Object classification method and device, storage medium and electronic equipment

Abstract

The invention discloses an object classification method and device, a storage medium and electronic equipment. Relates to the technical field of data processing, the field of financial science and technology and other related technical fields, and comprises the following steps: calculating according to attribute information and historical behavior information of a plurality of to-be-classified objects to obtain corresponding cumulative function curves of the plurality of to-be-classified objects under the attribute information; the cumulative function curve is solved according to the multiple classification numbers, and an initial classification result corresponding to each classification number is obtained; according to the attribute information and the historical behavior information, evaluating the initial classification result to obtain a plurality of evaluation values; and determining a target classification number in the plurality of classification numbers according to the plurality of evaluation values, and determining a target classification result according to an initial classification result corresponding to the target classification number. Through the target object classification method and device, the problem that the accuracy of target object classification is low due to the fact that the target objects are classified in a manual mode in the related technology is solved.

Description

Technical Field

[0001] The present application relates to the fields of data processing technology, financial technology and other related technical fields, and specifically to a method and device for classifying objects, a storage medium and an electronic device. Background Art

[0002] In the financial, gaming, and new media fields, it is often necessary to bin (i.e., classify) customers based on the relationship between variables and targets. For example, in the credit field, when the target is whether a customer defaults, if the historical product usage behavior attributes (variables) and product default signs (targets) of all customers are known, IV maximization binning is performed for each historical product usage behavior attribute, so that each historical product usage behavior attribute of each customer can obtain the WOE value of the interval to which the corresponding historical product usage behavior attribute belongs based on the corresponding binning results. However, since the rules and policies in various fields are usually in a step-like or linear relationship, while hoping that the IV is significant enough, the staff will also propose constraints such as interval WOE trend, interval WOE difference, and interval sample difference for variable binning. Most traditional methods bin customers based on basic binning methods such as equal frequency binning and equal interval binning, lack consideration of additional constraints, are prone to misinterpretation risks, and the binning process is too time-consuming and inefficient. At the same time, since the relationship between distinguishing ability and the number of bins, and the relationship between endogeneity and the number of bins are inverse, there is a contradiction between the traditional IV index and WOE coding in terms of distinguishing ability and endogeneity, and the staff also have difficulties in choosing the appropriate number of bins.

[0003] With regard to the problem that the target objects are classified manually in the related technologies, resulting in a relatively low accuracy rate in the classification of the target objects, no effective solution has been proposed so far. Summary of the invention

[0004] The main purpose of the present application is to provide an object classification method and device, a storage medium and an electronic device to solve the problem in the related art that the target objects are classified manually, resulting in a relatively low accuracy rate in the classification of the target objects.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for classifying an object is provided. The method comprises: calculating based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in a target organization, obtaining the cumulative function curve corresponding to the multiple objects to be classified under the attribute information; solving the cumulative function curve respectively according to multiple classification quantities, obtaining the initial classification results corresponding to the multiple objects to be classified under each classification quantity; evaluating the initial classification results corresponding to the multiple objects to be classified under each classification quantity according to the attribute information and the historical behavior information, obtaining multiple evaluation values; determining the target classification quantity from the multiple classification quantities according to the multiple evaluation values, and determining the target classification result according to the initial classification result corresponding to the target classification quantity, wherein the target organization determines the processing strategy corresponding to the multiple objects to be classified according to the target classification result.

[0006] Furthermore, calculations are performed based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in the target organization to obtain a cumulative function curve corresponding to the multiple objects to be classified under the attribute information, including: determining a first value corresponding to the multiple objects to be classified based on the attribute information; calculating based on the historical behavior information of each object to be classified in the target organization to obtain a second value; and obtaining the cumulative function curve based on the first value and the second value.

[0007] Furthermore, the cumulative function curve is solved separately according to multiple classification numbers to obtain the initial classification results corresponding to the multiple objects to be classified under each classification number, including: determining the starting point in the cumulative function curve; for each classification number, solving the second numerical value corresponding to the starting point and the historical behavior information of the multiple objects to be classified in the cumulative function curve through the target cutting algorithm to obtain a first coordinate point, and determining the classification range according to the first coordinate point and the starting point; judging whether the number of categories after classifying the multiple objects to be classified under the classification range is equal to the number of categories; if the number of categories is not equal to the number of categories, repeatedly executing the step of solving the first coordinate point and the second numerical value corresponding to the historical behavior information of the multiple objects to be classified in the cumulative function curve through the target cutting algorithm to obtain the second coordinate point, until the initial classification result corresponding to the classification number is obtained.

[0008] Furthermore, for each number of classifications, the target cutting algorithm is used to solve the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the starting point and the cumulative function curve to obtain the first coordinate point, including: for each number of classifications, the target cutting algorithm is used to calculate the cumulative function curve based on the first sensitivity coefficient, the second sensitivity coefficient and the starting point to obtain a candidate point set, wherein the first sensitivity coefficient is used to characterize the degree of difference in the mean between the historical behavior information corresponding to each category after classification, and the second sensitivity coefficient is used to characterize the degree of difference in the number of objects corresponding to each category after classification; multiple candidate coordinate points in the candidate point set are calculated according to the IV index to obtain the index value corresponding to each candidate coordinate point; according to the index value corresponding to each candidate coordinate point, the first coordinate point is determined from the multiple candidate coordinate points in the candidate point set.

[0009] Furthermore, based on the attribute information and the historical behavior information, the initial classification results corresponding to the multiple objects to be classified under each classification quantity are evaluated to obtain multiple evaluation values, including: encoding the attribute information of the multiple objects to be classified to obtain the encoded attribute information; performing calculations based on the historical behavior information, the encoded attribute information and the attribute information to obtain a first evaluation value corresponding to the initial classification result; performing calculations based on the attribute information and the initial classification result to obtain a second evaluation value corresponding to the initial classification result; performing calculations based on the first evaluation value and the second evaluation value to obtain the multiple evaluation values.

[0010] Furthermore, calculations are performed based on the historical behavior information, the encoded attribute information, and the attribute information to obtain a first evaluation value corresponding to the initial classification result, including: calculations are performed based on the encoded attribute information and the attribute information to obtain a random error term corresponding to each object to be classified; calculations are performed based on the random error term, the encoded attribute information, and the historical behavior information to obtain a first calculation result; calculations are performed based on the encoded attribute information and the historical behavior information to obtain a second calculation result; calculations are performed based on the first calculation result and the second calculation result to obtain the first evaluation value.

[0011] Furthermore, based on the initial classification result corresponding to the target classification quantity, determining the target classification result includes: screening the multiple objects to be classified and the initial classification result according to preset screening conditions to obtain a first sample quantity corresponding to the screened multiple objects to be classified and a second sample quantity corresponding to the screened initial classification result; calculating based on the first sample quantity and the second sample quantity to obtain a credibility index value corresponding to the initial classification result; adjusting the initial classification result based on the credibility index value to obtain an adjusted initial classification result; and determining the adjusted initial classification result as the target classification result.

[0012] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a device for classifying an object is provided. The device comprises: a calculation unit, which is used to calculate according to the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in a target organization, and obtain the cumulative function curve corresponding to the multiple objects to be classified under the attribute information; a solution unit, which is used to solve the cumulative function curve respectively according to multiple classification quantities, and obtain the initial classification results corresponding to the multiple objects to be classified under each classification quantity; an evaluation unit, which is used to evaluate the initial classification results corresponding to the multiple objects to be classified under each classification quantity according to the attribute information and the historical behavior information, and obtain multiple evaluation values; a determination unit, which is used to determine the target classification quantity among the multiple classification quantities according to the multiple evaluation values, and determine the target classification result according to the initial classification result corresponding to the target classification quantity, wherein the target organization determines the processing strategy corresponding to the multiple objects to be classified according to the target classification result.

[0013] Furthermore, the calculation unit includes: a first determination module, used to determine the first numerical value corresponding to the multiple objects to be classified based on the attribute information; a first calculation module, used to calculate according to the historical behavior information of each object to be classified in the target organization to obtain a second numerical value; a second determination module, used to obtain the cumulative function curve based on the first numerical value and the second numerical value.

