CN115204285A - Method, device and equipment for establishing rating model - Google Patents
Method, device and equipment for establishing rating model Download PDFInfo
- Publication number
- CN115204285A CN115204285A CN202210806356.9A CN202210806356A CN115204285A CN 115204285 A CN115204285 A CN 115204285A CN 202210806356 A CN202210806356 A CN 202210806356A CN 115204285 A CN115204285 A CN 115204285A
- Authority
- CN
- China
- Prior art keywords
- rating
- target objects
- grading
- target
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The embodiment of the invention relates to the technical field of data processing, and discloses a method, a device and equipment for establishing a rating model, wherein the establishing method comprises the following steps: acquiring the rating data of a first preset number of target objects; sequentially grading and dividing the target objects based on a second preset number of expert experience grading schemes to obtain grading results; and taking the rating result as a rating target, and establishing a rating model by utilizing an integration algorithm or a neural network algorithm based on the rating data. According to the method and the device, the target object is graded and divided by utilizing various different expert experience grading schemes, the obtained grading result is used as a grading target to establish a grading model, the various expert experience grading schemes are effectively fused, and an artificial intelligent grading model is established, so that the technical effects of improving the grading accuracy and stability and efficiently and quickly grading the target object are achieved.
Description
Technical Field
The embodiment of the invention relates to the technical field of data processing, in particular to a method, a device and equipment for establishing a rating model.
Background
At present, most of the grading techniques of the objects to be graded adopt expert methods to carry out experience evaluation, and the value potential of the objects to be graded is evaluated by collecting relevant information of the objects to be graded.
Generally, the qualified experts can construct an accurate rating system according to their own rich experience, but most of such rating system evaluation tables can only be applied to a specific field range, and the personal experience preferences of each expert are also different, which may cause that for the same object to be rated, the evaluation results of two experts may have a large difference, and the judgments of both experts may be inaccurate.
Disclosure of Invention
The embodiment of the invention provides a method, a device and equipment for establishing a rating model, and solves the technical problems of more human participation and unstable rating accuracy caused by that a target object can only be rated according to expert experience in the prior art.
In a first aspect, an embodiment of the present invention provides a method for building a rating model, where the method includes:
obtaining rating data of a first preset number of target objects, wherein the target objects are objects to be rated, and the rating data is related information of the target objects determined according to rating requirements;
sequentially grading and dividing the target object based on a second preset number of expert experience grading schemes to obtain grading results;
and taking the rating result as a rating target, and establishing a rating model by utilizing an integrated algorithm or a neural network algorithm based on the rating data.
In a second aspect, an embodiment of the present invention further provides a device for building a rating model, where the device includes:
the device comprises a data acquisition unit, a storage unit and a display unit, wherein the data acquisition unit is used for acquiring rating data of a first preset number of target objects, the target objects are objects to be rated, and the rating data are related information of the target objects determined according to rating requirements;
the experience rating unit is used for sequentially rating and dividing the target objects based on a second preset number of expert experience rating schemes to obtain rating results;
and the model establishing unit is used for establishing a rating model by utilizing an integration algorithm or a neural network algorithm based on the rating data by taking the rating result as a rating target.
In a third aspect, an embodiment of the present invention further provides a device for establishing a rating model, where the device for establishing a rating model includes:
one or more processors;
storage means for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement a method of building a rating model as any of the first aspect of embodiments of the invention.
In a fourth aspect, the embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the method for building a rating model according to any of the first aspect of the embodiment of the present invention.
In a fifth aspect, embodiments of the present invention further provide a computer program product, including a computer program, which when executed by a processor, implements the method for establishing a rating model according to any of the first aspects of the embodiments of the present invention.
The embodiment of the invention discloses a method, a device and equipment for establishing a rating model, wherein the establishing method comprises the following steps: acquiring rating data of a first preset number of target objects, wherein the target objects are to-be-rated objects, and the rating data are related information of the target objects determined according to rating requirements; sequentially grading and dividing the target objects based on a second preset number of expert experience grading schemes to obtain grading results; and taking the rating result as a rating target, and establishing a rating model by utilizing an integration algorithm or a neural network algorithm based on the rating data. According to the method and the device, the target object is graded and divided by utilizing various different expert experience grading schemes, the obtained grading result is used as the grading target to establish the grading model, the various expert experience grading schemes are effectively fused to establish the artificial intelligent grading model, the technical problems that in the prior art, the target object can be graded only according to the expert experience, so that more artificial participation is caused, and the grading accuracy is unstable are solved, the grading accuracy and stability are improved, and the technical effect of grading the target object efficiently and quickly is achieved.
