CN110837895A - Model interpretation method and device, electronic equipment and computer readable storage medium - Google Patents
Model interpretation method and device, electronic equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN110837895A CN110837895A CN201911079742.7A CN201911079742A CN110837895A CN 110837895 A CN110837895 A CN 110837895A CN 201911079742 A CN201911079742 A CN 201911079742A CN 110837895 A CN110837895 A CN 110837895A
- Authority
- CN
- China
- Prior art keywords
- model
- data
- processed
- target
- processing result
- 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
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the application provides a model interpretation method and device, an electronic device and a computer readable storage medium, and relates to the technical field of model interpretation. In the embodiment of the present application, first, each target feature of a single piece of data to be processed is obtained. And secondly, processing each target characteristic according to a preset first model to obtain a processing result of a single piece of data to be processed. And then, calculating a contribution value of at least one target feature to the processing result according to a preset second model to obtain at least one contribution value, wherein the second model is an algorithm model based on a dual distribution solution of a cooperative game theory. By the method, the accuracy of model interpretation can be improved.
Description
Technical Field
The present application relates to the field of model interpretation technologies, and in particular, to a model interpretation method and apparatus, an electronic device, and a computer-readable storage medium.
Background
At present, various algorithms are applied to achieve practical effects, but generally, only one processing result is obtained for most of projects based on machine learning, and the specific reason of the processing result cannot be known. Especially in deep learning, the algorithm which is basically a black box is important to explain the algorithm.
However, the inventor researches and finds that in the prior art, the contribution of a single feature to the processing result is still difficult to give reasonable explanation, so that the accuracy of model explanation is not high.
Disclosure of Invention
In view of the above, an object of the present application is to provide a model interpretation method and apparatus, an electronic device, and a computer-readable storage medium, so as to solve the problems in the prior art.
In order to achieve the above purpose, the embodiment of the present application adopts the following technical solutions:
a model interpretation method comprising:
acquiring each target characteristic of data to be processed;
processing each target characteristic according to a preset first model to obtain a processing result;
and calculating a contribution value of at least one target characteristic to the processing result according to a preset second model to obtain at least one contribution value, wherein the second model is an algorithm model based on a dual distribution solution of a cooperative game theory.
In a preferred option of the embodiment of the present application, the step of calculating a contribution value of at least one of the target features to the processing result according to a preset second model includes:
calculating to obtain a dual distribution solution corresponding to at least one target feature according to a preset second model;
and calculating to obtain a contribution value corresponding to the target feature according to the dual distribution solution corresponding to the target feature, the processing result and a preset first formula.
In a preferred selection of the embodiment of the present application, the step of obtaining a dual assignment solution corresponding to at least one of the target features by calculation according to a preset second model includes:
carrying out relaxation treatment on the first model to obtain a second model;
and calculating to obtain a dual distribution solution corresponding to the target feature according to at least one target feature and the second model.
In a preferred option of the embodiment of the present application, the step of performing relaxation processing on the first model to obtain the second model includes:
and multiplying the Lagrange multiplier by each target characteristic included in the first model to obtain the second model.
In a preferred selection of the embodiment of the present application, the first formula includes:
wherein, ηiRepresenting the dual assignment solution corresponding to the ith target feature,
and C (N) represents the processing result.
In a preferred option of the embodiment of the present application, the step of obtaining each target feature of a single piece of data to be processed includes:
performing data cleaning processing on the acquired original data to obtain data to be processed;
and performing characteristic screening processing on the data to be processed to obtain each target characteristic.
In a preferred selection of the embodiment of the present application, the step of performing feature screening processing on the data to be processed to obtain each of the target features includes:
acquiring the characteristics of the data to be processed;
and performing characteristic screening processing on the characteristics of the data to be processed to obtain each target characteristic.
An embodiment of the present application further provides a model interpretation apparatus, including:
the target characteristic acquisition module is used for acquiring each target characteristic of a single piece of data to be processed;
the target feature processing module is used for processing each target feature according to a preset first model to obtain a processing result of a single piece of data to be processed;
and the contribution value calculating module is used for calculating a contribution value of at least one target feature to the processing result according to a preset second model to obtain at least one contribution value, wherein the second model is an algorithm model based on a dual distribution solution of a cooperative game theory.
An embodiment of the present application further provides an electronic device, which includes a memory and a processor, where the processor is configured to execute an executable computer program stored in the memory to implement the above-mentioned model interpretation method.
Embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored, which when executed, implements the steps of the above-described model interpretation method.
According to the model interpretation method and device, the electronic device and the computer readable storage medium, the contribution values of the target features of the data to be processed to the processing result of the data to be processed are respectively calculated according to the preset second model, the contribution of the target features to the processing result is obtained, and the contribution of a single feature to the processing result is interpreted, so that the accuracy of model interpretation is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.
Fig. 1 is a block diagram of an electronic device according to an embodiment of the present disclosure.
Fig. 2 is a schematic flowchart of a model interpretation method provided in an embodiment of the present application.
Fig. 3 is a schematic flowchart of step S110 according to an embodiment of the present application.
Fig. 4 is a schematic flowchart of step S130 according to an embodiment of the present application.
Fig. 5 is a block diagram of a model interpretation apparatus according to an embodiment of the present application.
Icon: 10-an electronic device; 12-a memory; 14-a processor; 100-a model interpretation means; 110-a target feature acquisition module; 120-target feature processing module; 130-contribution calculation module.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
As shown in fig. 1, an embodiment of the present application provides an electronic device 10. The electronic device 10 may include, among other things, a memory 12, a processor 14, and a model interpretation apparatus 100.
In detail, the memory 12 and the processor 14 are electrically connected directly or indirectly to enable data transmission or interaction. For example, they may be electrically connected to each other via one or more communication buses or signal lines. The model interpretation means 100 comprises at least one software functional module which may be stored in the form of software or firmware (firmware) in the memory 12. The processor 14 is used for executing executable computer programs stored in the memory 12, such as software functional modules and computer programs included in the model interpretation apparatus 100, so as to implement the model interpretation method.
The Memory 12 may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like.
The processor 14 may be an integrated circuit chip having signal processing capabilities. The Processor 14 may be a general-purpose Processor including a Central Processing Unit (CPU), a Network Processor (NP), a System on Chip (SoC), and the like.
It will be appreciated that the configuration shown in FIG. 1 is merely illustrative and that the electronic device 10 may include more or fewer components than shown in FIG. 1 or may have a different configuration than shown in FIG. 1.
With reference to fig. 2, an embodiment of the present application further provides a model interpretation method applicable to the electronic device 10. Wherein, the method steps defined by the flow related to the model interpretation method can be implemented by the electronic device 10, and the specific flow shown in fig. 2 will be described in detail below.
Step S110, each target feature of a single piece of data to be processed is obtained.
In this embodiment of the present application, the data to be processed may be obtained through user input. After obtaining the single piece of data to be processed, the electronic device 10 may process the single piece of data to be processed to obtain each target feature of the single piece of data to be processed.
And step S120, processing each target characteristic according to a preset first model to obtain a processing result of a single piece of data to be processed.
In this embodiment of the application, after obtaining each target feature of the single piece of data to be processed in step S110, each target feature may be processed according to a preset first model, so as to obtain a processing result of the single piece of data to be processed.
Step S130, calculating a contribution value of at least one target feature to the processing result according to a preset second model, to obtain at least one contribution value.
In this embodiment of the application, after the processing result is obtained in step S120, a contribution value of at least one target feature to the processing result may be calculated according to a preset second model, so as to obtain at least one contribution value.
And the second model is an algorithm model based on a dual distribution solution of the cooperative game theory.
By the method, the contribution values of the target characteristics of the data to be processed to the processing result of the data to be processed can be respectively calculated according to the preset second model, so that the contribution of the target characteristics to the processing result is obtained, the contribution of the single characteristic to the processing result is explained, and the accuracy of model explanation is improved. For step S110, it should be noted that, the specific manner of obtaining each target feature of a single piece of to-be-processed data is not limited, and may be set according to the actual application requirement.
For example, in an alternative example, in conjunction with fig. 3, step S110 may include step S111 and step S112.
And step S111, performing data cleaning processing on the acquired original data to obtain data to be processed.
Step S112, performing feature screening processing on the data to be processed to obtain each target feature.
For step S111, it should be noted that, the original data may be obtained according to the actual application requirement, and the data cleaning processing may be performed on the original data to delete the useless characters and to complement the missing data, so as to obtain the data to be processed.
For step S112, it should be noted that the specific manner of step S112 is not limited, and may be set according to the actual application requirement.
For example, in an alternative example, step S112 may include the following sub-steps:
acquiring the characteristics of the data to be processed; and performing characteristic screening processing on the characteristics of the data to be processed to obtain each target characteristic.