[0014] Furthermore, the solving unit includes: a third determination module, used to determine the starting point in the cumulative function curve; a solving module, used to solve the second numerical value corresponding to the starting point and the historical behavior information of multiple objects to be classified in the cumulative function curve for each classification quantity through a target cutting algorithm to obtain a first coordinate point, and determine the classification range based on the first coordinate point and the starting point; a judgment module, used to judge whether the number of classifications after classifying the multiple objects to be classified within the classification range is equal to the number of classifications; an execution module, used to repeat the step of solving the first coordinate point and the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the cumulative function curve through the target cutting algorithm to obtain a second coordinate point if the number of classifications is not equal to the number of classifications, until an initial classification result corresponding to the number of classifications is obtained.

[0015] Furthermore, the solution module includes: a first calculation submodule, used for calculating the cumulative function curve for each classification quantity through the target cutting algorithm based on the first sensitivity coefficient, the second sensitivity coefficient and the starting point to obtain a candidate point set, wherein the first sensitivity coefficient is used to characterize the degree of difference in the mean between the historical behavior information corresponding to each category after classification, and the second sensitivity coefficient is used to characterize the degree of difference in the number of objects corresponding to each category after classification; a second calculation submodule, used for calculating multiple candidate coordinate points in the candidate point set according to the IV index to obtain the index value corresponding to each candidate coordinate point; a determination submodule, used to determine the first coordinate point from the multiple candidate coordinate points in the candidate point set according to the index value corresponding to each candidate coordinate point.

[0016] Furthermore, the evaluation unit includes: an encoding module, used to encode the attribute information of the multiple objects to be classified to obtain the encoded attribute information; a second calculation module, used to perform calculations based on the historical behavior information, the encoded attribute information and the attribute information to obtain a first evaluation value corresponding to the initial classification result; a third calculation module, used to perform calculations based on the attribute information and the initial classification result to obtain a second evaluation value corresponding to the initial classification result; and a fourth calculation module, used to perform calculations based on the first evaluation value and the second evaluation value to obtain the multiple evaluation values.

[0017] Furthermore, the second calculation module includes: a third calculation submodule, used to perform calculations based on the encoded attribute information and the attribute information to obtain a random error term corresponding to each object to be classified; a fourth calculation submodule, used to perform calculations based on the random error term, the encoded attribute information and the historical behavior information to obtain a first calculation result; a fifth calculation submodule, used to perform calculations based on the encoded attribute information and the historical behavior information to obtain a second calculation result; and a sixth calculation submodule, used to perform calculations based on the first calculation result and the second calculation result to obtain the first evaluation value.

[0018] Furthermore, the determination unit includes: a screening module, which is used to screen the multiple objects to be classified and the initial classification results according to preset screening conditions, and obtain a first sample number corresponding to the multiple objects to be classified after screening and a second sample number corresponding to the initial classification results after screening; a fifth calculation module, which is used to calculate based on the first sample number and the second sample number to obtain a credibility index value corresponding to the initial classification result; an adjustment module, which is used to adjust the initial classification result according to the credibility index value to obtain an adjusted initial classification result; and a fourth determination module, which is used to determine the adjusted initial classification result as the target classification result.

[0019] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a computer-readable storage medium is provided, wherein the storage medium stores a program, wherein when the program is running, the device where the storage medium is located is controlled to execute any one of the object classification methods described above.

[0020] In order to achieve the above-mentioned purpose, according to another aspect of the present application, an electronic device is further provided, the electronic device includes one or more processors and a memory, and the memory is used to store the classification method of the object described in any one of the above-mentioned methods implemented by the one or more processors.

[0021] Through the present application, the following steps are adopted: according to the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in the target organization, calculations are performed to obtain the cumulative function curves corresponding to the multiple objects to be classified under the attribute information; according to multiple classification quantities, the cumulative function curves are respectively solved to obtain the initial classification results corresponding to the multiple objects to be classified under each classification quantity; according to the attribute information and the historical behavior information, the initial classification results corresponding to the multiple objects to be classified under each classification quantity are evaluated to obtain multiple evaluation values; according to the multiple evaluation values, the target classification quantity is determined among the multiple classification quantities, and according to the initial classification results corresponding to the target classification quantity, the target classification result is determined, wherein the target organization determines the processing strategies corresponding to the multiple objects to be classified according to the target classification result. Through the present application, the problem of manually classifying the target objects in the related art, resulting in a relatively low accuracy rate in classifying the target objects, is solved.

[0022] In this scheme, based on the attribute information corresponding to multiple objects to be classified and their historical behavior information in the target organization, a cumulative function curve corresponding to each object to be classified under the attribute information is constructed. Then, through multiple preset classification numbers, the cumulative function curve is solved respectively to obtain the initial classification results corresponding to multiple objects to be classified under each classification number. Finally, the initial classification results under each classification number are evaluated to determine the appropriate classification number (i.e., the target classification number) and classification results (i.e., the target classification results).

[0023] Through the attribute information and historical behavior information of multiple objects to be classified, the corresponding cumulative function curves of multiple objects to be classified under the attribute information can be quickly constructed to better capture the characteristics and change trends of the objects to be classified. At the same time, by solving the cumulative function curves based on multiple classification quantities, the initial classification results of multiple objects to be classified under each classification quantity can be obtained more intuitively, which is helpful to further analyze and understand the characteristics and behaviors of the objects to be classified, improve the accuracy and reliability of classification, and help decision makers make correct classification decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0025] Figure 1 is a flow chart of a method for classifying objects provided in an embodiment of the present application;

[0026] Figure 2 is a schematic diagram of a cumulative function curve provided according to an embodiment of the present application;

[0027] Figure 3 is a schematic diagram of a candidate point set provided according to an embodiment of the present application;

[0028] Figure 4 is a schematic diagram of a device for classifying objects provided according to an embodiment of the present application;

[0029] Figure 5 is a schematic diagram of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0031] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0034] WOE (Weight of Evidence): It represents the difference between positive and negative samples of feature variable groups. Generally speaking, the larger the absolute value of WOE, the better the prediction ability of the group; the smaller the absolute value of WOE, the worse the prediction ability of the group.

[0035] IV (Information Value): Information value is used to indicate the contribution of a feature to the target prediction, that is, the predictive ability of the feature. The larger the IV value, the stronger the predictive ability of the feature.

[0036] Sum of Squares Between Treatments or Factors (SSA): reflects the sum of the differences between the average level of each characteristic group and the overall average level.

[0037] Constrained optimization binning: For each variable and the corresponding target, modelers usually need to divide the sample variable values ​​into several intervals for variable discretization. In addition to maximizing indicators such as IV, staff usually hope that after a variable is discretized, it can also meet other constraints. The industry generally calls this binning method that maximizes indicators while taking into account other constraints constrained optimization binning.

[0038] WOE coding: Before modeling, modelers need to perform WOE coding, that is, according to the binning results, convert the independent variable of any sample into the WOE value of the interval to which the independent variable belongs, and use the WOE value instead of the original independent variable for modeling.

[0039] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. For example, an interface is set up between this system and the relevant users or organizations. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain relevant information after receiving the consent information fed back by the aforementioned user or organization.

[0040] The present invention is described below in conjunction with preferred implementation steps. Figure 1 is a flow chart of a method for classifying objects according to an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0041] Step S101 , performing calculations based on the attribute information of a plurality of objects to be classified and the historical behavior information of each object to be classified in a target organization, to obtain a cumulative function curve corresponding to the plurality of objects to be classified under the attribute information.

[0042] Optionally, attribute information (which can also be understood as an independent variable in the classification process) includes but is not limited to: age information, occupation information and asset information of the user (i.e., each object to be classified). Historical behavior information (which can also be understood as a dependent variable in the classification process) includes but is not limited to: whether the user has potential risks in the target institution, or whether the user can bring benefits to the target institution, and other binary classification information. Target institutions include but are not limited to: financial institutions, Internet institutions, media institutions and game institutions. Modeling is performed based on attribute information and historical behavior information to obtain cumulative function curves corresponding to multiple objects to be classified under attribute information.

[0043] It should be noted that the independent variables in the classification process are numerical independent variables, which include but are not limited to: age information and asset information of multiple objects to be classified.

[0044] It should be noted that the attribute information can be screened according to the corresponding thresholds corresponding to the missing rate, correlation, etc. (it can also be understood as independent variable screening) to obtain the screened attribute information (it can also be understood as the input independent variable in the modeling process).