Drawings
FIG. 1 is a flow chart of a method for building a rating model according to an embodiment of the present invention;
FIG. 2 is a flow chart of another method for building a rating model provided by an embodiment of the present invention;
FIG. 3 is a block diagram of an apparatus for building a rating model according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a rating model establishing device according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
It should be noted that the terms "first", "second", and the like in the description and claims of the present invention and the accompanying drawings are used for distinguishing different objects, and are not used for limiting a specific order. The following embodiments of the present invention may be implemented individually, or may be implemented in combination with each other, and the embodiments of the present invention are not limited in this respect.
Fig. 1 is a flowchart of a rating model establishing method according to an embodiment of the present invention. The rating model building method may be performed by a rating model building apparatus, which may be implemented in hardware and/or software, and may be generally integrated in a server. According to the technical scheme, the data acquisition, storage, use, processing and the like meet relevant regulations of national laws and regulations.
As shown in fig. 1, the method for establishing the rating model specifically includes the following steps:
s101, obtaining rating data of a first preset number of target objects, wherein the target objects are to-be-rated objects, and the rating data are related information of the target objects determined according to rating requirements.
Specifically, the target object may be a target customer group of the marketing service, a product to be rated, or the like, which is not limited herein. When the target object is a target guest group of the marketing service, the rating data can be consumption level, credit condition, consumption range and the like of the target guest group; when the target object is a product to be rated, the rating data may be basic factory parameters, factory detection information, and the like of the product. In order to ensure the randomness and diversity of the samples, the first preset number is as large as possible, and may be set to 1000, for example.
After the rating data of the target object is obtained, the rating data needs to be cleaned and processed, for example, some rating data of the target object may be abnormal or missing, and the rating data may be estimated, supplemented, deleted or directly replaced by 0 according to actual requirements.
And S102, carrying out rating division on the target objects in sequence based on a second preset number of expert experience rating schemes to obtain rating results.
Specifically, after the rating data of a first preset number of target objects are obtained, the target objects are graded and divided in sequence by using a plurality of expert experience grading schemes, specifically, one grading scheme is selected from a second preset number of expert experience grading schemes, the target objects are graded based on the obtained grading data, the highest and lowest grades are selected, then the rest target objects are graded by using other expert experience grading schemes until the target objects are graded completely, and a grading result is obtained.
Specifically, a second preset number of expert experience rating schemes are used for performing step-by-step rating, namely, different expert experience rating schemes are used for rating the target object step by step, and a part of rating division results with the highest rating and the lowest rating are selected as true values (namely, the rating results) for subsequent artificial intelligence algorithm learning (namely, subsequent modeling use). And after the highest and lowest rating parts are selected each time, continuously rating the highest and lowest levels of the remaining target objects by adopting other expert experience rating schemes, extracting the highest and lowest levels, and sequentially proceeding until the grade division of all the target objects is completed.
It should be noted that, in the second preset number of expert experience rating schemes, the expert experience rating scheme that considers the most accurate rating effect as the first scheme may be preferentially used for rating and dividing the target object according to experience, the expert experience rating scheme that considers the least accurate rating effect as the first scheme may also be preferentially used for rating and dividing the target object, and the expert experience rating schemes of the second preset number may also be randomly ordered to sequentially rate and divide the target object.
And S103, taking the rating result as a rating target, and establishing a rating model by utilizing an integrated algorithm or a neural network algorithm based on the rating data.
Specifically, a rating result obtained by rating and dividing a target object by using various expert experience rating schemes is used as a rating target, then target characteristic data is selected from the rating data and substituted into an integration algorithm or a neural network algorithm to establish a rating model. The target characteristic data are data selected from the rating data of the target object according to rating requirements, illustratively, the anti-falling-down capability of the toy car needs to be rated, anti-falling-down data measured when a certain toy car falls from different heights needs to be selected from the rating data, and correspondingly, data related to the electric quantity cruising capability of the toy car can be used without being selected for modeling.