For example, in an alternative example, in connection with table 1, the pending data may include name, age, occupation, income, and the like of a bank customer, and the processing result of the pending data indicates the credit degree of the customer, and the bank may decide whether to approve the loan of the customer according to the credit degree, where the name, age, occupation, and income are characteristics of the pending data.
Table 1 pending data 1 (processing result is credit degree of bank client)
| Name (I) | Age (year of old) | Occupation of the world | Income (year) |
| Zhang three | 16 | Have no business | 0 |
| Li four | 32 | Teacher | 20 ten thousand |
| Wangwu tea | 27 | |
10 ten thousand |
For example, in another alternative example, in combination with table 2, the to-be-processed data may include data of age, gender, occupation, height, weight, and the like of the crowd, and the processing result of the to-be-processed data indicates a likeness of the crowd to the online game, and the age, gender, occupation, height, and weight are characteristics of the to-be-processed data.
Table 2 pending data 2 (processing result is likeness of the crowd to the network game)
| Age (year of old) | Sex | Occupation of the world | Height (cm) | Body weight (kg) |
| 16 | For male | Student's desk | 170 | 80 |
| 32 | Woman | Teacher | 165 | 50 |
| 24 | For male | Game tester | 180 | 70 |
Further, after the features of the data to be processed are obtained, preliminary feature screening may be performed on the features of the data to be processed, and a feature having a relatively large correlation with the processing result is obtained as a target feature.
For example, in an alternative example, the name and age characteristics have little effect on the level of credit, and the occupation and income characteristics have some effect on the level of credit. In detail, the credit rating of the general client is higher when the client's occupation is stable, and the credit rating of the general client is lower when the client's occupation is unstable. The credit rating of the general customer is higher when the income of the customer is higher, and the credit rating of the general customer is lower when the income of the customer is lower. Thus, occupational and income characteristics may be targeted.
For example, in another alternative example, the height and weight characteristics have little effect on the like of the online game, while the age, gender, and occupation characteristics have some effect on the like of the online game. In detail, the network game is generally preferred to be higher when the age is small, and is generally preferred to be lower when the age is large. The network game is generally preferred to be high when the sex is male, and to be low when the sex is female. The preference degree of the online game is generally high when the occupation is related to the internet, and the preference degree of the online game is generally low when the occupation is not related to the internet. Thus, age, gender and occupational characteristics may be targeted.
For step S120, it should be noted that, according to the difference between the data to be processed and the processing result, the specific type of the preset first model is not limited, and may be set according to the actual application requirement.
For example, in an alternative example, when the data to be processed is the age, gender and occupation characteristics of the crowd, in order to obtain the processing result of the likeness of the crowd to the online game, historical data including the likeness of the crowd with different ages, genders and occupation characteristics to the online game may be obtained. And constructing a two-classification model through historical data and a supervised machine learning algorithm, and classifying the people needing to be predicted by using the two-classification model to obtain the like degree of the online game.
For step S130, it should be noted that when the data to be processed is crowd information including age, gender and occupation characteristics, and the processing result obtained by model calculation is a like degree for the network game, there are two methods for explaining the model in the prior art:
1. the whole data set is considered integrally to predict which features can influence the processing result;
2. for a single datum, perhaps among all, age is the most important feature, with younger people more likely to prefer computer games. However, if a person is a 50 year old video game tester, his occupation may be much more important than his age in determining whether he likes computer games. Here identifying which features are most important to the person, i.e. finding the importance of each feature at the single data level, to arrive at an interpretation of each result predicted by the classification model.
First, a very simple model can be considered: linear regression, the output of the linear regression model is:
f(x1,x2,…,xn)＝φ1x1+φ2x2+…+φnxn；
in the linear regression model described above, for each feature xiAssigning a coefficient phiiThen all ofThe contents are added to obtain an output and the input features will be (age, gender, occupation), the influence of each feature is easily found. If phi isiHaving a very large absolute value, the feature xiHas a large influence on the final result (e.g. if phiAgeIs large, age is an important characteristic).
However, the above linear regression model has a disadvantage in that it is very simple and can represent only a linear relationship. For example, age may be an important characteristic, and computer games are generally preferred over others of age if the age is between 12 and 18. Since this is a non-linear relationship, linear regression will not be able to calculate such a result. To reveal this more complex relationship, a more complex machine learning model is required to solve this dichotomy problem.