[0045] Step S102, solving the cumulative function curves respectively according to multiple classification numbers to obtain initial classification results corresponding to multiple objects to be classified under each classification number.

[0046] Optionally, a potential number of bins (i.e., multiple numbers of classifications) is looped through, and for each number of bins (i.e., the number of classifications used in each loop), a cumulative function curve corresponding to multiple objects to be classified under the number of bins is solved by a geometric cutting algorithm to obtain initial binning results (i.e., initial classification results) corresponding to the multiple objects to be classified under the number of bins.

[0047] Step S103 , evaluating the initial classification results corresponding to the multiple objects to be classified under each classification quantity according to the attribute information and the historical behavior information, and obtaining multiple evaluation values.

[0048] Optionally, the performance of the initial classification results corresponding to the multiple objects to be classified under each number of classifications is calculated through comprehensive performance indicators (such as endogenous indicators and distinguishing ability indicators, etc.) to obtain multiple evaluation values.

[0049] Step S104, determining a target classification number from a plurality of classification numbers based on a plurality of evaluation values, and determining a target classification result based on an initial classification result corresponding to the target classification number, wherein the target organization determines a processing strategy corresponding to a plurality of objects to be classified based on the target classification result.

[0050] Optionally, the optimal binning number and the corresponding initial binning result are found among the multiple evaluation values ​​as the target classification number and target classification result corresponding to the multiple objects to be classified.

[0051] It should be noted that in the actual binning process of multiple objects to be classified, in addition to requiring the IV of the independent variable to be as high as possible after binning, the staff also requires the practical meaning and practicality of the variable to be taken into account. For example: the staff will have the following binning constraints: 1) The WOE values ​​of the previous and next intervals show a monotonically increasing or monotonically decreasing trend; 2) The difference in the WOE values ​​of the previous and next intervals must be obvious enough; 3) The difference in the number of samples in each interval should not be too great. This binning method that takes into account other constraints while maximizing the target indicator is called constrained optimization binning in the industry.

[0052] In summary, based on the attribute information corresponding to multiple objects to be classified and their historical behavior information in the target organization, the cumulative function curve corresponding to each object to be classified under the attribute information is constructed. Then, through the preset multiple classification numbers, the cumulative function curve is solved respectively to obtain the initial classification results corresponding to multiple objects to be classified under each classification number. Finally, the initial classification results under each classification number are evaluated to determine the appropriate classification number (i.e., the target classification number) and classification results (i.e., the target classification results).

[0053] Through the attribute information and historical behavior information of multiple objects to be classified, the corresponding cumulative function curves of multiple objects to be classified under the attribute information can be quickly constructed to better capture the characteristics and change trends of the objects to be classified. At the same time, by solving the cumulative function curves based on multiple classification quantities, the initial classification results of multiple objects to be classified under each classification quantity can be obtained more intuitively, which is helpful to further analyze and understand the characteristics and behaviors of the objects to be classified, improve the accuracy and reliability of classification, and help decision makers make correct classification decisions.

[0054] Optionally, in the object classification method provided in the embodiment of the present application, calculations are performed based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in the target organization to obtain a cumulative function curve corresponding to the multiple objects to be classified under the attribute information, including: determining a first value corresponding to the multiple objects to be classified based on the attribute information; calculating based on the historical behavior information of each object to be classified in the target organization to obtain a second value; and obtaining a cumulative function curve based on the first value and the second value.

[0055] In an optional embodiment, assuming that the independent variable (i.e., attribute information) is x, the dependent variable (i.e., historical behavior information) is y, and there are n samples (i.e., n customers) when modeling, the independent variable x of the tth sample is marked as , the dependent variable y is labeled Sort the n customers according to the value of the independent variable x from small to large, and number the customers in sequence. The customer numbers are , the value of the independent variable x of the pth customer is marked as (ie the first value).

[0056] Constructing cumulative dependent variables , represents the accumulation of the dependent variable from the first customer to the pth customer (i.e., the second value), that is, It can be expressed as:

[0057]

[0058] because ,so , that is, the cumulative dependent variable is a function of the cumulative number of customers.

[0059] The customer number p from small to large according to the independent variable x is used as the horizontal axis, and the cumulative dependent variable As the vertical axis, draw the curve of the cumulative dependent variable (that is, the cumulative function curve). When the horizontal axis is equal to 0, it is marked as , since there is no cumulative dependent variable at this time, , so the origin (0,0) can be expressed as .

[0060] It should be noted that Figure 2 Schematic diagram of a cumulative function curve provided according to an embodiment of the present application. Figure 2 As shown in the figure, assuming that the staff wants to obtain k classification results, k horizontal coordinates are taken on the horizontal axis and numbered respectively. , , …, Among them, the last horizontal coordinate (i.e., the kth horizontal coordinate) is fixed to n, that is, The above k points correspond to k numbers, that is, to k customers, and their independent variable x values ​​are , , …, The corresponding ordinates of the k horizontal coordinates on the curve of the cumulative dependent variable are , ,……, . This constitutes k points on the cumulative dependent variable curve The first point on the curve With origin After that, each point is connected to the previous point, that is, and connected to form a polyline consisting of k line segments. The horizontal coordinates of the two endpoints of each line segment form the upper and lower bounds of an interval. For example: the i-th line segment is point With point So it corresponds to the i-th interval, which is .therefore, The corresponding interval before was , and The corresponding interval is , and The corresponding interval is , and The corresponding interval is Dividing k segments on the cumulative dependent variable curve is equivalent to dividing the independent variable x into k boxes.

[0061] It should be noted that for the i-th interval For , the WOE value can be expressed as:

[0062]

[0063] in, represents the number of samples in which the dependent variable is equal to 1 in the i-th interval, represents the number of samples where all dependent variables are equal to 1, represents the number of samples in which the dependent variable is equal to 0 in the i-th interval, It represents the number of samples where all dependent variables are equal to 0.

[0064] Requiring that the mean returns of each interval after the independent variable is binned be monotonic actually requires that the slopes of the k line segments on the "cumulative dependent variable" curve increase or decrease one by one.

[0065] It should be noted that the constraints can be visualized as geometric properties: (1) The WOE values ​​of the previous and next intervals on the cumulative function curve show a monotonically increasing or monotonically decreasing trend, which is equivalent to requiring that the slopes of the k line segments on the cumulative function curve increase or decrease one by one. (2) The difference in the WOE values ​​of the previous and next intervals on the cumulative function curve should be obvious enough, which is equivalent to requiring that the difference in the slopes of the k line segments on the cumulative function curve should be obvious enough. (3) The difference in the number of samples in each interval on the cumulative function curve should not be too great, which is equivalent to requiring that the difference in the projection length of the slopes of the k line segments on the cumulative function curve on the horizontal axis should not be too great.

[0066] By comprehensively considering the attribute information and historical behavior information of multiple objects to be classified to construct a cumulative function curve, it is possible to more comprehensively reflect the characteristics and behaviors of multiple objects to be classified, thereby improving the accuracy of classification.

[0067] Optionally, in the object classification method provided in the embodiment of the present application, the cumulative function curve is solved separately according to multiple classification quantities to obtain the initial classification results corresponding to multiple objects to be classified under each classification quantity, including: determining the starting point in the cumulative function curve; for each classification quantity, solving the second numerical value corresponding to the starting point and the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm to obtain a first coordinate point, and determining the classification range based on the first coordinate point and the starting point; judging whether the number of categories after classifying multiple objects to be classified under the classification range is equal to the number of categories; if the number of categories is not equal to the number of categories, repeatedly executing the step of solving the first coordinate point and the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm to obtain a second coordinate point, until the initial classification result corresponding to the classification quantity is obtained.

[0068] In an optional embodiment, the geometric cutting algorithm (i.e., the target cutting algorithm) includes three parameters, namely: the interval number parameter k, the sensitivity coefficient δ of the difference between the mean values ​​of the previous and next intervals, and the sensitivity coefficient θ of the difference between the number of samples of the previous and next intervals. Each parameter corresponds to a different modeling requirement, that is, it represents a different meaning. Interval number parameter k: The staff hopes that an independent variable can be divided into k boxes. The sensitivity coefficient δ of the difference between the mean values ​​of the previous and next intervals: The staff usually hopes that after the binning, the difference in the mean returns of each interval should be obvious enough, that is, the difference in the mean returns of each interval should be greater than , which is equivalent to the k line segments of the cumulative function curve after geometric cutting. The difference in slope between the front and rear segments should not be too small, that is, the difference in slope between the front and rear segments should be greater than δ. According to industry practice, the default value of δ is 0.05. The sensitivity coefficient of the difference in the number of samples between the front and rear intervals θ: The staff usually hopes that after binning, the difference in the number of samples in each interval will not be too great, that is, the number of samples in the interval with the largest number of samples ÷ the number of samples in the interval with the smallest number of samples ≤ θ. According to industry practice, the default value of θ is 2.