In the embodiment of the invention, the target characteristic data required by rating is determined from the rating data and is used as the characteristic index for establishing the rating model, and meanwhile, the rating results obtained by rating and dividing various expert experience rating schemes are used as the rating targets for machine learning simulation, so that each expert experience rating scheme can be rapidly learned, the characteristics of each scheme are fused, and the rating model with higher accuracy is established.
According to the method and the device, the target object is graded and divided by utilizing various different expert experience grading schemes, the obtained grading result is used as the grading target to establish the grading model, the various expert experience grading schemes are effectively fused to establish the artificial intelligent grading model, the technical problems that in the prior art, the target object can be graded only according to the expert experience, so that more artificial participation is caused, and the grading accuracy is unstable are solved, the grading accuracy and stability are improved, and the technical effect of grading the target object efficiently and quickly is achieved.
On the basis of the foregoing technical solutions, when the first preset number is S, the second preset number is m, and n levels are obtained by rating division, where m is less than n, S102 specifically includes:
selecting any one of the m expert experience rating schemes to perform rating division on the s target objects, selecting x target objects with the highest grade from the division results as a 1 st grade, and selecting x target objects with the lowest grade as an nth grade, wherein x is the number determined by multiplying s by a first preset percentage;
selecting any one of the remaining m-1 expert experience rating schemes to perform rating division on s-2x target objects, selecting y target objects with the highest grade from the division results as a 2 nd grade, and selecting y target objects with the lowest grade as an n-1 st grade, wherein y is the number determined by multiplying (s-2 x) by a second preset percentage;
repeating the grading process until the rest s1 target objects are graded and divided by using the last 1 expert experience grading scheme, excluding the z target objects with the highest grade and the z target objects with the lowest grade from the dividing result, and completely dividing the rest target objects into the rest grades in the n grades, wherein z is the number determined by multiplying the s1 by a third preset percentage.
Illustratively, the first preset number s is 1000, the second preset number m is 3, and the rank n is 7, that is, there are three expert experience rating schemes A, B, C, and it is necessary to divide 1000 target customers into 7 ranks. Then scheme a is used to perform a first round of rating on 1000 target customers, and the 100 people with the highest rating score (x above) and the 100 people with the lowest rating score (x above) are taken out and labeled with rating labels 1 and 7, respectively. The remaining 800 people (s-2 x above) are then ranked using scheme B for a second round, and the 100 people with the highest ranking score (y above) and the 100 people with the lowest ranking score (y above) are removed and given rank labels 2 and 6, respectively. Finally, the remaining 600 people (s 1 above) are rated for a third round using scheme C, the 150 people with the highest rating score (z above) and the 150 people with the lowest rating score (z above) are taken out and are labeled with rating labels 3 and 5, respectively. The remaining 300 people all marked a level 4 label.
It should be noted that the first preset percentage, the second preset percentage, and the third preset percentage may be set according to actual needs.
Alternatively, when n is an odd number, after z target objects of the highest rank and z target objects of the lowest rank are excluded from the division results, the remaining target objects are all divided into (n + 1)/2 levels; when n is an even number, after z target objects of the highest rank and z target objects of the lowest rank are excluded from the division result, the remaining target objects are equally divided into n/2 ranks and (n/2) +1 ranks.
Illustratively, when n is 7 (odd number), and 2m > n, after excluding z target objects of the highest rank and z target objects of the lowest rank from the division results, all the remaining target objects are divided into 4 ranks; when n is 7 (odd number) and 2m < n, for example, when m is 3 as described above, after excluding z target objects of the highest rank and z target objects of the lowest rank from the division results, the remaining target objects are divided into (n-1)/2 levels and (n + 3)/2 levels, and then the remaining target clients are directly divided into (n + 1)/2 levels.
When n is an even number, after z target objects with the highest rank and z target objects with the lowest rank are excluded from the division result, the remaining target objects are equally allocated and divided into n/2 ranks and (n/2) +1 ranks.
It should be noted that, if the expert experience rating scheme is too many, that is, m is large, the first preset number of the target objects is small, that is, s is small, and the division level n is small, there may be a case where the target objects are already divided but the expert experience rating scheme is not used up, at this time, the number of the levels n may be adjusted as needed, and the values of the first preset percentage, the second preset percentage, and the third preset percentage may be adjusted as needed.
