Target object grading method, device, equipment and readable storage medium

Abstract

The application provides a target object grading method and device, equipment and readable storage medium, and relates to the technical field of data processing. The method comprises the following steps: determining a plurality of indexes of each target object according to business requirements; for each index of each target object, analyzing a calculation formula of the index of the target object, obtaining a score value of the index, and calculating a score value of each target object based on the score values of the plurality of indexes; generating a plurality of grading schemes from a plurality of target objects according to a number of to-be-graded levels, wherein each grading scheme comprises a plurality of levels, and each level comprises a plurality of target objects; calculating the size-weighted distance of each grading scheme; and determining the optimal grading scheme based on the size-weighted distances of the plurality of grading schemes by using a differential evolution algorithm. The application can realize grading of target objects with high efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, device, and readable storage medium for classifying target objects. Background Technology

[0002] Banks generate thousands of target entities during their operations, such as operational entities. To facilitate performance evaluation and rating, it is crucial to categorize these entities. Each entity has a large amount of data on its metrics, with varying implications for different business operations and different levels of importance for rating. If the rating of each target entity were to go through four stages—requirements analysis, development, deployment, and verification—the process would be lengthy and iterative. Summary of the Invention

[0003] This invention proposes a target object classification method to achieve high efficiency in classifying target objects. The method includes:

[0004] Based on business needs, determine multiple metrics for each target object;

[0005] For each indicator of each target object, the calculation formula of that indicator for that target object is parsed to obtain the score value of that indicator. Based on the score values ​​of multiple indicators, the score value of each target object is calculated.

[0006] Based on the number of levels to be classified, multiple classification schemes are generated for multiple target objects. Each classification scheme includes multiple levels, and each level includes multiple target objects.

[0007] Calculate the size-weighted distance for each hierarchical scheme;

[0008] The differential evolution algorithm is used to determine the optimal hierarchical scheme based on the size-weighted distance of multiple hierarchical schemes.

[0009] This invention provides a target object classification device for achieving high efficiency in classifying target objects. The device includes:

[0010] The metrics determination module is used to determine multiple metrics for each target object based on business needs.

[0011] The rating calculation module is used to parse the calculation formula of each indicator for each target object, obtain the rating value of the indicator, and calculate the rating value of each target object based on the rating values ​​of multiple indicators.

[0012] The grading scheme acquisition module is used to generate multiple grading schemes from multiple target objects based on the number of grades to be graded. Each grading scheme includes multiple grades, and each grade includes multiple target objects.

[0013] The scale-weighted distance calculation module is used to calculate the scale-weighted distance for each hierarchical scheme;

[0014] The optimal grading scheme determination module is used to determine the optimal grading scheme by employing a differential evolution algorithm based on the size-weighted distance of multiple grading schemes.

[0015] This invention also proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described target object classification method.

[0016] This invention also proposes a computer-readable storage medium storing a computer program that performs the above-described target object classification method.

[0017] In this embodiment of the invention, multiple indicators are determined for each target object based on business requirements. For each indicator of each target object, the calculation formula for that indicator is parsed to obtain its score. Based on the scores of multiple indicators, the score of each target object is calculated. Multiple grading schemes are generated for the multiple target objects according to the number of levels to be graded, wherein each grading scheme includes multiple levels, and each level includes multiple target objects. The scale-weighted distance of each grading scheme is calculated. A differential evolution algorithm is used to determine the optimal grading scheme based on the scale-weighted distance of the multiple grading schemes. The beneficial effects achieved in the above process are as follows:

[0018] 1. High real-time performance: From obtaining the indicator calculation formula to the visualization of the hierarchical results, the feedback results can be seen instantly, which facilitates rapid business iteration.

[0019] 2. High scalability: It can quickly expand the built-in operators, optional indicators, and intra-class size-weighted distance formulas to flexibly support new business needs.

[0020] 3. High flexibility: Based on the built-in indicator library and operator library, business personnel can flexibly define various indicator calculation formulas.

