data d6, c7 correspond to the first position forecast sample data d7, c8, the first position forecast sample data d8, c9 correspond to the first position forecast sample data d9, and c10 correspond to the first position forecast sample data d10.

Step 204c: Determine m reference deviation degrees based on the m actual amounts of the first positions and the m first position forecast sample data.

In the embodiment of the present application, based on the m actual amount of the first position and the m first position forecast sample data, the degree of deviation between the actual amount of each first position and the corresponding first position forecast sample data is calculated, Get m reference deviation degrees. Exemplarily, based on the first position forecast sample data d1 corresponding to c1 and c1, determine and obtain the reference deviation degree corresponding to c1; based on the first position forecast sample data d2 corresponding to c2 and c2, determine and obtain the reference deviation degree corresponding to c2; Based on the first position forecast sample data d3 corresponding to c3 and c3, determine the reference deviation degree corresponding to c3, and so on, until the reference deviation degree corresponding to c10 is determined based on the first position forecast sample data d10 corresponding to c10 and c10 .

Step 204d, based on the first ratio and the m reference deviation degrees, determine a quality assessment result.

In the embodiment of the present application, the quality evaluation result for the object to be evaluated is determined based on the m reference deviation degrees and the first ratio, ie, n/m.

Based on the foregoing embodiments, in other embodiments of the present application, step 204c may be implemented by the following steps: through the formula

It is determined to obtain m reference deviation degrees; among them, i=1, 2..., m, P _{actual i} is the actual amount of the i-th first position among the actual amount of the m first positions, and P _{pre i} is the actual amount of the i-th position The first position forecast sample data corresponding to the actual amount of the first position, D _i is the reference deviation degree corresponding to the actual amount of the ith first position.

Based on the foregoing embodiments, in other embodiments of the present application, step 204d may be implemented by steps a11-a15:

Step a11, from the m reference deviation degrees, determine the reference deviation degrees smaller than the preset reference deviation degree threshold, and obtain x target deviation degrees.

Wherein, x is an integer greater than or equal to 0 and less than or equal to m.

In the embodiment of the present application, the preset reference deviation threshold is an empirical value obtained from a large number of experiments, which can be continuously corrected according to actual conditions. Comparing each of the m reference deviation degrees with a preset reference deviation degree threshold, to determine the reference deviation degree smaller than the preset reference deviation degree threshold from the m reference deviation degrees as the target deviation degree . In some application scenarios, the reference deviation degrees among the m reference deviation degrees may all be greater than the preset reference deviation degree threshold, that is, none of the target deviation degrees exists.

Step a12. Determine x first forecast score values corresponding to x target deviation degrees.

In the embodiment of the present application, when determining x first forecast score values corresponding to x target deviation degrees, it may be determined according to the relationship between the deviation degree and the forecast score value, and the difference between the deviation degree and the forecast score value The relationship between is usually an empirical relationship obtained from a large number of experiments or user experience. The relationship between the deviation degree and the forecast score value can be an algorithmic formula between the deviation degree and the forecast score value, or it can be a relationship list including the deviation degree corresponding to different forecast score values, which can be determined by the actual situation. There is no specific limit.

Exemplarily, assuming that there are three degrees of deviation: degree of deviation 1, degree of deviation 2 and degree of deviation 3, the corresponding determination obtains that the first forecast score value corresponding to degree of deviation 1 is score 1, and the first forecast value corresponding to degree of deviation 2 is The score value is 2 points, and the first forecast score value corresponding to the deviation degree 3 is 3 points.

Step a13. From the m actual amounts of the first positions, determine y additional actual amounts of the fourth positions in addition to the actual amounts of the n second positions.

Among them, the sum of n and y is m.

In this embodiment of the application, the actual amount of the fourth position in a normal state is determined from the actual amount of the m first positions, that is, the actual amount of the n second positions included in the field of the actual amount of the m first positions If the amount is removed, the actual amount of y fourth positions in the normal state can be obtained.

Exemplarily, for the 10 actual occurrences of the first position of product A, c1, c2, c3, c4, c5, c6, c7, c8, c9 and c10, since it is determined that the 3 second positions in the abnormal state actually occurred c1, c3 and c4, therefore, it is possible to determine the actual amounts c2, c5, c6, c7, c8, c9 and c10 corresponding to the product A corresponding to the 7 fourth positions in a normal state.