In the embodiment of the invention, one of the main reasons for adopting the gradual grading and layering mode is that the method can skillfully contain various expert experience grading results, and the expert experience grading results contain a large amount of evaluation rule information in an expert experience grading scheme, so that the expert experience grading scheme can be effectively learned during modeling; another reason is that for any expert experience rating scheme, the two parts with the highest and lowest scores are relatively accurate, so called stepwise, that is, gradually expanding from the beginning to the end towards the middle. And the customers with scores in the middle part have the most difficult evaluation accuracy, so that the difficulty is directly avoided, the part which is most difficult to evaluate in the middle is directly and uniformly classified into a middle grade, and the middle grade requires more people than other grades, so that the effectiveness of the empirical method can be ensured to the maximum extent, and the false information learned by the artificial intelligence algorithm is avoided.
On the basis of the foregoing technical solutions, when the first preset number is S, the second preset number is m, and n levels are obtained by rating division, where m is greater than n, S102 specifically includes:
selecting at least two of the m expert experience rating schemes to rate and divide the s target objects;
adding the scores of the two expert experience rating schemes and taking the average value to obtain an average dividing result, selecting a fourth preset percentage of target objects from the average dividing result as the 1 st level, and selecting a fourth preset percentage of target objects as the nth level;
selecting at least two target objects with the remaining s2 from the remaining m-2 expert experience rating schemes for rating and dividing, selecting the target objects with the fifth preset percentage from the dividing results as the 2 nd level, and selecting the target objects with the fifth preset percentage from the dividing results as the n-1 st level;
repeating the grading process until the last 1 expert experience grading scheme is used for grading and dividing the rest s3 target objects, eliminating the target objects with the first sixth preset percentage and the target objects with the second sixth preset percentage from the grading result, and completely grading the rest target objects into the rest levels of the n levels.
Specifically, when the number of the expert experience rating schemes is large relative to the number of the grades, that is, m is greater than n, it is only necessary to perform weighting by using more than one expert experience rating scheme in each grade-by-grade rating to obtain a composite score.
Illustratively, the first preset number s is 1000, the second preset number m is 8, and the level n is 6, that is, there are 8 expert experience rating schemes and 6 levels, two expert experience rating schemes may be directly adopted to rate 1000 target objects in the first round of rating, scores of the two expert experience rating schemes are subjected to weighted average calculation, wherein values of weights may be set according to rating emphasis points and rating requirements of different expert experience rating schemes, after obtaining the weighted average, the former ten percent (the fourth preset percentage) is taken as level 1, the latter ten percent (the fourth preset percentage) is taken as level 6, and the remaining target objects are subjected to weighted rating continuously by using the other two expert experience rating schemes until all ratings are completed, so as to obtain a rating result.
Optionally, when n is an odd number, after the sixth preset percentage of the target objects and the sixth preset percentage of the target objects are excluded from the division result, dividing all the remaining target objects into (n + 1)/2 levels; when n is an even number, after the first sixth preset percentage of target objects and the second sixth preset percentage of target objects are excluded from the division result, the remaining target objects are equally divided into n/2 levels and (n/2) +1 levels.
Illustratively, when n is 7 (odd), after the first sixth preset percentage of target objects and the second sixth preset percentage of target objects are excluded from the division result, the remaining target objects are all divided into 4 levels, and when n is 8 (even), after the first sixth preset percentage of target objects and the second sixth preset percentage of target objects are excluded from the division result, the remaining target objects are equally divided into 4 levels and 5 levels.
On the basis of the above technical solutions, S103 specifically includes:
and taking the rating result as a rating target, and establishing a rating model by utilizing a lightGBM algorithm based on the rating data.
On the basis of the above technical solutions, S103 specifically includes:
and taking the rating result as a rating target, and establishing a rating model by utilizing a full-connection neural network based on the rating data.
In particular, to complete the modeling work, the appropriate input variables, i.e., data characteristics, need to be determined. And selecting target characteristic data from the rating data as characteristic indexes required by the model, and matching with the rating result given by the expert experience rating scheme in the step S102 as a rating target, namely, quickly learning the comprehensive experience of a plurality of experts by adopting a model algorithm, wherein the target characteristic data is selected from the rating data of the target object according to the rating requirement.
The model algorithm adopted in the embodiment of the present invention mainly refers to an artificial intelligence technique and a big data method, which includes a large number of models that can be used for problem classification, and no specific limitation is imposed on the model, but an integrated algorithm and a neural network model are preferably used, wherein a Light Gradient Boosting Machine (Light Gradient Boosting Machine) is preferred in the integrated algorithm, and a fully connected neural network (DNN) is preferred in the neural network algorithm.