[0021] 4. Strong business interpretability: The calculation formula of the indicator is directly defined according to the business meaning, which makes it highly interpretable. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0023] Figure 1 This is a flowchart of the target object classification method in an embodiment of the present invention;

[0024] Figure 2 This is a flowchart of the reverse Polish notation algorithm in an embodiment of the present invention;

[0025] Figure 3 This is a flowchart of the differential evolution algorithm calculation in an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram of the target object grading device in an embodiment of the present invention;

[0027] Figure 5 This is another schematic diagram of the target object grading device in an embodiment of the present invention;

[0028] Figure 6 This is a schematic diagram of a computer device in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0030] In the description of this specification, the terms "comprising," "including," "having," and "containing" are open-ended terms, meaning that they include but are not limited to. The terms "an embodiment," "a specific embodiment," "some embodiments," and "for example," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example that are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. The order of steps involved in the various embodiments is used to illustrate the implementation of this application, and the order of steps is not limited and can be adjusted appropriately as needed.

[0031] First, the terminology involved in the embodiments of the present invention will be explained.

[0032] Unsupervised learning: Machine learning where samples are not labeled.

[0033] Reverse Polish notation: a type of postfix expression that is well-suited for machine execution and can mask the precedence of operators caused by parentheses.

[0034] Evolutionary algorithms are a general term for a class of metaheuristic algorithms that draw on the evolution, selection, and elimination mechanisms of organisms in nature. They typically start by creating a population, which then evolves and is eliminated until superior individuals are produced as the target solution.

[0035] Differential Evolutionary Algorithm: A population-based evolutionary algorithm that simulates the process of cooperation and competition among individuals in a population.

[0036] Figure 1 This is a flowchart of the target object hierarchical method in an embodiment of the present invention, such as... Figure 1 As shown, the method includes:

[0037] Step 101: Based on business needs, determine multiple metrics for each target object;

[0038] Step 102: For each indicator of each target object, parse the calculation formula of that indicator for that target object to obtain the score value of that indicator. Based on the score values ​​of multiple indicators, calculate the score value of each target object.

[0039] Step 103: Based on the number of levels to be classified, generate multiple classification schemes for multiple target objects, wherein each classification scheme includes multiple levels, and each level includes multiple target objects.

[0040] Step 104: Calculate the scale-weighted distance for each hierarchical scheme;

[0041] Step 105: Using the differential evolution algorithm, the optimal grading scheme is determined based on the size-weighted distance of multiple grading schemes.

[0042] In this embodiment of the invention, the time complexity of the scale-weighted distance calculation is very low, which greatly speeds up the evaluation of the merits of the grading scheme; the introduction of the evolutionary differential algorithm to search for the optimal grading scheme further improves the efficiency of determining the optimal grading scheme.

[0043] In practice, each target entity has multiple indicators. For example, the target entity of Yutong Service Point has risk-related indicators, performance-related indicators, and operational indicators in four aspects: agricultural assistance services, digital customer acquisition, assets and liabilities, and non-financial services. The business meaning of each indicator is different, and its importance to rating and assessment varies. At this time, business personnel need to select the indicators they are interested in from the system's indicator database in a visual way according to actual needs.

[0044] In one embodiment, the calculation formula for each indicator is written using predefined operators according to business requirements.

[0045] In the above embodiments, predefined operators include addition, subtraction, multiplication, division, squaring, square root, maximum / minimum truncation, log, etc.

[0046] After assigning a symbolic name to the indicator selected in the previous step, use the above operators to write the calculation formula for the indicator according to business requirements.

[0047] In one embodiment, parsing the calculation formula of the indicator of the target object includes:

[0048] The calculation formula for this index of the target object is parsed using the Reverse Polish Notation algorithm.

[0049] Figure 2 The flowchart of the Reverse Polish Notation algorithm in this embodiment of the invention is as follows:

[0050] Step 201: Scan the calculation formula of the indicator (also known as the infix expression) from left to right to obtain the read content. If the read content is an operand, proceed to step 202; if the read content is an operator, proceed to step 203.