Step a14, determining y second forecast score values corresponding to y actual amounts of the fourth positions.

In this embodiment of the application, the implementation process of determining y second forecast score values corresponding to y fourth position actual occurrences can refer to the realization of x first forecast score values corresponding to x target deviation degrees in step a12 The process will not be described in detail here.

Step a15, through the formula

Determine the quality assessment result S.

Wherein, θ is the first ratio, S _1i is the i-th first predicted score value among the x first predicted score values, and S _2j is the j-th second predicted score value among the y second predicted score values.

In this way, the outlier detection algorithm is used to identify the special events in the forecast, that is, the position sample data in an abnormal state, and then filter the forecasts of these special events by the degree of deviation corresponding to these special events, so as to predict the qualified events. The score of the score is assigned a smaller weight, so that the score of this part occupies a lesser importance in the final overall product forecast score, which not only takes into account the difficulty of position forecasting under special circumstances, but also evaluates the accuracy of position forecasting under special circumstances. an effective measure.

Based on the foregoing embodiments, referring to FIG. 3 , after executing step 204, the position forecasting and evaluating device is also used to execute step 205:

Step 205, storing the quality evaluation result in the message queue corresponding to the object to be evaluated.

In the embodiment of the present application, the corresponding quality assessment results of the objects to be assessed within the period to be assessed are stored in the message queue, so that downstream data users can obtain the quality assessment results from the message queue for subsequent analysis and application.

Based on the aforementioned embodiments, the embodiment of the present application provides an application embodiment of a position forecasting and evaluation method, and the specific implementation process of the position forecasting and evaluation device can refer to Figure 4, including:

Step 301, obtaining position forecast data of each product.

Wherein, the position forecast data of each product may be obtained from a database.

Step 302, grouping process.

Among them, the obtained position forecast data of each product is grouped according to the product dimension, and the grouped position forecast data sets of each product in each quality evaluation period are obtained, which is recorded as D={D1, D2,..., Dz} . Among them, D1 is the position forecast data set corresponding to product D1; D2 is the position forecast data set corresponding to product D2; ..., Dz is the position forecast data set corresponding to product Dz. Among them, D1 also includes position forecast data sets in multiple different quality assessment periods corresponding to product D1.

Step 303, data preprocessing process.

Among them, data preprocessing is performed on the position forecast data of each product after grouping processing, and the data preprocessing methods include processing missing values and/or replacing special characters in the data. In this way, after data preprocessing is performed on the position forecast data set of each product in the D set, the m actual occurrences of the first positions included in each quality evaluation period for each product can be obtained.

Step 304, an outlier detection process.

Among them, the outlier detection is performed on the position forecast data of each product after data preprocessing, that is, the outlier detection is performed on the m actual occurrences of the first positions included in each quality evaluation cycle of each product, and each product The actual amount of the first position in the abnormal state and the actual amount of the first position in the normal state included in each quality assessment period.

Step 305, weighted position forecast scoring process.

Among them, count the number of sessions of the actual amount of the first position that is in an abnormal state included in each pending quality assessment cycle for each product and the number of sessions of the actual amount of all first positions included in each pending quality assessment cycle for each product Ratio θ; determine the degree of deviation between the actual amount of the first position corresponding to the abnormal state and the corresponding first position forecast sample data; and judge the degree of deviation corresponding to the actual amount of the first position corresponding to the abnormal state and the preset The relationship between the deviation degree thresholds; obtain the actual amount of the first position corresponding to the abnormal state that is less than the preset deviation degree threshold, and determine the actual amount of x first positions corresponding to the abnormal state that is less than the preset deviation degree threshold The corresponding forecast score, and the forecast score corresponding to the actual amount of y first positions determined to be in a normal state; use the formula

Calculate the position forecast summary score of a certain product in a period of pending quality assessment, where S _1i is the forecast score corresponding to the actual amount of the i-th first position corresponding to the abnormal state, and S _2j is the forecast score corresponding to the first position in the normal state The forecast score corresponding to the actual amount of j first positions.

Step 306, store in the message queue for use by the downstream data consumer.