Illustratively, taking a neural network algorithm as an example, a conventional neuron function is
When the data transmitted by the upper level neuron exceeds a certain threshold value theta, the current neuron outputs a value y to the lower level neuron, wherein,
f is a function, the basic configuration of which has no specific requirement and only one specific setting: when the input value is less than 0, the function output is 0, and when the input value is greater than 0, the function output is a positive value. For the deep neural network model, the model needs to be set to be a multi-classification problem scene setting, the output result of the model is the prediction probability values of multiple categories, the number of the categories corresponds to the number of levels in a grading scene, and the model takes the boundary corresponding to the category with the maximum prediction probability as the prediction result of the model.
It should be noted that, in the embodiment of the present invention, the specific design of the model may also be changed in various ways according to actual requirements, for example, the above neural network algorithm may not be limited to the conventional neural network structure, and for the neuron function
If the variation trend corresponding to the real data is not current, the function setting of f can be adjusted according to the observation result, for example, a polynomial structure is adopted
Curve of square term
Or only part of specific upper-level neurons (assuming that the specific upper-level neuron set is I) are adopted for data information feedback transmission:
on the basis of the foregoing technical solutions, fig. 2 is a flowchart of another method for establishing a rating model according to an embodiment of the present invention, and as shown in fig. 2, after the rating model is established in S103, the method further includes:
s201, rating the target object by using a rating model.
Specifically, after the rating model is trained, rating data of some new target objects needs to be input, a prediction result is output by using the rating model, namely, the new target objects are graded, then the graded grades are given to an experienced expert for manual evaluation, so as to detect whether the grading result of the model is reasonable and accurate, and after the manual inspection is obtained, the inspected rating model is applied to a required rating scene.
In the actual use process, new rating data of a target object can be continuously input according to the requirements of a real rating scene, a new rating result is given, and the practical experience is further input into a rating model to further optimize iteration, so that the rating model is continuously updated and evolved, and the accuracy and the stability of rating are improved.
Fig. 3 is a structural diagram of a rating model establishing apparatus according to an embodiment of the present invention, and as shown in fig. 3, the rating model establishing apparatus includes:
the data acquiring unit 31 is configured to acquire rating data of a first preset number of target objects, where the target objects are to-be-rated objects, and the rating data is related information of the target objects determined according to rating requirements;
the experience rating unit 32 is configured to sequentially rate and divide the target objects based on a second preset number of expert experience rating schemes to obtain rating results;
and the model establishing unit 33 is used for establishing a rating model by using an integrated algorithm or a neural network algorithm based on the rating data by taking the rating result as a rating target.
Optionally, when the first preset number is s and the second preset number is m, the ranking division obtains n levels, and m is less than n, the experience ranking unit 32 is specifically configured to:
selecting any one of the m expert experience rating schemes to perform rating division on the s target objects, selecting x target objects with the highest grade from the division results as a 1 st grade, and selecting x target objects with the lowest grade as an nth grade, wherein x is the number determined by multiplying s by a first preset percentage;
selecting any one of the remaining m-1 expert experience rating schemes to perform rating division on s-2x target objects, selecting y target objects with the highest grade from the division results as a 2 nd grade, and selecting y target objects with the lowest grade as an n-1 st grade, wherein y is the number determined by multiplying (s-2 x) by a second preset percentage;
repeating the grading process until the last 1 expert experience grading scheme is used for grading and dividing the rest s1 target objects, excluding the z target objects with the highest grade and the z target objects with the lowest grade from the grading result, and completely grading the rest target objects into the rest grades in the n grades, wherein z is the number determined by multiplying the s1 by a third preset percentage.
Alternatively, when n is an odd number, after z target objects of the highest rank and z target objects of the lowest rank are excluded from the division results, the remaining target objects are all divided into (n + 1)/2 levels; when n is an even number, after z target objects of the highest rank and z target objects of the lowest rank are excluded from the division result, the remaining target objects are equally divided into n/2 ranks and (n/2) +1 ranks.