[0051] Step 202: Determine the type of the operand and store it in the operand stack; Proceed to step 207;

[0052] Step 203: Determine the operator being read. If it is a left parenthesis "(", proceed to step 204; if it is a right parenthesis "")", proceed to step 205; if it is not a parenthesis, proceed to step 206.

[0053] Step 204: Store the operator into the operator stack; Proceed to step 207;

[0054] Step 205: Output the operators in the operator stack to the operand stack until a left parenthesis is encountered, then pop the left parenthesis "("; Proceed to step 207;

[0055] Step 206: If the operator at the top of the operator stack is a left parenthesis, the read operator is directly stored into the operator stack; if the read operator has a higher priority than the operator at the top of the operator stack, the read operator is directly stored into the operator stack; if the read operator has a lower priority than or equal to the operator at the top of the operator stack, the top operator is output to the operand stack, until the top operator of the operator stack has a lower priority than (but not equal to) that operator, or is a left parenthesis, the read operator is pushed onto the operator stack; proceed to step 207.

[0056] Step 207: When the expression has been read and there are still operators in the operator stack, the operators are sequentially removed and added to the operand stack until the operator stack is empty.

[0057] Step 208: Output the top value of the operand stack as the score of the indicator.

[0058] In one embodiment, multiple hierarchical schemes are generated for multiple target objects based on the number of levels to be classified, including:

[0059] Sort all target objects according to their rating scores from smallest to largest;

[0060] The sorted target objects are divided into several sets of levels to be classified, with each set corresponding to a level. Different segments can form multiple classification schemes.

[0061] In practice, the goal of dividing the data into M levels is to control the percentage range of target objects included in each level while fully considering the characteristics of the overall distribution of the scoring data, so as to achieve a balance between scientific rigor and business implications.

[0062] Assuming the number of target objects is N and the number of levels is M, the total number of possible grading schemes is N×(N-1)×(N-2)×Λ×(N-M+1). When N is large, the total number of possible grading schemes is huge, making it impossible to achieve the real-time calculation requirement through brute-force evaluation. In this embodiment of the invention, the evolutionary differential algorithm is used to achieve real-time calculation of grading.

[0063] In one embodiment, calculating the scale-weighted distance for each grading scheme includes:

[0064] For each grading scheme, calculate the percentage difference of each level in the grading scheme, and calculate the intra-class distance of each level in the grading scheme;

[0065] Calculate the size-weighted distance for each grading scheme based on the percentage difference of each level and the intra-class distance in each grading scheme.

[0066] In the above embodiments, the percentage difference refers to the difference between the percentage of samples included in each level and the target percentage, and the intra-class distance refers to the overall difference in the rating values ​​of the target objects included in each level.

[0067] In one embodiment, the percentage difference of each level in the grading scheme is calculated using the following formula:

[0068] Diff i =|pi-pi′|

[0069] Among them, Diff i pi is the percentage difference of level i; pi is the percentage of target objects in level i out of the total number of target objects; pi′ is the target percentage.

[0070] The number of samples in each level can be enormous, and direct calculation has a time complexity of O(N), which is computationally expensive. To accelerate the calculation of intra-class distances for each level, an accumulation array is used to reduce the time complexity to O(logN), as follows.

[0071] In one embodiment, calculating the intra-class distance for each level in the grading scheme includes:

[0072] For each level of each grading scheme, based on the score array S of that level, a cumulative array C is constructed for that level, where the score array is an array formed by sorting the score values ​​of multiple target objects for that level in ascending order; wherein,

[0073] Based on the rating array and accumulated data, determine the mean of the rating array for that level within a preset range, where the preset range is between a first value and a second value; the mean of the rating array for level i within the preset range is Mean. i It can be represented as

[0074] The time complexity of this step, which uses the binary search method to find the insertion position of the mean in the score array, is log(yx).

[0075] Calculate the intra-class distance for this level based on the first value and the insertion position;

[0076] Calculate the intra-class distance for this level based on the insertion position and the second value;

[0077] Calculate the intra-class distance for this level based on the intra-class distance of the first class and the intra-class distance of the second class.

[0078] The overall complexity of the above calculation process is log(N).

[0079] In one embodiment, the first intra-class distance of the level is calculated using the following formula, based on a first numerical value and the insertion position:

[0080]

[0081] Among them, Dist i (x, z) represents the first intraclass distance of rank i; x is the first numerical value, and z is the insertion position; Mean i C[z] is the mean of the score array for level i within a preset range; C[x] is the value of the cumulative array at the insertion position; C[x] is the value of the cumulative array at the first value.

[0082] The intra-class distance for this level is calculated using the following formula, based on the insertion position and the second value:

[0083]

[0084] Among them, Dist i (z,y) is the second-class intra-distance of class i;

[0085] The intra-class distance for this level is calculated using the following formula, based on the intra-class distances of the first and second classes:

[0086]

[0087] Among them, Dist i Let be the intra-class distance of level i.

[0088] In one embodiment, the size-weighted distance of each grading scheme is calculated using the following formula, based on the difference in the proportion of each level and the intra-class distance in each grading scheme:

[0089]

[0090] Where D is the size-weighted distance for each hierarchical scheme; Diff i The difference in the proportion of level i in each grading scheme; Dist i is the intra-class distance of level i in each grading scheme; w is the weighting factor.

[0091] It should be noted that the weighting factor, used to adjust the weighting influence of percentages, is generally set to 1. (This is just a linear weighting scheme; more complex non-linear weighting schemes can be adopted according to business needs). The smaller the size-weighted distance, the smaller the intra-class distance, and the smaller the difference from the percentages set by business personnel for each level. The smaller this value, the better the grading scheme.

[0092] The objective of this invention is to find the hierarchical scheme with the smallest size-weighted distance from all possible hierarchical schemes. Since the number of possible hierarchical schemes is enormous, and real-time response to user requests is required, a brute-force approach is not feasible. This invention proposes using an evolutionary differential algorithm to quickly find an approximately optimal hierarchical scheme.

[0093] The Differential Evolution Algorithm (DE) is a population-based evolutionary algorithm that simulates the cooperation and competition among individuals in a population. The algorithm is simple in principle, has few control parameters (only crossover probability and scaling factor), is robust, and is easy to implement.

[0094] In the differential evolution algorithm, the genes of each individual represent a candidate solution to the problem to be solved. The embodiments of this invention employ the following gene encoding method:

[0095] For a sorted array S of object ratings, with N objects and M levels, the boundary index value of each level is represented by (M-1) integers in the range [0, N], sorted from smallest to largest.

[0096] That is, the gene is X = (v1, v2, Λ, vM ), where 0≤v1≤v2≤Λ≤v M ≤N

[0097] This gene encoding method only involves boundary constraints.

[0098] Figure 3 The flowchart of the differential evolution algorithm in this embodiment of the invention includes:

[0099] Step 301, Initialize the population: Randomly generate genes for PN individuals, and let PN be the number of individuals in the differential evolution algorithm population;

[0100] Step 302, Mutation: Randomly select 3 individuals X r1 X r2 X r3 Calculate the mutated gene using the following formula;

[0101] U = X r1 +F(X r2 -X r3 ), where U is the mutated gene and F is the scaling factor, which is generally set to 0.5;

[0102] Step 303, Crossover: First, generate a random number 1≤d rand ≤M, then randomly select an individual Xr, and now cross the genes of Xr and U to generate a new individual W = (w1, w2, Λ, w M The formula for the generation process is as follows:

[0103] Where CR is the crossover probability;

[0104] Step 304, calculate the fitness function: calculate the size-weighted distance between the two hierarchical schemes represented by the genes of W and Xr;

[0105] Step 305, Selection: Select the individual with the smaller size-weighted distance value between the two grading schemes represented by the genes of W and Xr, and discard the other individual;

[0106] Step 306, determine whether to end: If the optimal individual has not changed after a specified number of iterations, then end the iteration and select the scheme represented by the gene of the optimal individual in the population as the optimal grading scheme.

[0107] Note: If time permits, the diversity of individuals in the population can be increased by shrinking the scaling factor F and increasing the crossover factor CR, which may lead to a better solution.

[0108] In one embodiment, the method further includes:

[0109] The optimal grading scheme is visualized, allowing business personnel to perform manual analysis. If the grading results do not meet business needs, the above steps can be quickly repeated for a new round of adjustments.

[0110] This invention also proposes a target object classification device, the principle of which is similar to the target object classification method, and will not be described in detail here.

[0111] Figure 4 This is a schematic diagram of the target object grading device in an embodiment of the present invention, such as... Figure 4 As shown, the device includes:

[0112] The indicator determination module 401 is used to determine multiple indicators for each target object based on business needs;

[0113] The rating calculation module 402 is used to parse the calculation formula of each indicator of each target object, obtain the rating value of the indicator, and calculate the rating value of each target object based on the rating values ​​of multiple indicators.

[0114] The grading scheme acquisition module 403 is used to generate multiple grading schemes from multiple target objects based on the number of grades to be graded, wherein each grading scheme includes multiple grades, and each grade includes multiple target objects.

[0115] The scale-weighted distance calculation module 404 is used to calculate the scale-weighted distance for each hierarchical scheme;

[0116] The optimal grading scheme determination module 405 is used to determine the optimal grading scheme by employing a differential evolution algorithm based on the size-weighted distance of multiple grading schemes.

[0117] In one embodiment, the calculation formula for each indicator is written using predefined operators according to business requirements.

[0118] In one embodiment, the scoring calculation module is specifically used for:

[0119] The calculation formula for this index of the target object is parsed using the Reverse Polish Notation algorithm.

[0120] In one embodiment, the hierarchical scheme acquisition module is specifically used for:

[0121] Sort all target objects according to their rating scores from smallest to largest;

[0122] The sorted target objects are divided into several sets of levels to be classified, with each set corresponding to a level. Different segments can form multiple classification schemes.

[0123] In one embodiment, the scale-weighted distance calculation module is specifically used for:

[0124] For each grading scheme, calculate the percentage difference of each level in the grading scheme, and calculate the intra-class distance of each level in the grading scheme;

[0125] Calculate the size-weighted distance for each grading scheme based on the percentage difference of each level and the intra-class distance in each grading scheme.

[0126] In one embodiment, the scale-weighted distance calculation module is specifically used for:

[0127] The percentage difference of each level in this grading scheme is calculated using the following formula:

[0128] Diff i =|pi-pi′|

[0129] Among them, Diff i pi is the percentage difference of level i; pi is the percentage of target objects in level i out of the total number of target objects; pi′ is the target percentage.

[0130] In one embodiment, the scale-weighted distance calculation module is specifically used for:

[0131] For each level of each grading scheme, an accumulation array for that level is constructed based on the score array for that level. The score array is an array formed by sorting the score values ​​of multiple target objects for that level in ascending order.

[0132] Based on the scoring array and the accumulated data, the mean value of the scoring array for that level within a preset range is determined, where the preset range is between a first value and a second value.

[0133] The insertion position of the mean in the score array is described using the binary search method;

[0134] Calculate the intra-class distance for this level based on the first value and the insertion position;

[0135] Calculate the intra-class distance for this level based on the insertion position and the second value;

[0136] Calculate the intra-class distance for this level based on the intra-class distance of the first class and the intra-class distance of the second class.

[0137] In one embodiment, the scale-weighted distance calculation module is specifically used for:

[0138] The intra-class distance for this level is calculated using the following formula, based on the first value and the insertion position:

[0139]

[0140] Among them, Dist i(x, z) represents the first intraclass distance of rank i; x is the first numerical value, and z is the insertion position; Mean i C[z] is the mean of the score array for level i within a preset range; C[x] is the value of the cumulative array at the insertion position; C[x] is the value of the cumulative array at the first value.

[0141] The intra-class distance for this level is calculated using the following formula, based on the insertion position and the second value:

[0142]

[0143] Among them, Dist i (z,y) is the second-class intra-distance of class i;

[0144] The intra-class distance for this level is calculated using the following formula, based on the intra-class distances of the first and second classes:

[0145]

[0146] Among them, Dist i Let be the intra-class distance of level i.

[0147] In one embodiment, the scale-weighted distance calculation module is specifically used for:

[0148] The following formula is used to calculate the size-weighted distance of each grading scheme based on the difference in the proportion of each level and the intra-class distance:

[0149]

[0150] Where D is the size-weighted distance for each hierarchical scheme; Diff i The difference in the proportion of level i in each grading scheme; Dist i is the intra-class distance of level i in each grading scheme; w is the weighting factor.

[0151] Figure 5 This is another schematic diagram of the target object classification device in an embodiment of the present invention; in one embodiment, it also includes a display module 406, used to: visually display the optimal classification scheme.

[0152] Embodiments of this application also provide a computer device. Figure 6 This is a schematic diagram of a computer device in an embodiment of the present invention. The computer device is capable of implementing all steps in the target object classification method described above. Specifically, the computer device includes the following components:

[0153] Processor 601, memory 602, communications interface 603, and communication bus 604;

[0154] The processor 601, memory 602, and communication interface 603 communicate with each other through the communication bus 604; the communication interface 603 is used to realize information transmission between server-side devices, detection devices, and user-side devices and other related devices.

[0155] The processor 601 is used to call the computer program in the memory 602. When the processor executes the computer program, it implements all the steps in the target object classification method in the above embodiments.

[0156] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the target object classification method in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the target object classification method in the above embodiments.

[0157] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0158] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0159] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0161] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for classifying target objects, characterized in that, include: Based on business needs, determine multiple metrics for each target object; For each indicator of each target object, the calculation formula of that indicator for that target object is parsed to obtain the score value of that indicator. Based on the score values ​​of multiple indicators, the score value of each target object is calculated. Based on the number of levels to be assigned, sort all target objects in ascending order of their score values; The sorted target objects are divided into several sets of categories to be classified, with each set corresponding to a category. Different categories can form multiple classification schemes, where each classification scheme includes multiple categories, and each category includes multiple target objects. The goal of dividing into multiple categories is to control the percentage range of target objects included in each category while considering the characteristics of the overall distribution of the scoring data. Calculate the size-weighted distance for each hierarchical scheme; The differential evolution algorithm is used to determine the optimal hierarchical scheme based on the size-weighted distance of multiple hierarchical schemes; The optimal grading scheme is visualized. The size-weighted distance for each grading scheme is calculated as follows: for each grading scheme, the percentage difference of each grade in the grading scheme is calculated, and the intra-class distance of each grade in the grading scheme is calculated; based on the percentage difference and intra-class distance of each grade in each grading scheme, the size-weighted distance of each grading scheme is calculated; the percentage difference refers to the difference between the percentage of samples included in each grade and the target percentage, and the intra-class distance refers to the overall difference in the rating values ​​of the target objects included in each grade; Calculating the intra-class distance for each level in the grading scheme includes: for each level of each grading scheme, constructing a cumulative array for that level based on the score array for that level, wherein the score array is an array formed by sorting the score values ​​of multiple target objects for that level in ascending order; determining the mean of the score array for that level within a preset range based on the score array and the cumulative data, wherein the preset range is between a first value and a second value; using a binary search method to find the insertion position of the mean in the score array; calculating the first intra-class distance for that level based on the first value and the insertion position; calculating the second intra-class distance for that level based on the insertion position and the second value; and calculating the intra-class distance for that level based on the first intra-class distance and the second intra-class distance.