However, once the more complex model is started, the ease of interpretation of the linear model described above is lost. In practice, it is found that there is a non-linear or even cross relationship in practical applications, for example, if the importance of the age and gender characteristics is relevant, it becomes very difficult to interpret the model. There is therefore a need to trade off between easy to interpret models (only simple relationships can be found) and complex models (relationships that may be found to be difficult to interpret), as well as between interpretability and complexity.
Ideally, a complex model that can be interpreted is obtained, and the embodiment of the present application may perform model interpretation according to the linear regression model: for each feature xiAssigning a coefficient phiiIt is possible to describe linearly how this feature affects the output of the model. Over many data sets, the coefficient φiComplex relationships cannot be obtained, but at the single data level, the prediction of each variable impact model is constant.
For example, taking Frank 50 as an example, he loves a network game, he is a video game tester 50 years old, and for him, φJobWill be very high, andAgewill be low. However, for Bobby 14 years old, φAgeWould be high because a 14 year old teenager generally tends to prefer the netAnd (6) playing online games.
With a complex model, the non-linear patterns in the data to be processed can be decomposed into many linear models describing a single piece of data. Wherein the interpretation coefficient phiiIt is not the result of the processing of the model, but is used to interpret the contents of the model. Rather than attempting to interpret the entire complex model, it is possible to do so by using a linear interpretation model, rather than attempting to interpret the results of the complex model's processing on a single piece of data. To further simplify the model, the coefficient φ is used if the target feature is presentiMultiply by 1, otherwise multiply by 0.
In the example of predicting who likes to play the online game, the following formula is obtained:
wherein g isFrank＝pFrank，pFrankIs the original prediction of the Frank model.
Wherein the coefficients included in the above formula are applicable only to Frank. If one wants to find the behavior of the model to Bobby, one needs to find a new set of coefficients. Since Bobby does not work, Bobby can be expected to work
Multiplied by 0. Thus, a simple model for Bobby would be:
at the same time, this will be done for all data points and aggregated, and it can be understood how the model processes all data. Further, the question of how to interpret a complex model needs to be considered is how to derive the weight values
The method of cooperative game theory can be adopted for solving
The value is obtained.
In practice, when a group of people is playing the game, a certain reward may be obtained by playing the game, and how to reflect their respective contributions to allocate the reward may represent an interpretation of the model. As can be appreciated from cooperative gaming theory, satisfying the following conditions will mean that cooperation is fair and stable:
1. the sum of the rewards each person receives should equal the total reward;
2. if two people contribute the same value, they should get the same amount from the reward;
3. people who do not contribute have no rewards;
4. if the team plays two games, the individual's prize from the two games should equal the prize won in the first game plus the prize won in the second game.
The above condition is a fairly intuitive rule that must be followed when distributing rewards, which translates well into the machine learning problem we are trying to solve. In a machine learning problem, the reward may represent the final prediction of a complex model, while the participants in the game may represent features, these rules translate into symbols in the instance:
1.gFrankshould be reacted with pFrankEqual, pFrankIs the prediction probability of the model to Frank who likes online games;
2. if the two features x contribute the same to the final predicted value, their coefficients φ should have the same value;
3. if a feature does not have any contribution (or lack) to the final predicted value, its contribution to g should be 0;
4.gFrank+Bobby＝gFrank+gBobby。
optionally, the specific way of calculating the contribution value is not limited, and may be set according to the actual application requirement. For example, in an alternative example, in conjunction with fig. 4, step S130 may include step S131 and step S132.
Step S131, calculating according to a preset second model to obtain a dual distribution solution corresponding to at least one target feature.
And step S132, calculating to obtain a contribution value corresponding to the target feature according to the dual distribution solution corresponding to the target feature, the processing result and a preset first formula.
In detail, the specific manner of step S131 is not limited, and may be set according to the actual application requirement. For example, in an alternative example, step S131 may include the following sub-steps:
firstly, the first model can be subjected to relaxation treatment to obtain the second model; secondly, a dual distribution solution corresponding to the target feature can be calculated according to at least one target feature and the second model.
Wherein the step of performing relaxation processing on the first model may include: and multiplying the Lagrange multiplier with each target characteristic included in the first model to obtain the second model, adding one Lagrange multiplier to each target characteristic to obtain the second model belonging to a Lagrange function, and fusing the target characteristics into the first model through the Lagrange function to facilitate the explanation of the first model.
For example, in an alternative example, where the first model may represent an inequality constrained optimization problem:
minxf(x)；
s.t.mi(x)≤0i＝1,2,...…m；
said first model can be transformed into a first function of lagrange form:
the constraint function of the first function is:
minxmaxλ,ηL(x,λ,η)；
further, the dual problem of the constraint function may be solved by constructing a second model:
maxλ,ηminxL(x,λ,η)；
wherein x represents the target feature, λ represents the lagrangian multiplier corresponding to the target feature, and η represents the dual distribution solution corresponding to the target feature.
For step S132, it should be noted that the specific form of the first formula is not limited, and may be set according to the actual application requirement. For example, in an alternative example, the first formula may include:
wherein, ηiRepresenting the dual assignment solution corresponding to the ith target feature,
and C (N) represents the processing result.
With reference to fig. 5, an embodiment of the present invention further provides a model interpretation apparatus 100, which can be applied to the electronic device 10. The model interpretation apparatus 100 may include a target feature acquisition module 110, a target feature processing module 120, and a contribution value calculation module 130, among others.
The target feature obtaining module 110 is configured to obtain each target feature of a single piece of data to be processed. In this embodiment, the target feature acquiring module 110 may be configured to execute step S110 shown in fig. 2, and for the relevant content of the target feature acquiring module 110, reference may be made to the foregoing detailed description of step S110.
The target feature processing module 120 is configured to process each target feature according to a preset first model to obtain a processing result of a single piece of data to be processed. In this embodiment, the target feature processing module 120 may be configured to execute step S120 shown in fig. 2, and reference may be made to the foregoing detailed description of step S120 for relevant contents of the target feature processing module 120.
The contribution value calculating module 130 is configured to calculate a contribution value of at least one target feature to the processing result according to a preset second model to obtain at least one contribution value, where the second model is an algorithm model based on a dual distribution solution of a cooperative game theory. In this embodiment, the contribution value calculating module 130 may be configured to execute step S130 shown in fig. 2, and for the relevant content of the contribution value calculating module 130, reference may be made to the foregoing detailed description of step S130.
In an embodiment of the present application, there is also provided a computer-readable storage medium, in which a computer program is stored, and the computer program executes the steps of the model interpretation method when the computer program runs.
The steps executed when the computer program runs are not described in detail herein, and reference may be made to the explanation of the model interpretation method above.
In summary, according to the model interpretation method and apparatus, the electronic device, and the computer-readable storage medium provided in the embodiments of the present application, the contribution values of the target features of the to-be-processed data to the processing result of the to-be-processed data are respectively calculated according to the preset second model, so as to obtain the contribution of the target features to the processing result, and to explain the contribution of a single feature to the processing result, thereby improving the accuracy of model interpretation.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (10)
1. A model interpretation method, comprising:
acquiring each target characteristic of a single piece of data to be processed;
processing each target characteristic according to a preset first model to obtain a processing result of a single piece of data to be processed;
and calculating a contribution value of at least one target characteristic to the processing result according to a preset second model to obtain at least one contribution value, wherein the second model is an algorithm model based on a dual distribution solution of a cooperative game theory.
2. The model interpretation method of claim 1, wherein said step of calculating a contribution value of at least one of said target features to said processing result according to a preset second model comprises:
calculating to obtain a dual distribution solution corresponding to at least one target feature according to a preset second model;
and calculating to obtain a contribution value corresponding to the target feature according to the dual distribution solution corresponding to the target feature, the processing result and a preset first formula.
3. The model interpretation method according to claim 2, wherein said step of calculating a solution of dual assignment corresponding to at least one of said target features according to a predetermined second model comprises:
carrying out relaxation treatment on the first model to obtain a second model;
and calculating to obtain a dual distribution solution corresponding to the target feature according to at least one target feature and the second model.
4. The model interpretation method of claim 3, wherein said step of relaxing said first model to obtain said second model comprises:
and multiplying the Lagrange multiplier by each target characteristic included in the first model to obtain the second model.
5. The model interpretation method of claim 2, wherein the first formula comprises:
wherein, ηiRepresenting the dual assignment solution corresponding to the ith target feature,
and C (N) represents the processing result.
6. The model interpretation method of claim 1, wherein said step of obtaining target features for a single piece of data to be processed comprises:
performing data cleaning processing on the acquired original data to obtain data to be processed;
and performing characteristic screening processing on the data to be processed to obtain each target characteristic.
7. The model interpretation method of claim 6, wherein said step of performing feature screening processing on said data to be processed to obtain each of said target features comprises:
acquiring the characteristics of the data to be processed;
and performing characteristic screening processing on the characteristics of the data to be processed to obtain each target characteristic.
8. A model interpretation apparatus, comprising:
the target characteristic acquisition module is used for acquiring each target characteristic of a single piece of data to be processed;
the target feature processing module is used for processing each target feature according to a preset first model to obtain a processing result of a single piece of data to be processed;
and the contribution value calculating module is used for calculating a contribution value of at least one target feature to the processing result according to a preset second model to obtain at least one contribution value, wherein the second model is an algorithm model based on a dual distribution solution of a cooperative game theory.
9. An electronic device comprising a memory and a processor for executing an executable computer program stored in the memory to implement the model interpretation method of any of claims 1 to 7.
10. A computer-readable storage medium, having stored thereon a computer program which, when executed, carries out the steps of the model interpretation method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911079742.7A CN110837895A (en) | 2019-11-07 | 2019-11-07 | Model interpretation method and device, electronic equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911079742.7A CN110837895A (en) | 2019-11-07 | 2019-11-07 | Model interpretation method and device, electronic equipment and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110837895A true CN110837895A (en) | 2020-02-25 |
Family
ID=69576224
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911079742.7A Pending CN110837895A (en) | 2019-11-07 | 2019-11-07 | Model interpretation method and device, electronic equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110837895A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112116028A (en) * | 2020-09-29 | 2020-12-22 | 联想(北京)有限公司 | Model decision interpretation implementation method and device and computer equipment |
| CN113570260A (en) * | 2021-07-30 | 2021-10-29 | 北京房江湖科技有限公司 | Task allocation method, computer-readable storage medium and electronic device |
-
2019
- 2019-11-07 CN CN201911079742.7A patent/CN110837895A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112116028A (en) * | 2020-09-29 | 2020-12-22 | 联想(北京)有限公司 | Model decision interpretation implementation method and device and computer equipment |
| CN112116028B (en) * | 2020-09-29 | 2024-04-26 | 联想(北京)有限公司 | Model decision interpretation realization method and device and computer equipment |
| CN113570260A (en) * | 2021-07-30 | 2021-10-29 | 北京房江湖科技有限公司 | Task allocation method, computer-readable storage medium and electronic device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kremer et al. | Sparse portfolio selection via the sorted ℓ1-norm | |
| Middleton et al. | Unbiased estimation of the average treatment effect in cluster-randomized experiments | |
| Hill | Bayesian nonparametric modeling for causal inference | |
| Lim et al. | Bounded rationality and group size in Tullock contests: Experimental evidence | |
| Formby et al. | Mobility measurement, transition matrices and statistical inference | |
| Chowdhury et al. | Overbidding and overspreading in rent-seeking experiments: Cost structure and prize allocation rules | |
| Hahn | Decision making with uncertain judgments: A stochastic formulation of the analytic hierarchy process | |
| Pitt et al. | Credit programs for the poor and the health status of children in rural Bangladesh | |
| Lee et al. | Binary choice models with social network under heterogeneous rational expectations | |
| Ramos-Guajardo et al. | K-sample tests for equality of variances of random fuzzy sets | |
| Chakrabartty et al. | Scoring and analysis of likert scale: Few approaches | |
| Bornn et al. | Moment conditions and Bayesian non-parametrics | |
| CN106485585A (en) | Method and system for ranking | |
| CN110837895A (en) | Model interpretation method and device, electronic equipment and computer readable storage medium | |
| Okumura et al. | Concave‐monotone treatment response and monotone treatment selection: With an application to the returns to schooling | |
| Stahl | Is step-j thinking an arbitrary modelling restriction or a fact of human nature? | |
| Poleto et al. | A product-multinomial framework for categorical data analysis with missing responses | |
| Rossellini et al. | Integrating uncertainty awareness into conformalized quantile regression | |
| Conte et al. | A Test of the Rational Expectations Hypothesis using data from a Natural Experiment | |
| Hartinger | Do generalized expected utility theories capture persisting properties of individual decision makers? | |
| Wigren et al. | Marketing Mix Modelling: A comparative study of statistical models | |
| Hasebe | Treatment effect estimators for count data models | |
| Fan et al. | Estimation and inference in an ecological inference model | |
| Nicolau et al. | Measuring wage inequality under right censoring | |
| Marques et al. | Performance of combined models in discrete binary classification |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200225 |