Optionally, when the first preset number is s, the second preset number is m, the ranking division obtains n levels, and m > n, the experience ranking unit 32 is specifically configured to:
selecting at least two of the m expert experience rating schemes to rate and divide the s target objects;
adding the scores of the two expert experience rating schemes and taking the average value to obtain an average dividing result, selecting a fourth preset percentage of target objects from the average dividing result as the 1 st level, and selecting a fourth preset percentage of target objects as the nth level;
selecting at least two target objects with the remaining s2 from the remaining m-2 expert experience rating schemes for rating and dividing, selecting the target objects with the fifth preset percentage from the dividing results as the 2 nd level, and selecting the target objects with the fifth preset percentage from the dividing results as the n-1 st level;
repeating the grading process until the last 1 expert experience grading scheme is used for grading and dividing the rest s3 target objects, eliminating the target objects with the first sixth preset percentage and the target objects with the second sixth preset percentage from the grading result, and completely grading the rest target objects into the rest levels of the n levels.
Alternatively, when n is an odd number, after the first sixth preset percentage of target objects and the second sixth preset percentage of target objects are excluded from the division result, dividing all the remaining target objects into (n + 1)/2 levels; when n is an even number, after the first sixth preset percentage of target objects and the second sixth preset percentage of target objects are excluded from the division result, the remaining target objects are equally divided into n/2 levels and (n/2) +1 levels.
Optionally, the model building unit 33 is specifically configured to:
and taking the rating result as a rating target, and establishing a rating model by utilizing a lightGBM algorithm based on the rating data.
Optionally, the model establishing unit 33 is specifically configured to:
and taking the rating result as a rating target, and establishing a rating model by utilizing a full-connection neural network based on the rating data.
Optionally, after the model building unit 33 builds the rating model, the building apparatus further includes:
and the rating application unit is used for rating the target object by utilizing the rating model.
The device provided by the embodiment of the present invention has the same implementation principle and technical effect as the method embodiments, and for the sake of brief description, reference may be made to the corresponding contents in the method embodiments without reference to the device embodiments.
The device for establishing the rating model provided by the embodiment of the invention has the same technical characteristics as the method for establishing the rating model provided by the embodiment, so that the same technical problems can be solved, and the same technical effects can be achieved.
Fig. 4 is a schematic structural diagram of a rating model establishing apparatus according to an embodiment of the present invention, as shown in fig. 4, the rating model establishing apparatus includes a processor 41, a memory 42, an input device 43, and an output device 44; the number of processors 41 in the device for establishing the rating model may be one or more, and one processor 41 is taken as an example in fig. 4; the processor 41, the memory 42, the input means 43 and the output means 44 in the rating model building apparatus may be connected by a bus or other means, as exemplified by the bus connection in fig. 4.
The memory 42 is used as a computer-readable storage medium for storing software programs, computer-executable programs, and modules, such as program instructions/modules corresponding to the rating model establishing method in the embodiment of the present invention (for example, the data obtaining unit 31, the experience rating unit 32, and the model establishing unit 33 in the rating model establishing device). The processor 41 executes various functional applications and data processing of the rating model creation device by executing software programs, instructions, and modules stored in the memory 42, that is, implements the rating model creation method described above.
The memory 42 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the terminal, and the like. Further, the memory 42 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some instances, the memory 42 may further include memory located remotely from the processor 41, which may be connected to a rating model creation device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input means 43 may be used to receive input numeric or character information and to generate key signal inputs relating to user settings and function controls of the rating model building apparatus. The output device 44 may include a display device such as a display screen.
Embodiments of the present invention also provide a storage medium containing computer-executable instructions that, when executed by a computer processor, perform a method of building a rating model.
Specifically, the method for establishing the rating model comprises the following steps:
acquiring rating data of a first preset number of target objects, wherein the target objects are to-be-rated objects, and the rating data are related information of the target objects determined according to rating requirements;
sequentially grading and dividing the target objects based on a second preset number of expert experience grading schemes to obtain grading results;
and taking the rating result as a rating target, and establishing a rating model by utilizing an integration algorithm or a neural network algorithm based on the rating data.
Of course, the storage medium provided by the embodiment of the present invention contains computer-executable instructions, and the computer-executable instructions are not limited to the method operations described above, and may also perform related operations in the rating model establishing method provided by any embodiment of the present invention.