2. The target object classification method as described in claim 1, characterized in that, The calculation formula for each indicator is written using predefined operators according to business requirements.

3. The target object classification method as described in claim 1, characterized in that, The calculation formula for this indicator of the target object is analyzed, including: The calculation formula for this index of the target object is parsed using the Reverse Polish Notation algorithm.

4. The target object classification method as described in claim 1, characterized in that, The percentage difference of each level in this grading scheme is calculated using the following formula: in, For level The difference in proportion; For level The percentage of target objects in the total number of target objects; Percentage of the target.

5. The target object classification method as described in claim 1, characterized in that, The intra-class distance for this level is calculated using the following formula, based on the first value and the insertion position: in, For level The first type of interior distance; The first value; Indicates the insertion position; For level The average of the score array within a preset range; To accumulate the value of the array at the insertion position; To accumulate the value at the first position in the array. For level The rating array; The intra-class distance for this level is calculated using the following formula, based on the insertion position and the second value: in, For level The second type of interior distance; The intra-class distance for this level is calculated using the following formula, based on the intra-class distances of the first and second classes: in, For level Intra-class distance.

6. The target object classification method as described in claim 1, characterized in that, The following formula is used to calculate the size-weighted distance of each grading scheme based on the difference in the proportion of each level and the intra-class distance: in Scale-weighted distance for each grading scheme; In each grading scheme, the level The difference in proportion; In each grading scheme, the level Intra-class distance; As a weighting factor, This represents the total number of levels.

7. A target object grading device, characterized in that, include: The metrics determination module is used to determine multiple metrics for each target object based on business needs. The rating calculation module is used to parse the calculation formula of each indicator for each target object, obtain the rating value of the indicator, and calculate the rating value of each target object based on the rating values ​​of multiple indicators. The grading scheme acquisition module is used to sort all target objects in ascending order of their score values ​​based on the number of grades to be graded; the sorted target objects are divided into several sets of grades to be graded, each set corresponding to one grade, and different segments can form multiple grading schemes. Each grading scheme includes multiple grades, and each grade includes multiple target objects. The goal of dividing into multiple grades is to control the percentage range of target objects contained in each grade while considering the characteristics of the overall distribution of the score data. The scale-weighted distance calculation module is used to calculate the scale-weighted distance for each hierarchical scheme; The optimal grading scheme determination module is used to determine the optimal grading scheme by employing the differential evolution algorithm based on the size-weighted distance of multiple grading schemes. The display module is used to visually demonstrate the optimal grading scheme; The size-weighted distance for each grading scheme is calculated as follows: for each grading scheme, the percentage difference of each grade in the grading scheme is calculated, and the intra-class distance of each grade in the grading scheme is calculated; based on the percentage difference and intra-class distance of each grade in each grading scheme, the size-weighted distance of each grading scheme is calculated; the percentage difference refers to the difference between the percentage of samples included in each grade and the target percentage, and the intra-class distance refers to the overall difference in the rating values ​​of the target objects included in each grade; Calculating the intra-class distance for each level in the grading scheme includes: for each level of each grading scheme, constructing a cumulative array for that level based on the score array for that level, wherein the score array is an array formed by sorting the score values ​​of multiple target objects for that level in ascending order; determining the mean of the score array for that level within a preset range based on the score array and the cumulative data, wherein the preset range is between a first value and a second value; using a binary search method to find the insertion position of the mean in the score array; calculating the first intra-class distance for that level based on the first value and the insertion position; calculating the second intra-class distance for that level based on the insertion position and the second value; and calculating the intra-class distance for that level based on the first intra-class distance and the second intra-class distance.

8. The target object grading device as described in claim 7, characterized in that, The calculation formula for each indicator is written using predefined operators according to business requirements.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that performs the method according to any one of claims 1 to 6.

Images