Among them, write the calculated summary score of each position forecast into the message queue (Message Queue, MQ), and use different identification information such as group_id to identify the summary score of each position forecast, so that downstream data users can Use different group_ids to subscribe to the topic when the position forecast summary score is written, so that different downstream data users can further process the position forecast summary score in near real time, such as large-screen display, performance evaluation, or position management.

Wherein, the specific implementation process of the foregoing steps 301 to 303 can refer to that shown in FIG. 5 , including the following steps:

Step 301a, start.

Step 301b, using Structured Query Language (Structured Query Language, SQL) to read data and export it as a file.

Among them, use SQL to read the required data from the database, mainly including the following data: date, actual amount of the position, and the amount of the position forecast, which is the aforementioned sample data of the position forecast, and export the read data as, for example, the file type is comma-separated Value file format (Comma-Separated Values, CSV) file.

Step 301c, use computer programming language (Python) to group CSV files.

Among them, Python reads data from the CSV file as a distributed dataset (DataFrame) organized in named columns.

Step 301d, use Python to perform data preprocessing on the DataFrame.

Among them, the process of grouping and data preprocessing of CSV files by Python can be implemented with the following code:

In this way, through the above code, the format of the final output data is shown in Table 1 below:

Table 1 Product Details Position Forecast Score

Among them, ds is the date, y is the actual amount of the position, y_hat is the position forecast sample data, and score is the corresponding position forecast score.

Step 301e, end.

When performing outlier detection in the aforementioned step 304, the specific implementation process of an outlier detection method provided can refer to steps b11-b14:

Step b11. Determine the average value of the m actual amounts of the first positions.

Step b12, based on the actual amount of the m first positions and the average value, determine the standard deviation of the actual amount of the m first positions.

Step b13, respectively determine the first product of the first preset value and the standard deviation, and the second product of the second preset value and the standard deviation.

Wherein, both the first preset value and the second preset value are empirical values, the first preset value may be -3, and the second preset value may be 3.

Step b14. From the m actual first position amounts, determine the first position actual amounts that are smaller than the first product or greater than the second product, and obtain the first position actual amounts that are in an abnormal state, which can be recorded as V ₁ .

When performing outlier detection in the aforementioned step 304, the specific implementation process of another outlier detection method provided can refer to steps c11-c13:

Step c11: Sort the m actual balances of the first positions in descending order, and obtain the sorted m actual balances of the first positions.

Step c12, from the sorted m actual balances of the first positions, determine the actual balance of the first position at the first preset ratio position and the actual balance of the first position at the second preset ratio position.

Wherein, both the first preset ratio and the second preset ratio are empirical values. Further, the first preset ratio may be 25%, and the second preset ratio may be 75%.

Step c13. From the actual amount of the m first positions, determine the first positions that are smaller than the actual amount of the first position at the first preset ratio position or greater than the actual amount of the first position at the second preset ratio position The actual occurrence amount is to obtain the actual occurrence amount of the first position in an abnormal state, which can be recorded as V ₂ .

When performing outlier detection in the aforementioned step 304, the specific implementation process of another outlier detection method provided can refer to steps d11-d16:

Step d11. From the m actual amounts of the first positions, determine the distance between the actual amounts of each first position and the other m-1 actual amounts of the first positions.

Among them, the method of determining the distance between the actual amount of each first position and the actual amount of other m-1 first positions can be realized by using various distance algorithms, which is not specifically limited here.

Step d12. From the distance between the actual amount of each first position and the actual amount of other m-1 first positions, determine the first position whose actual amount of each first position is within the distance of K neighbors The actual occurrence amount, and the quantity N(i) of the actual occurrence amount of the first position within the range of K neighbor distances corresponding to each distance of the actual occurrence amount of the first position. Wherein, i=1, 2, ..., m, which represents the actual amount of the i-th first position among the actual amount of the m first positions.

Step d13, by formula

To determine the first value of the actual amount of the i-th first position. j is the actual amount of the jth first position among the m-1 actual amount of the first position in addition to the actual amount of the i-th first position among the actual amount of the first position.

Step d14. Determine the first value of the actual amount of the first position whose actual amount is within the range of K neighbor distances for each first position, and calculate the first value of the actual amount of each first position within the range of K neighbor distances. Calculate the average value of the first value of the actual amount of a position to obtain the first average value corresponding to the distance from the actual amount of each first position.

As an example, take the actual amount of the i-th first position among the actual amount of the m first positions as an example, assuming that the actual amount of the i-th first position within the range of K neighbors The occurrences are m1, m2, m3, m4 and mi, and the first values obtained by analyzing and calculating m1, m2, m3, m4 and mi through step d13 are respectively m1, m2, m3, m4 and mi, and for m1 , m2, m3, m4 and mi are averaged to obtain the first average value L(i) corresponding to the actual amount of the i-th first position.

Step d15, calculate and obtain the second value of the i-th actual amount of the first position through the formula LOF(i)=L(i)/lrd(i).

Step d16: Determine the actual amount of the n first positions whose second value is smaller than the third preset value as the actual amount of the first position in an abnormal state, denoted as V ₃ .

Wherein, the third preset value is an empirical value related to m.

When performing outlier detection in the aforementioned step 304, the specific implementation process of another outlier detection method provided can refer to steps e11-e16:

Step e11. Randomly select an actual position value Value1 from the training samples of the actual position value.

Step e12. Randomly select an actual position value Value2 from the actual position amount training samples again, and judge whether Value2 is greater than or equal to Value1. If not, go to step e13. If yes, go to step e14.

Step e13, set Value2 as the left child node of Value1, and then execute step e12.

Step e14, set Value2 as the right child node of Value1, and then execute step e12.

Step e15 , until the height of the obtained binary tree reaches the limited height, end the operation, repeat the training in this way, and obtain the target binary tree.

Step e16, using the target binary tree to classify the actual amount of the m first positions, and determine the actual amount of the first position in an abnormal state, denoted as V ₄ .

In this way, after V ₁ , V ₂ , V ₃ and V ₄ are obtained, the intersection of the four outlier sets V=V ₁ ∩V ₂ ∩V ₃ ∩V ₄ is taken as the final outlier, and the aforementioned nth The actual amount of the second position.

The specific implementation flow chart of the foregoing step 306 can be implemented with reference to FIG. 6 , including the following steps:

Step f11, determine to obtain the position forecast score.

Among them, the position forecast score is the aforementioned quality evaluation result.

Step f12, write to MQ.

Step f13, the downstream data processor 1 obtains the topic corresponding to the identifier 1; the downstream data processor 2 obtains the topic corresponding to the identifier 2; the downstream data processor 3 obtains the topic corresponding to the identifier 3.

Wherein, the topic corresponding to each identifier has an association relationship with the corresponding position forecast score.

It should be noted that, for descriptions of the same steps and content in this embodiment as in other embodiments, reference may be made to the descriptions in other embodiments, and details are not repeated here.

The embodiment of the present application provides a position forecast evaluation method, as shown in FIG. 7 , the method is applied to a position forecast evaluation device, and the method includes the following steps:

Step 401. Determine the actual amount of m first positions of the object to be evaluated within the quality evaluation period.

Wherein, m is an integer greater than or equal to 1.

Step 402, using p kinds of outlier detection methods to detect the actual amount of positions in an abnormal state among the actual amount of m first positions, and obtain the actual amount of q third positions in p groups,

Step 403: Determine the actual amount of the position included in the actual amount of the q third positions in the group p, and obtain the actual amount of n second positions.

Step 404. From the m actual amounts of the first positions, determine y additional actual amounts of the fourth positions in addition to the actual amounts of the n second positions.

Among them, the sum of n and y is m.

Step 405. Determine y second forecast score values corresponding to y actual amounts of the fourth positions.

Step 406. Determine the average value of y second forecast score values to obtain a quality evaluation result.

In the embodiment of the present application, the average value of the y second forecast score values is directly calculated to obtain the quality evaluation result of the object to be evaluated within the evaluation period. Wherein, the method of calculating the average value of the y second forecast score values can be realized by directly summing and then averaging, or by other weighted averaging methods, which can be determined according to the actual application scenario, No specific limitation is made here.

Based on the foregoing embodiments, in other embodiments of the present application, referring to FIG. 8 , the position forecasting and evaluation device is also used to perform step 407:

Step 407, storing the quality evaluation result in the message queue corresponding to the object to be evaluated.

Based on the foregoing embodiments, the embodiment of the present application provides a position forecast evaluation device, as shown in FIG. 9 , the position forecast evaluation device 5 may include: a determination unit 51 and a processing unit 52; wherein:

A determination unit 51, configured to determine the actual amount of m first positions of the object to be assessed within the quality assessment period; wherein, m is an integer greater than or equal to 1;

The processing unit 52 is configured to determine the actual amount of n second positions in an abnormal state from the actual amount of m first positions; wherein, n is an integer greater than or equal to 0;

The processing unit is also used to determine the corresponding quality evaluation result of the object to be evaluated in the quality evaluation cycle based on the actual amount of n second positions and the actual amount of m first positions; wherein, the quality evaluation result is used to represent the The forecast quality of the position forecast for the subject to be assessed within the assessment period.

In other embodiments of the present application, the processing unit 52 includes: a detection module and a determination module; wherein:

The detection module is configured to use p kinds of outlier detection methods to detect the actual amount of positions in an abnormal state among the actual amount of m first positions, and obtain the actual amount of q third positions in p groups; wherein, p is greater than or an integer equal to 1, and q is an integer greater than or equal to 0;

The determining module is used to determine the actual amount of the position included in the actual amount of q third positions in p groups, and obtain n actual amounts of the second position; wherein, n is less than or equal to q.

In other embodiments of the present application, the processing unit 52 further includes: a processing module; the processing module is specifically configured to implement the following steps:

determining a first ratio of n to m;

Obtain m first position forecast sample data corresponding to the actual amount of m first positions;

Determine m reference deviation degrees based on the m actual amount of the first position and the m first position forecast sample data;

Based on the first ratio and the m reference deviations, a quality assessment result is determined.

In other embodiments of the present application, when the processing module executes the step of determining m reference deviations based on m actual occurrences of the first positions and m first position forecast sample data, it can be specifically implemented through the following steps:

by formula

In other embodiments of the present application, when the processing module executes the step of determining the quality evaluation result based on the first ratio and the m reference deviation degrees, it may specifically be implemented by the following steps:

From the m reference deviation degrees, determine the reference deviation degree less than the preset reference deviation degree threshold, and obtain x target deviation degrees; wherein, x is an integer greater than or equal to 0 and less than or equal to m;

Determining x first forecast score values corresponding to x target deviation degrees;

From the m actual amount of the first position, determine the additional y actual amount of the fourth position in addition to the actual amount of the n second position; wherein, the sum of n and y is m;

Determine y second forecast score values corresponding to the actual amount of y fourth positions;

by formula

Determine the quality assessment result S; where, θ is the first ratio, S _1i is the i-th first forecast score value among the x first forecast score values, and S _2j is the j-th one of the y second forecast score values Second forecast score value.

In other embodiments of the present application, when p is equal to 1, the processing module is also specifically configured to implement the following steps:

The average value of the y second forecast score values is determined to obtain a quality evaluation result.

In other embodiments of the present application, after the processing unit 52, the device further includes: a storage unit; wherein:

The storage unit is used to store the quality evaluation result in the message queue corresponding to the object to be evaluated.

It should be noted that, for the information exchange process between units and modules in this embodiment, reference may be made to the implementation process of the methods provided in the embodiments corresponding to FIGS. 1-3 and FIGS.

Based on the foregoing embodiments, the embodiments of the present application provide a position forecast evaluation device, which can be applied to the position forecast evaluation methods provided in the embodiments corresponding to Figures 1-3 and Figures 7-8, refer to Figure 10 As shown, the position forecast evaluation device 6 may include: a processor 61, a memory 62 and a communication bus 63, wherein:

Memory 62, used to store executable instructions;

A communication bus 63 is used to realize the communication connection between the processor 61 and the memory 62;

The processor 61 is configured to execute the position forecast evaluation program stored in the memory 62, so as to realize the following steps:

From the actual actual amount of m first positions, determine the actual amount of n second positions in an abnormal state; wherein, n is an integer greater than or equal to 0;

Based on the n actual amounts of the second positions and the m actual amounts of the first positions, the corresponding quality evaluation results of the objects to be evaluated within the quality evaluation period are determined.

In other embodiments of the present application, when the processor 61 executes the step of determining the actual amount of the n second positions in an abnormal state from the actual amount of the m first positions, it may be implemented through the following steps:

Use p outlier detection methods to detect the actual amount of positions in an abnormal state among the actual amount of m first positions, and obtain the actual amount of q third positions in p groups; where p is an integer greater than or equal to 1, q is an integer greater than or equal to 0;

Determine the actual amount of the position included in the actual amount of the q third positions in the p group, and obtain n actual amounts of the second position; wherein, n is less than or equal to q.

In other embodiments of the present application, when the processor 61 executes the step of determining the corresponding quality evaluation result of the object to be evaluated within the quality evaluation period based on the actual amount of the n second positions and the actual amount of the m first positions, it can be passed Follow these steps to achieve:

determining a first ratio of n to m;

In other embodiments of the present application, when the processor 61 executes the step of determining m reference deviation degrees based on m actual occurrences of first positions and m first position forecast sample data, it may be implemented through the following steps:

by formula

In other embodiments of the present application, when the processor 61 executes the step of determining the quality assessment result based on the first ratio and the m reference deviation degrees, the following steps may be implemented:

by formula

In other embodiments of the present application, when p is equal to 1, the processor 61 executes steps to determine the corresponding quality evaluation of the object to be evaluated within the quality evaluation period based on the actual amount of n second positions and the actual amount of m first positions The results can be achieved by the following steps:

In other embodiments of the present application, after the processor 61 executes the step of determining the corresponding quality evaluation result of the object to be evaluated within the quality evaluation period based on the actual amount of the n second positions and the actual amount of the m first positions, it also uses to perform the following steps:

Store the quality evaluation result in the message queue corresponding to the object to be evaluated.

It should be noted that one or more programs in the embodiments of the present application can be explained by the steps of one or more processors, and the implementation process can be implemented by referring to the methods provided in the embodiments corresponding to Figures 1-3 and Figures 7-8, I won't repeat them here.

Based on the foregoing embodiments, the embodiments of the present application provide a computer-readable storage medium, referred to as a storage medium for short, where one or more programs are stored in the computer-readable storage medium, and the one or more programs can be used by one or more The processor executes to realize the implementation process of the position forecasting evaluation method provided in the embodiments corresponding to Figs. 1-3 and Figs. 7-8, which will not be repeated here.

The above are only examples of the present application, and are not intended to limit the scope of protection of the present application. Any modifications, equivalent replacements and improvements made within the spirit and scope of the present application are included in the protection scope of the present application.

Industrial Applicability

An embodiment of the present application provides a position forecast evaluation method, device, equipment, and storage medium. The method includes: determining the actual amount of m first positions of the object to be evaluated within the quality evaluation period; wherein, m is greater than or equal to 1 is an integer; from the m actual amount of the first position, determine the actual amount of n second positions in an abnormal state; wherein, n is an integer greater than or equal to 0; based on the actual amount of the n second positions The amount and the m actual amount of the first position, determine the quality evaluation result corresponding to the object to be evaluated in the quality evaluation cycle; wherein, the quality evaluation result is used to represent the The prediction quality of the position forecast for the object to be evaluated solves the problem that the current implementation process of evaluating the quality of the position forecast is relatively simple, and realizes a realization method for evaluating the quality of the position forecast, which enriches the quality of the position forecast. The command evaluation method of forecast improves the accuracy of evaluating the quality of position forecast.

Claims

A position forecast evaluation method, said method comprising:

Determine the actual amount of m first positions of the object to be evaluated within the quality evaluation cycle; where m is an integer greater than or equal to 1;

From the actual actual amount of the m first positions, determine the actual amount of n second positions that are in an abnormal state; wherein, n is an integer greater than or equal to 0;

Based on the n actual occurrences of the second positions and the m actual occurrences of the first positions, determine the corresponding quality evaluation results of the object to be evaluated within the quality evaluation cycle; wherein, the quality evaluation results are used Yu represents the forecast quality of the position forecast for the object to be assessed within the quality assessment period.
The method according to claim 1, wherein the determining the actual amount of the n second positions in an abnormal state from the m actual amount of the first position includes:

Using p kinds of outlier detection methods to detect the actual amount of positions in an abnormal state among the m actual amount of the first position, and obtain the actual amount of p groups of q third positions; where p is greater than or equal to 1 Integer, q is an integer greater than or equal to 0;

Determine the actual amount of the positions included in the actual amount of the q third positions in the p group, and obtain n actual amounts of the second position; wherein, n is less than or equal to q.
The method according to any one of claims 1 to 2, wherein, based on the n actual occurrences of the second position and the m actual occurrences of the first position, it is determined that the object to be evaluated is in the The corresponding quality assessment results within the quality assessment cycle, including:

determining a first ratio of n to m;

Obtain m first position forecast sample data corresponding to the m first position actual occurrence amounts;

Determine m reference deviation degrees based on m actual occurrences of the first position and m first position forecast sample data;

The quality evaluation result is determined based on the first ratio and the m reference deviation degrees.
The method according to claim 3, wherein said determining m reference deviation degrees based on the m actual amount of the first position and the m first position forecast sample data includes:

by formula
It is determined to obtain m reference deviation degrees; wherein, i=1, 2..., m, P actual i is the i-th actual amount of the first position among the m actual amount of the first position, P pre i is the first position forecast sample data corresponding to the i-th actual amount of the first position, and D i is the reference deviation degree corresponding to the i-th actual amount of the first position.
The method according to claim 3, wherein said determining the quality evaluation result based on the first ratio and the m reference deviation degrees comprises:

From the m reference deviation degrees, determine a reference deviation degree smaller than a preset reference deviation degree threshold to obtain x target deviation degrees; wherein, x is an integer greater than or equal to 0 and less than or equal to m;

determining x first forecast score values corresponding to the x target deviation degrees;

From the m actual amounts of the first positions, determine y additional actual amounts of the fourth positions in addition to the actual amounts of the n second positions; wherein, the sum of n and y is m;

Determine y second forecast score values corresponding to y actual occurrences of the fourth positions;

by formula
Determine the quality assessment result S; wherein, θ is the first ratio, S 1i is the i-th first forecast score value among the x first forecast score values, and S 2j is the y first forecast score value. The j-th second forecast score value among the second forecast score values.
The method according to any one of claims 1 to 2, wherein when p is equal to 1, the determination of the pending amount is based on n actual amounts of the second position and m actual amounts of the first position The corresponding quality assessment results of the assessment object within the quality assessment cycle, including:

From the m actual amounts of the first positions, determine y additional actual amounts of the fourth positions in addition to the actual amounts of the n second positions; wherein, the sum of n and y is m;

Determine y second forecast score values corresponding to y actual occurrences of the fourth positions;

Determining the average value of the y second forecast score values to obtain the quality evaluation result.
The method according to claim 1, wherein, based on the n actual occurrences of the second position and the m actual occurrences of the first position, it is determined that the corresponding value of the object to be evaluated within the quality evaluation period is After the quality assessment results, the method also includes:

The quality evaluation result is stored in a message queue corresponding to the object to be evaluated.
A position forecast evaluation device, the device includes: a determination unit and a processing unit; wherein:

The determination unit is used to determine the actual amount of m first positions of the object to be assessed within the quality assessment period; wherein, m is an integer greater than or equal to 1;

The processing unit is configured to determine the actual amount of n second positions in an abnormal state from the m actual amount of the first position; wherein, n is an integer greater than or equal to 0;

The processing unit is further configured to determine the corresponding quality evaluation result of the object to be evaluated within the quality evaluation cycle based on the n actual occurrences of the second position and the m actual occurrences of the first position; Wherein, the quality assessment result is used to represent the forecast quality of the position forecast for the object to be assessed within the quality assessment period.
A position forecast evaluation device, the device includes: a memory, a processor and a communication bus; wherein:

The memory is used to store executable instructions;

The communication bus is used to realize the communication connection between the processor and the memory;

The processor is configured to execute the position forecast evaluation program stored in the memory, so as to realize the steps of the position forecast evaluation method according to any one of claims 1-7.
A storage medium, on which a position forecast evaluation program is stored, and when the position forecast evaluation program is executed by a processor, the steps of the position forecast evaluation method according to any one of claims 1 to 7 are realized.