From the above description of the embodiments, it is obvious for those skilled in the art that the present invention can be implemented by software and necessary general hardware, and certainly can be implemented by hardware, but the former is a better embodiment in many cases. Based on such understanding, the technical solutions of the present invention or portions thereof contributing to the prior art may be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a FLASH Memory (FLASH), a hard disk or an optical disk of a computer, and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute the methods according to the embodiments of the present invention.
It should be noted that, in the embodiment of the rating model building apparatus, the units and modules included in the rating model building apparatus are only divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.
Embodiments of the present invention also provide a computer program product comprising computer-executable instructions for performing the method of building a rating model provided in any of the embodiments of the present invention when executed by a computer processor.
Of course, the computer program product provided in the embodiments of the present application has computer-executable instructions that are not limited to the method operations described above, and may also perform related operations in the method provided in any embodiments of the present invention.
In the description of the embodiments of the present invention, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Finally, it should be noted that the above-mentioned embodiments are only preferred embodiments of the present invention and the technical principles applied thereto. Those skilled in the art will appreciate that the present invention is not limited to the particular embodiments described herein, and that various obvious changes, rearrangements and substitutions will now be apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (12)
1. A method of building a rating model, the method comprising:
acquiring rating data of a first preset number of target objects, wherein the target objects are to-be-rated objects, and the rating data are related information of the target objects determined according to rating requirements;
sequentially grading and dividing the target object based on a second preset number of expert experience grading schemes to obtain grading results;
and taking the rating result as a rating target, and establishing a rating model by utilizing an integrated algorithm or a neural network algorithm based on the rating data.
2. The method for establishing a rating model according to claim 1, wherein when the first preset number is s and the second preset number is m, the rating division obtains n levels, and m is less than n, the ranking division is performed on the target objects in sequence based on the expert experience rating scheme of the second preset number, and obtaining the rating result comprises:
selecting any one of m expert experience rating schemes to perform rating division on the s target objects, selecting x target objects with the highest grade from the division results as a 1 st grade, and selecting x target objects with the lowest grade as an nth grade, wherein x is the number determined by multiplying s by a first preset percentage;
selecting any one of the remaining m-1 expert experience rating schemes to perform rating division on s-2x target objects, selecting y target objects with the highest grade from the division results as a 2 nd grade, and selecting y target objects with the lowest grade as an n-1 th grade, wherein y is the number determined by multiplying (s-2 x) by a second preset percentage;
repeating the grading process until the last 1 expert experience grading scheme is used for grading and dividing the rest s1 target objects, excluding z target objects with the highest grade and z target objects with the lowest grade from the grading result, and completely grading the rest target objects into the rest grades in the n grades, wherein z is the number determined by multiplying s1 by a third preset percentage.
3. The method according to claim 2, wherein when n is an odd number, after z target objects of highest rank and z target objects of lowest rank are excluded from the division results, the remaining target objects are all divided into (n + 1)/2 ranks; when n is an even number, after z target objects of the highest rank and z target objects of the lowest rank are excluded from the division result, the remaining target objects are equally divided into n/2 ranks and (n/2) +1 ranks.
4. The method of claim 1, wherein when the first predetermined number is s, the second predetermined number is m, n levels are obtained by rating division, and m > n, the obtaining of the rating result comprises:
selecting at least two of the m expert experience rating schemes to rate and divide the s target objects;
adding the scores of the two expert experience rating schemes and taking the average value to obtain an average dividing result, selecting the target objects with the fourth preset percentage from the average dividing result as the 1 st level, and selecting the target objects with the fourth preset percentage as the nth level;
selecting at least two target objects which are remained by s2 from the rest m-2 expert experience rating schemes for rating and dividing, selecting the target objects with the fifth preset percentage from the dividing results as the 2 nd level, and selecting the target objects with the fifth preset percentage from the dividing results as the n-1 st level;
repeating the grading process until the last 1 expert experience grading scheme is used for grading and dividing the rest s3 target objects, eliminating the target objects with the first sixth preset percentage and the target objects with the second sixth preset percentage from the grading result, and completely grading the rest target objects into the rest levels of the n levels.
5. The method of establishing a rating model according to claim 4, wherein when n is an odd number, after excluding a sixth preset percentage of the target objects before and a sixth preset percentage of the target objects after the target objects are excluded from the division result, all the remaining target objects are divided into (n + 1)/2 levels; when n is an even number, after the first sixth preset percentage of the target objects and the second sixth preset percentage of the target objects are excluded from the division result, the remaining target objects are equally divided into n/2 levels and (n/2) +1 levels.