[0069] For each number of classifications (for example, assuming that k is initially set to 6, then the number of classifications is 6, 5, 4, 3, and 2 respectively), the geometric cutting algorithm is used to solve the origin of the cumulative function (i.e., the starting point) and the second numerical value corresponding to the historical behavior information of multiple objects to be classified based on the above parameters to obtain the first coordinate point, and the range between the first coordinate point and the origin is used as the first classification range, and the initial classification results corresponding to the multiple objects to be classified under the attribute information are obtained according to the first classification range.

[0070] It should be noted that if (i-1) groups have been successfully obtained, that is, the first (i-1) segments have been successfully separated on the cumulative dependent variable curve, but the number of categories obtained is not equal to the current number of categories (that is, the current k value), now we want to get the i-th segmentation point, so as to separate the i-th segment and then get the i-th group. First, draw the lines that pass through the points at the same time. (i.e. the first coordinate point) and (i.e. the second coordinate point) . And draw the passing points And the slope is (δ is the sensitivity coefficient of the difference between the mean values ​​of the previous and next intervals, The straight line representing the slope of the i-1th line segment on the cumulative function curve , on the cumulative dependent variable curve, located on the straight line Below and on the line Above and with the horizontal axis between (θ is the sensitivity coefficient of the difference in the number of samples between the previous and next intervals, k is the number of each category, and n is the number of samples) and (θ is the sensitivity coefficient of the difference in the number of samples in the previous and next intervals, k is the number of categories, and n is the number of samples) The points between are the potential feasible point sets. The points in the feasible point set are prioritized and the point with the highest priority is selected as the i-th segmentation point. Its coordinates are .connect and , forming a line segment, the slope of the line segment can be obtained as If the feasible point set is empty, it means that this recursion is invalid and you need to return to the previous step. In the feasible point set formed in the previous step, select The point with a lower priority level is used as the i-1th segmentation point, and then the above operation is repeated again. The geometric cutting algorithm will eventually end after completing the k-box binning, that is, it will end when the k-1th segmentation point is found, and all k segmentation points and k intervals (i.e., the initial classification results) will be output. Since the geometric cutting algorithm will return to the previous step when it cannot continue to segment, in extreme cases, the algorithm will continue to return until it returns to the feasible point set used to determine the first segmentation point. If the feasible point set used for the first segmentation point has been explored and no suitable i-th segmentation point can be found, it means that the classification has failed and the algorithm will output an empty set.

[0071] It should be noted that we first assume that the final classification trend of multiple objects to be classified is monotonically increasing. If the output result is an empty set when it is assumed that the final classification trend of multiple objects to be classified is monotonically increasing, then we assume that the final classification trend of multiple objects to be classified is monotonically decreasing, and solve the initial classification result corresponding to the cumulative function under the current number of classifications according to the above method.

[0072] It should be noted that the solution objectives of the geometric cutting algorithm are: (1) to achieve a monotonic trend after binning, that is, after binning, the woe values ​​of each interval are monotonically increasing or monotonically decreasing. (2) to ensure that after binning, the differences in the woe values ​​of each interval are significant enough, that is, the woe value of the high-risk customer group is as high as possible, and the woe value of the low-risk customer group is as low as possible, and at the same time, the difference in the number of samples contained in each interval should not be too great.

[0073] By repeatedly executing the geometric cutting algorithm, it is possible to ensure that the initial classification results corresponding to the current number of classifications are obtained, thereby improving the accuracy and reliability of the classification and ensuring that each object to be classified is correctly classified into the corresponding group.

[0074] Optionally, in the object classification method provided in the embodiment of the present application, for each number of classifications, the target cutting algorithm is used to solve the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the starting point and the cumulative function curve to obtain the first coordinate point, including: for each number of classifications, the target cutting algorithm is used to calculate the cumulative function curve based on the first sensitivity coefficient, the second sensitivity coefficient and the starting point to obtain a candidate point set, wherein the first sensitivity coefficient is used to characterize the degree of difference in the mean between the historical behavior information corresponding to each category after classification, and the second sensitivity coefficient is used to characterize the degree of difference in the number of objects corresponding to each category after classification; multiple candidate coordinate points in the candidate point set are calculated according to the IV indicator to obtain the index value corresponding to each candidate coordinate point; based on the index value corresponding to each candidate coordinate point, the first coordinate point is determined from the multiple candidate coordinate points in the candidate point set.

[0075] In an optional embodiment, for each number of categories, if (i-1) groups have been successfully obtained, that is, the first (i-1) line segments have been successfully separated on the cumulative dependent variable curve, but the number of categories obtained is not equal to the current number of categories (that is, the current k value), it is now desired to obtain the i-th segmentation point, thereby separating the i-th line segment and then obtaining the i-th group. First, draw the points that pass through the points at the same time. (i.e. the first coordinate point) and (i.e. the second coordinate point) . And draw the passing points And the slope is (δ is the sensitivity coefficient of the difference between the mean values ​​of the previous and next intervals, Represents the slope of the i-1th line segment on the cumulative function curve The straight line , on the cumulative dependent variable curve, located on the straight line Below and on the line Above and with the horizontal axis between (θ is the sensitivity coefficient of the difference in the number of samples between the previous and next intervals, k is the number of each category, and n is the number of samples) and The points between (θ is the sensitivity coefficient of the difference in the number of samples between the previous and next intervals, k is the number of categories, and n is the number of samples) are the potential feasible point sets. Among them, the sensitivity coefficient of the difference in the mean value between the previous and next intervals δ is the first sensitivity coefficient, and the sensitivity coefficient of the difference in the number of samples between the previous and next intervals θ is the second sensitivity coefficient.

[0076] It should be noted that Figure 3 is a schematic diagram of a candidate point set provided according to an embodiment of the present application. Figure 3 As shown in , if the number of points in the feasible point set is greater than 1, then the potentially feasible points (i.e., candidate coordinate points) need to be prioritized in the feasible point set. Suppose a point in the feasible point set is , and this point is the i-th split point, which can constitute the potential i-th interval. Calculate the IV value of the mean return of the first k intervals corresponding to this point, expressed as , and follow To sort, The higher the point, the higher the priority.

[0077] The geometric cutting algorithm is used to comprehensively evaluate the historical behavior information of multiple objects to be classified, so as to find the optimal first coordinate point, thereby improving the accuracy and efficiency of classification.

[0078] Optionally, in the object classification method provided in the embodiment of the present application, initial classification results corresponding to multiple objects to be classified under each classification quantity are evaluated based on attribute information and historical behavior information, and multiple evaluation values ​​are obtained, including: encoding the attribute information of multiple objects to be classified to obtain encoded attribute information; calculating based on historical behavior information, encoded attribute information and attribute information to obtain a first evaluation value corresponding to the initial classification result; calculating based on the attribute information and the initial classification result to obtain a second evaluation value corresponding to the initial classification result; calculating based on the first evaluation value and the second evaluation value to obtain multiple evaluation values.

[0079] In an optional embodiment, there is a contradiction between the traditional IV index and the WOE code. There is a causal relationship between the traditional WOE code and the dependent variable, that is, the endogeneity problem of statistics. When the number of bins increases, the IV value of the variable is higher, but at the same time, the potential endogeneity problem of the WOE code will become more obvious; when the number of bins decreases, the endogeneity problem of the WOE code will be alleviated, but the ability to distinguish the variable will be weakened, that is, the IV value will decline. Therefore, the essence of the contradiction between the IV index and the WOE code is that the relationship between the ability to distinguish and the number of bins, and the relationship between endogeneity and the number of bins are reversed. In order to resolve the relevant contradictions, it is necessary to evaluate the initial classification results to obtain multiple evaluations, and then find the optimal number of bins (that is, the number of target classifications) based on multiple evaluation values.

[0080] If there are m initial classification results corresponding to each classification number of multiple objects to be classified (i.e., m bins are completed for the independent variable), then for any sample among the n samples (i.e., n objects to be classified), if its independent variable (i.e. attribute information) belongs to the i-th interval , then the encoded variable (i.e. the encoded attribute information) will be assigned , the dependent variable is (i.e. historical behavior information), It can be expressed as follows:

[0081]

[0082] According to the coded variables , independent variable and the dependent variable is By calculating, we can get the endogenous index value of the initial classification results corresponding to each classification number of multiple objects to be classified (i.e. the first evaluation value). The larger the value of this indicator is, the less endogeneity there is in terms of causality, that is, the better the endogeneity problem is alleviated.

[0083] The m initial classification results are marked as , construct the difference index corresponding to the initial classification result , The expression is as follows:

[0084]

[0085] in, Represents the difference index corresponding to the initial classification result, represents the number of samples where all dependent variables are equal to 1, It indicates the number of samples where all dependent variables are equal to 0. SSA is the sum of squares of deviations between groups in statistics. The meanings of the various components of this indicator are as follows: is the scaling factor, SSA reflects the sum of the differences between the average level of each interval and the average level of the entire sample, and n-1 is the penalty term.

[0086] After simplifying the difference index, we can know It obeys the chi-square distribution, that is:

[0087]

[0088] Therefore, the difference index values ​​of the initial classification results corresponding to multiple objects to be classified under each classification number can be obtained. (i.e. the second evaluation value), which is expressed as follows:

[0089]

[0090] The larger the value of this indicator is, the more obvious the difference between the final samples is, that is, the better the discrimination ability is.

[0091] By multiplying the endogeneity index value and the difference index value, we can get the comprehensive index value (i.e., multiple evaluation values) of the initial classification results corresponding to multiple objects to be classified under each classification number. The larger the index value, the more the number of bins can balance the pros and cons of alleviating endogeneity and improving the ability to distinguish.

[0092] Among them, the comprehensive index value The expression is as follows:

[0093]

[0094] By multiplying the endogenous index value and the difference index value, the evaluation value of the initial classification result corresponding to each classification number of multiple objects to be classified can be obtained quickly and accurately.

[0095] Optionally, in the object classification method provided in the embodiment of the present application, calculation is performed based on historical behavior information, encoded attribute information and attribute information to obtain a first evaluation value corresponding to the initial classification result, including: calculation is performed based on the encoded attribute information and attribute information to obtain a random error term corresponding to each object to be classified; calculation is performed based on the random error term, the encoded attribute information and the historical behavior information to obtain a first calculation result; calculation is performed based on the encoded attribute information and the historical behavior information to obtain a second calculation result; calculation is performed based on the first calculation result and the second calculation result to obtain a first evaluation value.

[0096] In an optional embodiment, construct the encoded variable and the original independent variable The one-variable linear equation is expressed as follows:

[0097]

[0098] in, is the disturbance term (i.e., random error term), η and λ are fixed parameters.

[0099] After statistical operations, we can get the estimated values ​​of parameters η and λ. Let the estimated values ​​be and . Then, according to the univariate linear equation, the estimated value of the disturbance term of each sample can be obtained, and the expression is as follows:

[0100]

[0101] Construct the encoded variables separately With or without disturbance terms cases, and the dependent variable A linear equation in one variable.

[0102] Encoded variables In the presence of a disturbance cases, and the dependent variable The expression of the one-variable linear equation is as follows:

[0103]

[0104] Among them, α, β, ρ and ζ are fixed parameters, represents the encoded variable, represents the disturbance term.

[0105] It can be expressed as:

[0106]

[0107] in, Dependent variable The probability is equal to 1, Dependent variable The probability of being equal to 0, α, β and ρ are fixed parameters, represents the encoded variable, represents the disturbance term.

[0108] So we can construct the log-likelihood function , The expression is as follows:

[0109]

[0110] The log-likelihood function can be obtained The maximum value of (i.e. the first calculation result).

[0111] Encoded variables In the absence of disturbance terms cases, and the dependent variable The expression of the one-variable linear equation is as follows:

[0112]

[0113] Among them, α, β and ζ are fixed parameters, Represents the encoded variable.

[0114] It can be expressed as:

[0115]

[0116] in, Dependent variable The probability is equal to 1, Dependent variable The probability of being equal to 0, α and β are fixed parameters, Represents the encoded variable.

[0117] So we can construct the log-likelihood function , The expression is as follows:

[0118]

[0119] The log-likelihood function can be obtained The maximum value of (i.e. the second calculation result).

[0120] According to Taylor expansion and the law of large numbers, when the amount of data is large enough, the difference between the maximum values ​​of the above two log-likelihood functions obeys the chi-square distribution, and the expression is as follows:

[0121]

[0122] Therefore, by calculating based on the difference between the maximum values ​​of the two log-likelihood functions mentioned above, the endogenous index value (i.e., the first index value) corresponding to the initial classification result can be obtained.

[0123] Among them, the endogenous index value The expression is as follows:

[0124]

[0125] By calculating through the historical behavior information, the encoded attribute information and the attribute information, the first evaluation value corresponding to the initial classification result can be obtained quickly and accurately.

[0126] Optionally, in the object classification method provided in the embodiment of the present application, determining the target classification result based on the initial classification result corresponding to the target classification quantity includes: screening multiple objects to be classified and the initial classification result according to preset screening conditions to obtain a first sample quantity corresponding to the screened multiple objects to be classified and a second sample quantity corresponding to the screened initial classification result; calculating based on the first sample quantity and the second sample quantity to obtain a credibility index value corresponding to the initial classification result; adjusting the initial classification result based on the credibility index value to obtain an adjusted initial classification result; and determining the adjusted initial classification result as the target classification result.

[0127] In an optional embodiment, the largest evaluation value is selected from multiple evaluation values, and the number of categories corresponding to the evaluation value is determined as the optimal number of bins (i.e., the target number of categories). Woe coding correction is performed on the initial classification result corresponding to the target number of categories, and the Woe coding correction process is as follows:

[0128] First, according to (ie, the number of target categories) woe values ​​of the interval ,calculate The average value of woe in the interval , The expression is as follows:

[0129]

[0130] For n samples, obtain the maximum value max(x) and the minimum value min(x) of the dependent variable in the sample, and calculate the ideal sampling interval interval(x) for the sample. The expression is as follows:

[0131]

[0132] For each sample, calculate the minimum difference between the independent variable of other samples and this sample. , which is expressed as follows:

[0133]

[0134] Calculate n samples, The number of samples less than or equal to interval(x) (i.e. the first sample size), which is expressed as follows:

[0135]

[0136] It should be noted that the samples with a moderate distance from other samples are referred to as uniform points. Indicates that among n samples, A uniform point, so Reflects the density of the space point group. Assuming that the overall sample is an unbiased estimate of the true parameter, which means that in the current population space, The space point group of is denser, The density of the spatial point group means that if true random sampling is achieved, the density of the obtained sampling point group should be similar to the density of the spatial point group, otherwise the sampling is considered biased.

[0137] For each interval For each sample in the interval, calculate the minimum value of the difference between other samples in the interval and the independent variable of this sample , which is expressed as follows:

[0138]

[0139] For each interval , calculate the interval The number of samples less than or equal to interval(x) (i.e. the second sample size), its expression is as follows:

[0140]

[0141] It should be noted that reflects the density in this interval, so It reflects the overall density situation inferred based on the density situation in this interval.

[0142] According to the geometric meaning of Poisson distribution, for each interval , calculate the credibility index value , which is expressed as follows:

[0143]

[0144] in, Express Perform rounding operation, Indicates from 1 to The continuous multiplication of all natural numbers, Similarly, Indicates that the sample The number of samples less than or equal to interval(x), Indicates that the interval The number of samples less than or equal to interval(x).

[0145] in, The simplified expression is as follows:

[0146]

[0147] in, Express Perform rounding operation, Indicates from 1 to The continuous multiplication of all natural numbers, Similarly, Indicates from to The continuous multiplication of all natural numbers, Similarly, Indicates that the sample The number of samples less than or equal to interval(x), Indicates that the interval The number of samples less than or equal to interval(x).

[0148] It should be noted that the geometric meaning of the Poisson distribution shows that if n samples are randomly selected from the overall space, the expected value of the number of uniform points is , and then, the number of samples from any n sample uniform points should obey the Poisson distribution. At the same time, It reflects the density of the space point group. If true random sampling is achieved, the density of the sampling point group should be similar to the density of the space point group, otherwise the sampling is considered biased.

[0149] If each box of samples in the binning result is regarded as a resample, focusing on a certain interval, the density inferred based on the interval is , so we can get the probability value , which is the probability value divided by the peak value of the Poisson distribution , the final result Represents the similarity between the density of this interval and the overall spatial density. The larger the value is, the more reliable the sampling method of this interval is, which complies with the point group distribution law of the overall spatial density.

[0150] therefore, The interval can be calculated indivual , output each interval Corrected woe encoding value , to obtain the target classification result, The expression is as follows:

[0151]

[0152] It should be noted that traditional classification results rely on WOE coding, but traditional WOE coding may lead to extreme values ​​due to sampling bias. For example: if the density of the overall space is uniform, a certain interval is (0,100], there are 100 samples falling into this interval, but 90 samples are between (0,10], and the remaining samples are scattered between (10,100], then the WOE value of the entire interval is not very reliable, because there are a lot of gaps between (10,100], which may be caused by improper sampling, resulting in the sample information between (10,100] not being fully included in the coding value. The WOE value of this interval is very prone to extreme values. Therefore, it is necessary to calibrate and adjust the WOE coding.

[0153] Through the credibility index value, the initial classification result can be adjusted quickly and accurately, and then the target classification result can be obtained efficiently.

[0154] The object classification method provided in the embodiment of the present application performs calculations based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in a target organization to obtain a cumulative function curve corresponding to the multiple objects to be classified under the attribute information; solves the cumulative function curves respectively according to multiple classification quantities to obtain initial classification results corresponding to the multiple objects to be classified under each classification quantity; evaluates the initial classification results corresponding to the multiple objects to be classified under each classification quantity based on the attribute information and the historical behavior information to obtain multiple evaluation values; determines a target classification quantity from the multiple classification quantities based on the multiple evaluation values, and determines a target classification result based on the initial classification result corresponding to the target classification quantity, wherein the target organization determines the processing strategies corresponding to the multiple objects to be classified based on the target classification result, thereby solving the problem in the related art that the target objects are classified manually, resulting in a relatively low accuracy rate in classifying the target objects.

[0155] In summary, in this scheme, based on the attribute information corresponding to multiple objects to be classified and their historical behavior information in the target organization, a cumulative function curve corresponding to each object to be classified under the attribute information is constructed. Then, through multiple preset classification numbers, the cumulative function curve is solved respectively to obtain the initial classification results corresponding to multiple objects to be classified under each classification number. Finally, the initial classification results under each classification number are evaluated to determine the appropriate classification number (i.e., the target classification number) and classification results (i.e., the target classification results).

[0156] Through the attribute information and historical behavior information of multiple objects to be classified, the corresponding cumulative function curves of multiple objects to be classified under the attribute information can be quickly constructed to better capture the characteristics and change trends of the objects to be classified. At the same time, by solving the cumulative function curves based on multiple classification quantities, the initial classification results of multiple objects to be classified under each classification quantity can be obtained more intuitively, which is helpful to further analyze and understand the characteristics and behaviors of the objects to be classified, improve the accuracy and reliability of classification, and help decision makers make correct classification decisions.

[0157] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0158] The present application also provides an object classification device. It should be noted that the object classification device of the present application can be used to execute the object classification method provided by the present application. The object classification device provided by the present application is introduced below.

[0159] Figure 4 is a schematic diagram of a device for classifying objects according to an embodiment of the present application. Figure 4 As shown, the device includes: a calculation unit 401, a solution unit 402, an evaluation unit 403, and a determination unit 404.

[0160] A calculation unit 401 is used to calculate based on the attribute information of the multiple objects to be classified and the historical behavior information of each object to be classified in the target organization, and obtain the cumulative function curve corresponding to the multiple objects to be classified under the attribute information;

[0161] A solving unit 402 is used to solve the cumulative function curve according to multiple classification quantities to obtain initial classification results corresponding to multiple objects to be classified under each classification quantity;

[0162] An evaluation unit 403 is used to evaluate the initial classification results corresponding to each classification quantity of the multiple objects to be classified according to the attribute information and the historical behavior information to obtain multiple evaluation values;

[0163] The determination unit 404 is used to determine the target classification number from multiple classification numbers based on multiple evaluation values, and determine the target classification result based on the initial classification result corresponding to the target classification number, wherein the target organization determines the processing strategies corresponding to multiple objects to be classified based on the target classification result.

[0164] The object classification device provided in the embodiment of the present application comprises a calculation unit 401 which performs calculations based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in a target organization to obtain a cumulative function curve corresponding to the multiple objects to be classified under the attribute information; a solution unit 402 solves the cumulative function curves respectively based on multiple classification quantities to obtain initial classification results corresponding to the multiple objects to be classified under each classification quantity; an evaluation unit 403 evaluates the initial classification results corresponding to the multiple objects to be classified under each classification quantity based on the attribute information and the historical behavior information to obtain multiple evaluation values; a determination unit 404 determines a target classification quantity from the multiple classification quantities based on the multiple evaluation values, and determines a target classification result based on the initial classification result corresponding to the target classification quantity, wherein the target organization determines the processing strategies corresponding to the multiple objects to be classified based on the target classification result, thereby solving the problem in the related art that the target objects are classified manually, resulting in a relatively low accuracy rate in classifying the target objects.

[0165] In this scheme, based on the attribute information corresponding to multiple objects to be classified and their historical behavior information in the target organization, a cumulative function curve corresponding to each object to be classified under the attribute information is constructed. Then, through multiple preset classification numbers, the cumulative function curve is solved respectively to obtain the initial classification results corresponding to multiple objects to be classified under each classification number. Finally, the initial classification results under each classification number are evaluated to determine the appropriate classification number (i.e., the target classification number) and classification results (i.e., the target classification results).

[0166] Through the attribute information and historical behavior information of multiple objects to be classified, the corresponding cumulative function curves of multiple objects to be classified under the attribute information can be quickly constructed to better capture the characteristics and change trends of the objects to be classified. At the same time, by solving the cumulative function curves based on multiple classification quantities, the initial classification results of multiple objects to be classified under each classification quantity can be obtained more intuitively, which is helpful to further analyze and understand the characteristics and behaviors of the objects to be classified, improve the accuracy and reliability of classification, and help decision makers make correct classification decisions.

[0167] Optionally, in the object classification device provided in the embodiment of the present application, the calculation unit includes: a first determination module, used to determine the first numerical value corresponding to multiple objects to be classified based on attribute information; a first calculation module, used to calculate based on the historical behavior information of each object to be classified in the target organization to obtain a second numerical value; a second determination module, used to obtain a cumulative function curve based on the first numerical value and the second numerical value.

[0168] Optionally, in the object classification device provided in the embodiment of the present application, the solving unit includes: a third determination module, used to determine the starting point in the cumulative function curve; a solving module, used to solve the second numerical value corresponding to the starting point and the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm for each classification quantity, to obtain a first coordinate point, and determine the classification range based on the first coordinate point and the starting point; a judgment module, used to judge whether the number of classifications after classifying multiple objects to be classified within the classification range is equal to the number of classifications; an execution module, used to repeat the step of solving the first coordinate point and the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm to obtain a second coordinate point if the number of classifications is not equal to the number of classifications, until an initial classification result corresponding to the number of classifications is obtained.

[0169] Optionally, in the object classification device provided in the embodiment of the present application, the solution module includes: a first calculation submodule, used to calculate the cumulative function curve based on the first sensitivity coefficient, the second sensitivity coefficient and the starting point for each classification quantity through the target cutting algorithm to obtain a candidate point set, wherein the first sensitivity coefficient is used to characterize the degree of difference in the mean between the historical behavior information corresponding to each category after classification, and the second sensitivity coefficient is used to characterize the degree of difference in the number of objects corresponding to each category after classification; a second calculation submodule, used to calculate multiple candidate coordinate points in the candidate point set according to the IV indicator to obtain the indicator value corresponding to each candidate coordinate point; a determination submodule, used to determine the first coordinate point from the multiple candidate coordinate points in the candidate point set according to the indicator value corresponding to each candidate coordinate point.

[0170] Optionally, in the object classification device provided in the embodiment of the present application, the evaluation unit includes: an encoding module, which is used to encode the attribute information of multiple objects to be classified to obtain the encoded attribute information; a second calculation module, which is used to perform calculations based on the historical behavior information, the encoded attribute information and the attribute information to obtain a first evaluation value corresponding to the initial classification result; a third calculation module, which is used to perform calculations based on the attribute information and the initial classification result to obtain a second evaluation value corresponding to the initial classification result; and a fourth calculation module, which is used to perform calculations based on the first evaluation value and the second evaluation value to obtain multiple evaluation values.

[0171] Optionally, in the object classification device provided in the embodiment of the present application, the second calculation module includes: a third calculation submodule, which is used to perform calculations based on the encoded attribute information and the attribute information to obtain a random error term corresponding to each object to be classified; a fourth calculation submodule, which is used to perform calculations based on the random error term, the encoded attribute information and the historical behavior information to obtain a first calculation result; a fifth calculation submodule, which is used to perform calculations based on the encoded attribute information and the historical behavior information to obtain a second calculation result; and a sixth calculation submodule, which is used to perform calculations based on the first calculation result and the second calculation result to obtain a first evaluation value.

[0172] Optionally, in the object classification device provided in the embodiment of the present application, the determination unit includes: a screening module, which is used to screen multiple objects to be classified and the initial classification results according to preset screening conditions, and obtain a first sample number corresponding to the multiple objects to be classified after screening and a second sample number corresponding to the initial classification results after screening; a fifth calculation module, which is used to calculate based on the first sample number and the second sample number to obtain a credibility index value corresponding to the initial classification result; an adjustment module, which is used to adjust the initial classification result according to the credibility index value to obtain an adjusted initial classification result; and a fourth determination module, which is used to determine the adjusted initial classification result as the target classification result.

[0173] The object classification device includes a processor and a memory. The above-mentioned calculation unit 401, solution unit 402, evaluation unit 403, determination unit 404, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to achieve accurate classification of the object.

[0174] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the accurate classification of objects can be achieved by adjusting the kernel parameters.

[0175] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0176] An embodiment of the present invention provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a method for classifying objects is implemented.

[0177] An embodiment of the present invention provides a processor, which is used to run a program, wherein a classification method of an object is executed when the program is run.

[0178] like Figure 5 As shown, an embodiment of the present invention provides an electronic device, the device includes a processor, a memory and a program stored in the memory and executable on the processor, and the processor implements the following steps when executing the program: calculating based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in a target organization, and obtaining the cumulative function curve corresponding to the multiple objects to be classified under the attribute information; solving the cumulative function curve respectively according to multiple classification quantities, and obtaining the initial classification results corresponding to the multiple objects to be classified under each classification quantity; evaluating the initial classification results corresponding to the multiple objects to be classified under each classification quantity according to the attribute information and the historical behavior information, and obtaining multiple evaluation values; determining the target classification quantity from the multiple classification quantities according to the multiple evaluation values, and determining the target classification result according to the initial classification result corresponding to the target classification quantity, wherein the target organization determines the processing strategies corresponding to the multiple objects to be classified according to the target classification result.

[0179] Optionally, calculations are performed based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in the target organization to obtain a cumulative function curve corresponding to the multiple objects to be classified under the attribute information, including: determining a first value corresponding to the multiple objects to be classified based on the attribute information; calculating based on the historical behavior information of each object to be classified in the target organization to obtain a second value; and obtaining a cumulative function curve based on the first value and the second value.

[0180] Optionally, the cumulative function curve is solved separately according to multiple classification quantities to obtain initial classification results corresponding to multiple objects to be classified under each classification quantity, including: determining a starting point in the cumulative function curve; for each classification quantity, solving the second numerical value corresponding to the starting point and the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm to obtain a first coordinate point, and determining a classification range based on the first coordinate point and the starting point; judging whether the number of categories after classifying multiple objects to be classified under the classification range is equal to the number of categories; if the number of categories is not equal to the number of categories, repeatedly executing the step of solving the first coordinate point and the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm to obtain a second coordinate point, until the initial classification result corresponding to the classification quantity is obtained.

[0181] Optionally, for each number of classifications, the target cutting algorithm is used to solve the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the starting point and the cumulative function curve to obtain the first coordinate point, including: for each number of classifications, the target cutting algorithm is used to calculate the cumulative function curve based on the first sensitivity coefficient, the second sensitivity coefficient and the starting point to obtain a candidate point set, wherein the first sensitivity coefficient is used to characterize the degree of difference in the mean between the historical behavior information corresponding to each category after classification, and the second sensitivity coefficient is used to characterize the degree of difference in the number of objects corresponding to each category after classification; multiple candidate coordinate points in the candidate point set are calculated according to the IV indicator to obtain the indicator value corresponding to each candidate coordinate point; based on the indicator value corresponding to each candidate coordinate point, the first coordinate point is determined from the multiple candidate coordinate points in the candidate point set.

[0182] Optionally, based on the attribute information and the historical behavior information, the initial classification results corresponding to multiple objects to be classified under each classification quantity are evaluated to obtain multiple evaluation values, including: encoding the attribute information of multiple objects to be classified to obtain the encoded attribute information; calculating based on the historical behavior information, the encoded attribute information and the attribute information to obtain a first evaluation value corresponding to the initial classification result; calculating based on the attribute information and the initial classification result to obtain a second evaluation value corresponding to the initial classification result; calculating based on the first evaluation value and the second evaluation value to obtain multiple evaluation values.

[0183] Optionally, calculation is performed based on historical behavior information, encoded attribute information, and attribute information to obtain a first evaluation value corresponding to the initial classification result, including: calculation is performed based on the encoded attribute information and attribute information to obtain a random error term corresponding to each object to be classified; calculation is performed based on the random error term, encoded attribute information, and historical behavior information to obtain a first calculation result; calculation is performed based on the encoded attribute information and historical behavior information to obtain a second calculation result; calculation is performed based on the first calculation result and the second calculation result to obtain a first evaluation value.

[0184] Optionally, determining the target classification result based on the initial classification result corresponding to the target classification quantity includes: screening multiple objects to be classified and the initial classification result according to preset screening conditions to obtain a first sample quantity corresponding to the screened multiple objects to be classified and a second sample quantity corresponding to the screened initial classification result; calculating based on the first sample quantity and the second sample quantity to obtain a credibility index value corresponding to the initial classification result; adjusting the initial classification result based on the credibility index value to obtain an adjusted initial classification result; and determining the adjusted initial classification result as the target classification result.

[0185] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0186] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: calculating based on attribute information of multiple objects to be classified and historical behavior information of each object to be classified in a target organization, obtaining a cumulative function curve corresponding to the multiple objects to be classified under the attribute information; solving the cumulative function curve respectively according to multiple classification quantities, obtaining initial classification results corresponding to the multiple objects to be classified under each classification quantity; evaluating the initial classification results corresponding to the multiple objects to be classified under each classification quantity according to the attribute information and the historical behavior information, obtaining multiple evaluation values; determining a target classification quantity from the multiple classification quantities according to the multiple evaluation values, and determining a target classification result according to the initial classification result corresponding to the target classification quantity, wherein the target organization determines the processing strategies corresponding to the multiple objects to be classified according to the target classification result.

[0187] Optionally, calculations are performed based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in the target organization to obtain a cumulative function curve corresponding to the multiple objects to be classified under the attribute information, including: determining a first value corresponding to the multiple objects to be classified based on the attribute information; calculating based on the historical behavior information of each object to be classified in the target organization to obtain a second value; and obtaining a cumulative function curve based on the first value and the second value.

[0188] Optionally, the cumulative function curve is solved separately according to multiple classification quantities to obtain initial classification results corresponding to multiple objects to be classified under each classification quantity, including: determining a starting point in the cumulative function curve; for each classification quantity, solving the second numerical value corresponding to the starting point and the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm to obtain a first coordinate point, and determining a classification range based on the first coordinate point and the starting point; judging whether the number of categories after classifying multiple objects to be classified under the classification range is equal to the number of categories; if the number of categories is not equal to the number of categories, repeatedly executing the step of solving the first coordinate point and the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the cumulative function curve through a target cutting algorithm to obtain a second coordinate point, until the initial classification result corresponding to the classification quantity is obtained.

[0189] Optionally, for each number of classifications, the target cutting algorithm is used to solve the second numerical value corresponding to the historical behavior information of multiple objects to be classified in the starting point and the cumulative function curve to obtain the first coordinate point, including: for each number of classifications, the target cutting algorithm is used to calculate the cumulative function curve based on the first sensitivity coefficient, the second sensitivity coefficient and the starting point to obtain a candidate point set, wherein the first sensitivity coefficient is used to characterize the degree of difference in the mean between the historical behavior information corresponding to each category after classification, and the second sensitivity coefficient is used to characterize the degree of difference in the number of objects corresponding to each category after classification; multiple candidate coordinate points in the candidate point set are calculated according to the IV indicator to obtain the indicator value corresponding to each candidate coordinate point; based on the indicator value corresponding to each candidate coordinate point, the first coordinate point is determined from the multiple candidate coordinate points in the candidate point set.

[0190] Optionally, based on the attribute information and the historical behavior information, the initial classification results corresponding to multiple objects to be classified under each classification quantity are evaluated to obtain multiple evaluation values, including: encoding the attribute information of multiple objects to be classified to obtain the encoded attribute information; calculating based on the historical behavior information, the encoded attribute information and the attribute information to obtain a first evaluation value corresponding to the initial classification result; calculating based on the attribute information and the initial classification result to obtain a second evaluation value corresponding to the initial classification result; calculating based on the first evaluation value and the second evaluation value to obtain multiple evaluation values.

[0191] Optionally, calculation is performed based on historical behavior information, encoded attribute information, and attribute information to obtain a first evaluation value corresponding to the initial classification result, including: calculation is performed based on the encoded attribute information and attribute information to obtain a random error term corresponding to each object to be classified; calculation is performed based on the random error term, encoded attribute information, and historical behavior information to obtain a first calculation result; calculation is performed based on the encoded attribute information and historical behavior information to obtain a second calculation result; calculation is performed based on the first calculation result and the second calculation result to obtain a first evaluation value.

[0192] Optionally, determining the target classification result based on the initial classification result corresponding to the target classification quantity includes: screening multiple objects to be classified and the initial classification result according to preset screening conditions to obtain a first sample quantity corresponding to the screened multiple objects to be classified and a second sample quantity corresponding to the screened initial classification result; calculating based on the first sample quantity and the second sample quantity to obtain a credibility index value corresponding to the initial classification result; adjusting the initial classification result based on the credibility index value to obtain an adjusted initial classification result; and determining the adjusted initial classification result as the target classification result.

[0193] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0194] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0195] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0196] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0197] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0198] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0199] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0200] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0201] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0202] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for classifying an object, characterized in that: include: Calculating based on the attribute information of multiple objects to be classified and the historical behavior information of each object to be classified in the target organization, obtaining the cumulative function curve corresponding to the multiple objects to be classified under the attribute information; Solving the cumulative function curves respectively according to multiple classification numbers to obtain initial classification results corresponding to the multiple objects to be classified under each classification number; According to the attribute information and the historical behavior information, the initial classification results corresponding to the multiple objects to be classified under each classification quantity are evaluated to obtain multiple evaluation values; Based on the multiple evaluation values, a target classification number is determined from the multiple classification numbers, and a target classification result is determined based on an initial classification result corresponding to the target classification number, wherein the target mechanism determines a processing strategy corresponding to the multiple objects to be classified based on the target classification result.

2. The method according to claim 1, characterized in that The calculation is performed based on the attribute information of the multiple objects to be classified and the historical behavior information of each object to be classified in the target organization to obtain the cumulative function curve corresponding to the multiple objects to be classified under the attribute information, including: Determining first values ​​corresponding to the plurality of objects to be classified according to the attribute information; Calculate the historical behavior information of each object to be classified in the target organization to obtain a second value; The cumulative function curve is obtained according to the first value and the second value.

3. The method according to claim 1, characterized in that Solving the cumulative function curves respectively according to multiple classification numbers to obtain the initial classification results corresponding to the multiple objects to be classified under each classification number includes: Determining a starting point in the cumulative function curve; For each number of classifications, solving the second numerical value corresponding to the historical behavior information of the plurality of objects to be classified in the starting point and the cumulative function curve by a target cutting algorithm to obtain a first coordinate point, and determining a classification range according to the first coordinate point and the starting point; Determining whether the number of categories after classifying the plurality of objects to be classified within the classification range is equal to the number of categories; If the number of categories is not equal to the number of categories, then repeatedly execute the step of solving the second numerical value corresponding to the historical behavior information of the first coordinate point and the multiple objects to be classified in the cumulative function curve through the target cutting algorithm to obtain the second coordinate point until the initial classification result corresponding to the number of categories is obtained.

4. The method according to claim 3, characterized in that For each number of classifications, the target cutting algorithm is used to solve the second numerical value corresponding to the historical behavior information of the plurality of objects to be classified in the starting point and the cumulative function curve, and the first coordinate point is obtained including: For each number of classifications, the target cutting algorithm is used to calculate the cumulative function curve based on the first sensitivity coefficient, the second sensitivity coefficient and the starting point to obtain a candidate point set, wherein the first sensitivity coefficient is used to characterize the degree of difference in the mean values ​​of the historical behavior information corresponding to each category after classification, and the second sensitivity coefficient is used to characterize the degree of difference in the number of objects corresponding to each category after classification; Calculating multiple candidate coordinate points in the candidate point set according to the IV index to obtain an index value corresponding to each candidate coordinate point; The first coordinate point is determined from a plurality of candidate coordinate points in the candidate point set according to an index value corresponding to each candidate coordinate point.

5. The method according to claim 1, characterized in that Based on the attribute information and the historical behavior information, the initial classification results corresponding to the multiple objects to be classified under each classification quantity are evaluated to obtain multiple evaluation values ​​including: Encoding the attribute information of the plurality of objects to be classified to obtain encoded attribute information; Calculating according to the historical behavior information, the encoded attribute information and the attribute information to obtain a first evaluation value corresponding to the initial classification result; Calculating according to the attribute information and the initial classification result to obtain a second evaluation value corresponding to the initial classification result; The plurality of evaluation values ​​are obtained by performing calculations according to the first evaluation value and the second evaluation value.

6. The method according to claim 5, characterized in that Calculating according to the historical behavior information, the encoded attribute information, and the attribute information to obtain a first evaluation value corresponding to the initial classification result includes: Calculating according to the encoded attribute information and the attribute information to obtain a random error term corresponding to each object to be classified; Performing calculation according to the random error term, the encoded attribute information and the historical behavior information to obtain a first calculation result; Performing calculation based on the encoded attribute information and the historical behavior information to obtain a second calculation result; Calculation is performed according to the first calculation result and the second calculation result to obtain the first evaluation value.

7. The method according to claim 1, characterized in that Determining the target classification results according to the initial classification results corresponding to the target classification quantity includes: Screening the multiple objects to be classified and the initial classification results according to a preset screening condition to obtain a first sample quantity corresponding to the screened multiple objects to be classified and a second sample quantity corresponding to the screened initial classification results; Calculating according to the first sample quantity and the second sample quantity to obtain a credibility index value corresponding to the initial classification result; Adjusting the initial classification result according to the credibility index value to obtain an adjusted initial classification result; The adjusted initial classification result is determined as the target classification result.

8. An object classification device, characterized in that: include: A calculation unit, used to calculate according to the attribute information of the plurality of objects to be classified and the historical behavior information of each object to be classified in the target organization, to obtain the cumulative function curve corresponding to the plurality of objects to be classified under the attribute information; A solving unit, used for solving the cumulative function curve respectively according to a plurality of classification quantities to obtain initial classification results corresponding to the plurality of objects to be classified under each classification quantity; An evaluation unit, configured to evaluate the initial classification results corresponding to each number of classifications of the plurality of objects to be classified according to the attribute information and the historical behavior information, to obtain a plurality of evaluation values; A determination unit is used to determine a target classification number from the multiple classification numbers based on the multiple evaluation values, and to determine a target classification result based on an initial classification result corresponding to the target classification number, wherein the target mechanism determines a processing strategy corresponding to the multiple objects to be classified based on the target classification result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the storage medium is controlled to execute the object classification method according to any one of claims 1 to 7 on a device.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the object classification method described in any one of claims 1 to 7.

Images