6. A method for building a rating model according to claim 1, wherein the rating result is used as a rating target, and building a rating model using an integration algorithm or a neural network algorithm based on the rating data comprises:
and taking the rating result as a rating target, and establishing a rating model by utilizing a lightGBM algorithm based on the rating data.
7. A method for building a rating model according to claim 1, wherein the rating result is used as a rating target, and building a rating model using an integration algorithm or a neural network algorithm based on the rating data comprises:
and taking the rating result as a rating target, and establishing a rating model by utilizing a full-connection neural network based on the rating data.
8. A rating model building method according to claim 1, wherein after building a rating model, the building method further comprises:
and grading the target object by utilizing the grading model.
9. An apparatus for building a rating model, the apparatus comprising:
the system comprises a data acquisition unit, a data processing unit and a data processing unit, wherein the data acquisition unit is used for acquiring rating data of a first preset number of target objects, the target objects are objects to be rated, and the rating data are related information of the target objects determined according to rating requirements;
the experience rating unit is used for sequentially rating and dividing the target objects based on a second preset number of expert experience rating schemes to obtain rating results;
and the model establishing unit is used for establishing a rating model by utilizing an integration algorithm or a neural network algorithm based on the rating data by taking the rating result as a rating target.
10. A rating model creation apparatus, characterized by comprising:
one or more processors;
storage means for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of building a rating model as recited in any of claims 1-8.
11. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out a method of establishing a rating model according to any one of claims 1 to 8.
12. A computer program product comprising a computer program which, when executed by a processor, implements a method of building a rating model according to any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210806356.9A CN115204285A (en) | 2022-07-08 | 2022-07-08 | Method, device and equipment for establishing rating model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210806356.9A CN115204285A (en) | 2022-07-08 | 2022-07-08 | Method, device and equipment for establishing rating model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115204285A true CN115204285A (en) | 2022-10-18 |
Family
ID=83579870
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210806356.9A Pending CN115204285A (en) | 2022-07-08 | 2022-07-08 | Method, device and equipment for establishing rating model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115204285A (en) |
-
2022
- 2022-07-08 CN CN202210806356.9A patent/CN115204285A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110135630B (en) | Short-term load demand prediction method based on random forest regression and multi-step optimization | |
| CN109544399B (en) | Power transmission equipment state evaluation method and device based on multi-source heterogeneous data | |
| US9934470B2 (en) | Production equipment including machine learning system and assembly and test unit | |
| CN114780831A (en) | Sequence recommendation method and system based on Transformer | |
| CN115577152B (en) | Online book borrowing management system based on data analysis | |
| CN111210072A (en) | Prediction model training and user resource limit determining method and device | |
| CN116227745A (en) | Big data-based research and analysis method and system for fishing vessels | |
| CN114202174A (en) | Electricity price risk grade early warning method and device and storage medium | |
| CN109934352B (en) | Automatic evolution method of intelligent model | |
| CN112329879A (en) | Image scene graph generation method and system based on counterfactual multi-agent learning | |
| CN115204285A (en) | Method, device and equipment for establishing rating model | |
| CN116701647A (en) | Knowledge graph completion method and device based on fusion of embedded vector and transfer learning | |
| CN108363738B (en) | Recommendation method for industrial equipment data analysis algorithm | |
| CN112529637B (en) | Service demand dynamic prediction method and system based on context awareness | |
| CN114708117A (en) | Electricity safety inspection rating method, device and equipment integrating priori knowledge | |
| CN114742297A (en) | Power battery treatment method | |
| CN113987904A (en) | Method, device, equipment and storage medium for measuring and calculating repair cost of power transmission project | |
| CN115618921B (en) | Knowledge distillation method, apparatus, electronic device, and storage medium | |
| CN113435655B (en) | Sector dynamic management decision method, server and system | |
| CN108734265A (en) | Compression method and device, terminal, the storage medium of deep neural network model | |
| CN110059219A (en) | A kind of video preference prediction technique, device, equipment and readable storage medium storing program for executing | |
| CN116662922A (en) | Object type identification method and related equipment | |
| CN117350765A (en) | Variable determining method and device, storage medium and electronic equipment | |
| CN117648494A (en) | Recommendation processing method, device and computing equipment | |
| CN114758397A (en) | Model training method and device, face recognition method and device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |