CN116205536A - Object evaluation method, computing device, and readable storage medium - Google Patents
Object evaluation method, computing device, and readable storage medium Download PDFInfo
- Publication number
- CN116205536A CN116205536A CN202310231352.7A CN202310231352A CN116205536A CN 116205536 A CN116205536 A CN 116205536A CN 202310231352 A CN202310231352 A CN 202310231352A CN 116205536 A CN116205536 A CN 116205536A
- Authority
- CN
- China
- Prior art keywords
- user
- particles
- target
- coordinates
- particle
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
Embodiments of the present specification provide an object evaluation method, a computing device, and a readable storage medium, wherein the object evaluation method includes: providing first user data, mapping first user particles corresponding to the first user into a preset coordinate space according to target user attributes and first attribute values, determining the matching degree between the attribute values corresponding to the current coordinates of the first user particles and target objects as index scores of the first user particles, iteratively updating the coordinates of the first user particles by taking the change trend of the index scores as a target, and determining the user deviation degree of an object to be evaluated according to the difference value of the index scores among the iteratively updated user particles. The user particles are mapped in the coordinate space of any target user attribute, so that the method has high adaptability, and the user deviation degree is determined according to the difference value of index scores among the user particle coordinates after the self-iteration updating, so that the accuracy is ensured, and the processing efficiency is improved.
Description
Technical Field
The embodiment of the specification relates to the technical field of data processing, in particular to an object evaluation method.
Background
With the development of computer technology, the interactive data between the user data and the target object is deeply analyzed to obtain the matching degree between the user and the target object, and the target object recommendation by the target user is determined based on the matching degree, so that the method is widely applied to various fields such as commodity recommendation of an e-commerce platform, information recommendation of an information recommendation platform, personalized information search of the information search platform, course matching of an education platform and the like. Object recommendation is performed by determining target users based only on the matching degree, because of collection limitation of interaction data and setting bias of a matching degree analysis algorithm, resulting in a high degree of deviation between the determined target users, for example, for occupation of "air space", the determined target users are mostly female users, and for the occupation, there is a high user deviation.
Currently, the evaluation of the user deviation degree of the object to be evaluated mainly depends on a statistical index calculation method, such as accuracy, recall, F1-Score (an index used for measuring the accuracy of two classification models in statistics) and the like, so as to perform statistical calculation and obtain the user deviation degree of the object to be evaluated. However, in order to ensure accuracy of the evaluation result, a statistical analysis needs to be performed on a large amount of data, for example, the indexes of all user data are gradually calculated and analyzed by using a violent exhaustion method, so as to determine the user deviation degree, which results in insufficient evaluation efficiency, and the corresponding evaluation result is mainly a binary result, which is different from the actual diversified object recommendation result, so that an efficient and highly adaptive object evaluation method is needed to determine the user deviation degree of the object to be evaluated.
Disclosure of Invention
In view of this, the present embodiment provides an object evaluation method. One or more embodiments of the present disclosure relate to a method for evaluating a commodity recommendation system, a method for evaluating an information recommendation system, an object evaluating apparatus, an evaluating apparatus of a commodity recommendation system, an evaluating apparatus of an information recommendation system, a computing device, a computer-readable storage medium, and a computer program, to solve the technical drawbacks existing in the prior art.
According to a first aspect of embodiments of the present specification, there is provided an object assessment method, including:
providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target object as the index score of the first user particles;
iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target;
And determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the user particles after iterative updating.
According to a second aspect of embodiments of the present specification, there is provided an evaluation method of a commodity recommendation system, including:
providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target commodity as the index score of the first user particles;
iteratively updating the coordinates of the first user particles with the change trend of the index score as a target
And determining the user deviation degree of the commodity recommendation system according to the difference value of the index scores among the iteratively updated user particles.
According to a third aspect of embodiments of the present specification, there is provided an evaluation method of an information recommendation system, including:
providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
According to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target recommendation information as index scores of the first user particles;
iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target;
and determining the user deviation degree of the information recommendation system according to the difference value of the index scores among the iteratively updated user particles.
According to a fourth aspect of embodiments of the present specification, there is provided an object evaluation apparatus comprising:
a first providing module configured to provide first user data, wherein the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute;
the first mapping module is configured to map first user particles corresponding to the first user data into a preset coordinate space according to the target user attribute and the first attribute value;
the first index score determining module is configured to determine the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target object as the index score of the first user particles;
The first updating module is configured to iteratively update the coordinates of the first user particles by taking the change trend of the index score as a target;
the first deviation determining module is configured to determine the user deviation of the object to be evaluated according to the iteratively updated difference value of the index scores among the user particles.
According to a fifth aspect of embodiments of the present specification, there is provided an evaluation device of a commodity recommendation system, including:
a second providing module configured to provide first user data, wherein the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute;
the second mapping module is configured to map first user particles corresponding to the first user data into a preset coordinate space according to the target user attribute and the first attribute value;
the second index score determining module is configured to determine the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target commodity as the index score of the first user particles;
a second updating module configured to iteratively update the coordinates of the first user particles with respect to the trend of the index score
And the second deviation degree determining module is configured to determine the user deviation degree of the commodity recommendation system according to the iteratively updated difference value of the index scores among the user particles.
According to a sixth aspect of the embodiments of the present specification, there is provided an evaluation device of an information recommendation system, including:
a third providing module configured to provide first user data, wherein the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute;
the third mapping module is configured to map the first user particles corresponding to the first user data into a preset coordinate space according to the target user attribute and the first attribute value;
the third index score determining module is configured to determine the matching degree between the attribute value corresponding to the current coordinate of the first user particle and the target recommendation information as the index score of the first user particle;
the third updating module is configured to iteratively update the coordinates of the first user particles by taking the change trend of the index score as a target;
and the third deviation determining module is configured to determine the user deviation of the information recommendation system according to the iteratively updated difference value of the index scores among the user particles.
According to a seventh aspect of embodiments of the present specification, there is provided a computing device comprising:
a memory and a processor;
the memory is configured to store computer-executable instructions that, when executed by the processor, implement the steps of the object assessment method, the merchandise recommendation system assessment method, or the information recommendation system assessment method described above.
According to an eighth aspect of the embodiments of the present specification, there is provided a computer-readable storage medium storing computer-executable instructions which, when executed by a processor, implement the above-described object evaluation method, commodity recommendation system evaluation method, or information recommendation system evaluation method steps.
According to a ninth aspect of the embodiments of the present specification, there is provided a computer program, wherein the computer program, when executed in a computer, causes the computer to execute the steps of the above-described object evaluation method, the commodity recommendation system evaluation method, or the information recommendation system evaluation method.
In one or more embodiments of the present disclosure, first user data is provided, where the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute, and according to the target user attribute and the first attribute value, first user particles corresponding to the first user are mapped into a preset coordinate space, and a matching degree between the attribute value corresponding to a current coordinate of the first user particles and a target object is determined as an index score of the first user particles; and iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target, and determining the user deviation degree of the object to be evaluated according to the difference value of the index score among the user particles after the iterative updating. The user deviation degree of the object to be evaluated can be determined by only carrying out difference value calculation on the index scores of the user particles after the iterative updating is completed, so that the user deviation degree can be determined after the index of all user data is gradually calculated and analyzed, and the evaluation efficiency is improved while the accuracy of the user deviation degree is ensured.
Drawings
FIG. 1 is a flow chart of a method of object assessment provided in one embodiment of the present disclosure;
FIG. 2 is a flow chart of a method for evaluating a merchandise recommendation system according to one embodiment of the present disclosure;
FIG. 3 is a flowchart of an evaluation method of an information recommendation system according to an embodiment of the present disclosure;
FIG. 4 is a process flow diagram of an object assessment method according to one embodiment of the present disclosure;
FIG. 5 is a flow chart of an object assessment method according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram of an object assessment apparatus according to an embodiment of the present disclosure;
FIG. 7 is a schematic diagram of an evaluation device of a commodity recommendation system according to an embodiment of the present disclosure;
FIG. 8 is a schematic diagram of an evaluation device of an information recommendation system according to an embodiment of the present disclosure;
FIG. 9 is a block diagram of a computing device provided in one embodiment of the present description.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present description. This description may be embodied in many other forms than described herein and similarly generalized by those skilled in the art to whom this disclosure pertains without departing from the spirit of the disclosure and, therefore, this disclosure is not limited by the specific implementations disclosed below.
The terminology used in the one or more embodiments of the specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of the specification. As used in this specification, one or more embodiments and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present specification refers to and encompasses any or all possible combinations of one or more of the associated listed items.
It should be understood that, although the terms first, second, etc. may be used in one or more embodiments of this specification to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first may also be referred to as a second, and similarly, a second may also be referred to as a first, without departing from the scope of one or more embodiments of the present description. The word "if" as used herein may be interpreted as "at … …" or "at … …" or "responsive to a determination", depending on the context.
Furthermore, it should be noted that, user information (including, but not limited to, user equipment information, user personal information, etc.) and data (including, but not limited to, data for analysis, stored data, presented data, etc.) according to one or more embodiments of the present disclosure are information and data authorized by a user or sufficiently authorized by each party, and the collection, use, and processing of relevant data is required to comply with relevant laws and regulations and standards of relevant countries and regions, and is provided with corresponding operation entries for the user to select authorization or denial.
First, terms related to one or more embodiments of the present specification will be explained.
Object recommendation algorithm: and determining the matching degree between the user and the object based on the user attribute in the user data and the object attribute of the object, and screening and filtering a large number of objects to obtain an algorithm of a target object with higher matching degree.
Object recommendation model: an object recommendation algorithm is characterized in that an object recommendation model models and mines preference information of users based on interaction data between the users and objects, and a large number of objects are screened and filtered based on the matching degree between the users and the objects to obtain an artificial intelligent model of a target object with higher matching degree.
User bias degree: in the process of determining a target object, the artificial intelligent model algorithm makes object screening and filtering independent of sensitive information of natural attributes and social attributes, so that the sensitive information is protected, and the inherent user attributes or acquired user attributes of individual users or group users are not existed, so that prejudice and deviation of object recommendation are caused.
User bias evaluation: and under the evaluation index of the deviation degree of the specific user, based on the input and output results of the object recommendation model, evaluating whether the object recommendation model has deviation and deviation to the specific individual user or group user, and further evaluating the deviation and deviation degree.
In the present specification, there are provided an object evaluation method, the present specification relates to an evaluation method of a commodity recommendation system, an evaluation method of an information recommendation system, an object evaluation apparatus, an evaluation apparatus of a commodity recommendation system, an evaluation apparatus of an information recommendation system, a computing device, a computer-readable storage medium, and a computer program, which are described in detail one by one in the following embodiments.
Referring to fig. 1, fig. 1 shows a flowchart of an object evaluation method according to an embodiment of the present disclosure, including the following specific steps:
Step 102: first user data is provided, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute.
The embodiment of the present disclosure is applied to a server side or a cloud side device of an application for evaluating a user deviation degree of an object to be evaluated, where the cloud side device is a distributed virtual network device, and performs data transmission with the end side device through a protocol interface, which is not limited herein. In the embodiment of the present disclosure, the object to be evaluated is any artificial intelligence model. Artificial intelligence models corresponding to different application fields, such as a translation model, a commodity recommendation model, an information search model, a course screening model, and the like.
The user data is user data corresponding to a user on the object to be evaluated, the user data comprises at least one user attribute of the user and an attribute value corresponding to the user attribute, for example, the user data StatUser_1 comprises four user attributes of a region, an academic, a labor value and an interest and attribute values corresponding to the four user attributes: region-A, academy-family, labor value-20K, interest-skiing. The user data may be obtained by online direct collection, or may be provided by a user database of the object to be evaluated, and it should be noted that, in order to ensure accuracy and efficiency of the subsequent user deviation degree, the collection of the user data may be implemented according to a preset rule, for example, collection according to a preset attribute value distribution, collection according to a preset number threshold of the user data, and so on.
The target user attribute is at least one user attribute category corresponding to the user deviation degree of the object to be evaluated, the target user attribute is any one or more of the user attributes of the user data, corresponding replacement is carried out according to different evaluation dimensions of the user deviation degree of the object to be evaluated, and the attribute value corresponding to the target user attribute is a quantized numerical value corresponding to the target user attribute. The target user attribute can be regarded as an evaluation dimension of a user deviation degree, the attribute value corresponding to the target user attribute can be regarded as an evaluation reference value input in the evaluation dimension, and the user deviation degree can be regarded as an evaluation index value output in the evaluation dimension. For example, the object to be evaluated is a commodity recommendation model, the target object is a brand electronic product, the target user attribute is occupation, labor value, interest and the like, and the user deviation degree of the object to be evaluated is determined according to the attribute value of the target user attribute, namely, whether the commodity recommendation model has deviation and deviation to the user attribute such as occupation, labor value, interest and the like in the brand electronic product recommendation is determined.
The first user data is any one of user data of a plurality of users on the object to be evaluated, and the first user data comprises a target user attribute and an attribute value corresponding to the target user attribute, namely a first attribute value.
For example, 100 user data statsuser_i (i e [1,100 ]) of the object to be evaluated obtained by online collection are provided, any one first user data statsuser_1 is determined from the 100 user data statsuser_i, and the first user data statsuser_1 includes a target user attribute TargetUserProp and a first attribute value PropValue_1 corresponding to the target user attribute TargetUserProp.
And providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute, and providing data support for mapping the first user particles to a preset coordinate space subsequently.
Step 104: and according to the target user attribute and the first attribute value, mapping the first user particles corresponding to the first user data into a preset coordinate space.
The preset coordinate space is pre-constructed according to the target user attribute, is an abstract coordinate space taking the target user attribute as a coordinate dimension, and can be a coordinate space only comprising the coordinate dimension corresponding to the target user attribute, or can be a coordinate space comprising the coordinate dimension corresponding to the target user attribute and the coordinate dimension corresponding to other user attributes. Any one of the coordinates in the preset coordinate space corresponds to a specific attribute value, for example, the preset coordinate space is a coordinate dimension corresponding to the attribute of the regional target user in terms of height, the first user is a user of the area a of 180cm, the corresponding first attribute value is 180,1 (a ground), and the coordinates in the preset coordinate space are (180,1).
The first user particles are mass-free abstract particles corresponding to the first user data, and after the first user particles are mapped to a preset coordinate space, the first user particles can move according to a preset movement rule, namely the coordinates of the first user particles are updated.
And mapping the first user particles corresponding to the first user into a preset coordinate space according to the target user attribute and the first attribute value, wherein the specific mode is that corresponding coordinates in the preset coordinate space are determined according to the target user attribute and the first attribute value, and the coordinates are determined to be initial coordinates of the first user particles corresponding to the first user.
The preset coordinate space TargetSpace is a coordinate space with a target user attribute TargetUserProp as a coordinate dimension, corresponding coordinates (X1, Y1) in the preset coordinate space TargetSpace are determined according to the target user attribute TargetUserProp and the first attribute value propvalue_1, and the coordinates (X1, Y1) are determined as initial coordinates of the first user particle pad_1 corresponding to the first user data.
According to the target user attribute and the first attribute value, mapping the first user particle corresponding to the first user data into a preset coordinate space, laying a data base for the follow-up determination of the matching degree and the iterative updating of the coordinates of the first user particle, constructing the preset coordinate space for any target user attribute, and mapping the user particle corresponding to the user data under the coordinate space, wherein the user particle has high adaptability.
Step 106: and determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target object as the index score of the first user particles.
The current coordinates of the first user particles are coordinates of the current position of the first user particles in a preset coordinate space.
The matching degree is the degree of matching quantification between the attribute value and the target object determined by using an object recommendation algorithm of the object to be evaluated, and comprises, but is not limited to, recall rate, accuracy, click conversion rate, selection conversion rate, diversity, popularity, novelty and the like. For example, the attribute value of the user is academic-family, occupation-teacher, and the object is a mathematical teaching material, so that the two materials have high matching degree. The index score is a quantization score for evaluating the representativeness of the user particles, and the higher or lower index score represents the representativeness of the user particles, for example, for M user particles with adjacent coordinates, the user particles with the highest or lowest index score can be used for representing the M user particles, so that the data processing capacity can be reduced, and the evaluation efficiency can be improved.
The matching degree between the attribute value corresponding to the current coordinate of the first user particle and the target object is determined as the index score of the first user particle, and the matching degree is equivalent to the coordinate value of the current coordinate. Similarly, for a screen with 1980×1080 resolution, any pixel has corresponding coordinates, the coordinates of the pixel are unchanged, and the chromaticity value of the pixel is the coordinate value of the pixel. After the first user particles are mapped to a preset coordinate space, the user particles move, and coordinate values of all coordinates on a moving path in the space are correspondingly determined one by one, which is equivalent to the detection process of the coordinate values of all coordinates in the preset coordinate space by using the user particles.
And determining the matching degree between the attribute value corresponding to the current coordinate of the first user particle and the target object as the index score of the first user particle, wherein the matching degree between the attribute value and the target object is calculated by using an object recommendation algorithm of the object to be evaluated according to the attribute value corresponding to the current coordinate of the first user particle and the object data of the target object, and the matching degree is determined as the index score of the first user particle. The object data of the target object is the object data of the target object on the object to be evaluated, and the object data comprises at least one object attribute of the target object and an attribute value corresponding to the object attribute.
For example, according to the attribute value PropValue_1 corresponding to the current coordinate (X1, Y1) of the first user particle Patcle_1 and the object data StatTargetItem of the target object, using the object recommendation model of the object to be evaluated, calculating to obtain the matching degree Fit between the attribute value PropValue_1 and the target object, and determining the matching degree Fit as the index score of the first user particle Patcle_1.
And determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target object as the index score of the first user particles, and providing a numerical basis for iteratively updating the coordinates.
Step 108: and iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target.
Because the user particles are directly mapped into the preset coordinate space according to the target user attribute and attribute value in step 104, the distribution is more scattered and irregular, and the coordinates of the user particles need to be iteratively updated by taking the change trend of the index score as a target, so that the iteratively updated user particles meet the relatively concentrated and regular target coordinate distribution, and the user deviation degree is further determined according to the target coordinate distribution, so that the accurate user deviation degree is obtained, and the processing efficiency is improved.
The change trend of the index score is a particle motion rule preset to meet the change trend of the index score, the specific form may be a change trend equation of the index score, or may be monotonicity judgment of the change trend of the index score, for example, the index score meets the change trend of the unitary quadratic function equation form, and the index score meets the change trend of monotonic change (monotonically increasing or single-point decreasing). And taking the change trend of the index score as a target, and enabling the user particles subjected to iterative updating to meet the target coordinate distribution.
The iterative updating is to update the coordinates of the user particles according to the iteration speed, wherein the preset iteration speed is a vector for updating the user particles in one update, and the vector comprises a coordinate length and a coordinate direction, for example, the preset iteration speed is V, the coordinate length of the iteration speed is L, the coordinate direction is alpha, the current coordinates of the user particles are (X1, Y1), the coordinates of the updated user particles are (X2, Y2), then
tan(Y2-Y1)/(X2-X1)＝α。
And iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target, specifically, iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target according to the iteration speed.
Illustratively, the preset trend equation is satisfied with the index score: y=f (X) as a target, and iteratively updating the coordinates (Xi, yi) of the first user particle paycle_1 according to the iteration speed V.
And the change trend of the index score is taken as a target, the coordinates of the first user particles are iteratively updated, so that the user deviation degree is ensured to be determined according to the difference value of the index scores among the user particles which are more aggregated and regularly distributed, the accurate user deviation degree is ensured to be obtained, and the processing efficiency is improved.
Step 110: and determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the user particles after iterative updating.
The ending judgment condition for finishing the iterative updating can be preset iterative times or preset coordinate distribution conditions of user particles, for example, any user particle is in the coordinate radius R and contains at least one other user particle, for example, any two or more user particles are not in the coordinate radius R, so that the regularity of particle distribution is ensured. And are not limited thereto.
In an object recommendation algorithm of an object to be evaluated, the user deviation degree is an object recommendation difference value of a target user attribute under a target object. For example, if in a commodity recommendation system, for a certain brand of electronic product, there is a higher probability of object recommendation for users with higher labor value, and a lower probability of object recommendation for users with lower labor value, then there is a high difference value in object recommendation under the brand of electronic product for the labor value, the commodity recommendation algorithm of the commodity recommendation system has a higher user deviation degree, and the commodity recommendation algorithm of the commodity recommendation system can be adjusted according to the user deviation degree.
Determining the user deviation degree of the object to be evaluated according to the difference value of the index scores of the user particles after the iterative updating, specifically, marking the categories of the user particles according to the index scores of the user particles after the iterative updating, and determining the user deviation degree of the target user attribute under the target object according to the difference value of the index scores of the user particles under the categories. Further, one or more user particles in each category are determined to be category user particles representing the category, the index score of the category user particles is determined to be the index score of the category, and the user deviation degree of the target user attribute under the target object is determined according to the difference value of the index scores among the categories. Wherein the class user particles are representative user particles in the class, the class user particles can determine one or more user particles with the highest index score or the lowest index score in each class, or one or more user particles with the most concentrated coordinate distribution in each class,
And virtual user particles obtained by carrying out statistical calculation according to user particles in each category can be used, and the data processing capacity is reduced and the evaluation efficiency is improved by determining the category user particles. For example, after the coordinates of each user particle are iteratively updated, determining one or more user particles with the most concentrated coordinate distribution in each category, and determining that the index score of the user particle is the index score of the category for each representative user particle, for example, the index score of category 1 is 0.94, the index score of category 2 is 0.13, the difference value between the two is 0.81 and is greater than a preset difference value threshold (0.3), and determining that the user deviation degree of the object to be evaluated is high, thereby adjusting the object to be evaluated.
Illustratively, category labeling is performed on each user particle Pat_i according to the index score of each user particle Pat_i (i e [1,100 ]) after iterative updating, and the user deviation degree Fai rness of the object to be evaluated is determined according to the difference value of the index scores among the user particles under each category.
In the embodiment of the present disclosure, first user data is provided, where the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute, and according to the target user attribute and the first attribute value, first user particles corresponding to the first user data are mapped into a preset coordinate space, and a matching degree between the attribute value corresponding to a current coordinate of the first user particles and a target object is determined as an index score of the first user particles; and iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target, and determining the user deviation degree of the object to be evaluated according to the difference value of the index score among the user particles after the iterative updating. The user deviation degree of the object to be evaluated can be determined by only carrying out difference value calculation on the index scores of the user particles after the iterative updating is completed, so that the user deviation degree can be determined after the index of all user data is gradually calculated and analyzed, and the evaluation efficiency is improved while the accuracy of the user deviation degree is ensured.
Optionally, after step 104, the following specific steps are further included:
according to the first attribute value, determining an initial user particle swarm corresponding to the first user data, wherein the initial user particle swarm is at least two preset user particle swarms;
correspondingly, the step 110 includes the following specific steps:
determining a target user particle group corresponding to each user particle according to the coordinates of each user particle after iterative updating;
and determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the target user particle swarms.
The initial user particle swarm is a preset user particle classification group before iterative updating, and is divided according to the target user attribute. The target user particle swarm is divided into a target user particle swarm and an initial user particle swarm according to the target user attribute, the target user particle swarm and the initial user particle swarm are consistent, and the contained user particles are changed in the coordinate iterative updating process. In the process of the iterative updating of the coordinates of the user particles, the initial user particle swarm may be changed to a coordinate position closer to other user particle swarms, and at the moment, the user particle swarm corresponding to the user particles may be changed, that is, the initial user particle swarm is changed to the target user particle swarm. For example, according to the target user attribute of the occupation, the initial user particle groups of "teacher", "doctor" and "lawyer" are preset, and in the coordinate iterative update process, the user particles included in each user particle group may be changed, and the target user particle groups are also "teacher", "doctor" and "lawyer", but the included user particles may be different.
At least two initial user particle groups are preset, user particles mapped in a preset coordinate space are divided into corresponding user particle groups, at the moment, the distribution of the user particles in the user particle groups is scattered and irregular, after the coordinates of the user particles need to be updated in an iterative manner, the distribution of the user particles in each user particle group is gathered and regular, the difference value of index scores among the user particle groups can be directly determined, statistical analysis on a large amount of data is avoided, for example, the index score of at least one user particle in each user particle group is selected as the index score of the user particle group, the user particles can be randomly sampled in a random manner, the user particle at the center can be selected according to the coordinate distribution of each user particle, and the user particle with the highest or lowest index score can be selected without limitation.
And determining an initial user particle swarm corresponding to the first user data according to the first attribute value, wherein the initial user particle swarm is at least two preset user particle swarms. For example, the first user data corresponds to the consumers born in 1997, and is divided into the initial population of user particles "after 90" according to the year of birth "1997" and the year of birth threshold value of the initial population of user particles.
And determining a target user particle group corresponding to each user particle according to the coordinates of each user particle after iterative updating, specifically, determining the target user particle group corresponding to each user particle according to the coordinate distribution of each user particle after iterative updating. Optionally, determining the target user particle group corresponding to each user particle according to the distance between the coordinates of each user particle after iterative updating. For example, user particles included in a coordinate range where any user particle radius R is set belong to the same target user particle group.
And determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the target user particle groups, wherein one or more user particles in each target user particle group are determined to be representative user particles representing the target user particle group, the index score representing the user particle is determined to be the index score of the target user particle group, and the user deviation degree of the target user attribute under the target object is determined according to the difference value of the index scores among the target user particle groups. The representative user particles are representative user particles in the target user particle groups, and can determine one or more user particles with highest or lowest index scores in each target user particle group, one or more user particles with most concentrated coordinate distribution in each target user particle group, and virtual user particles obtained by carrying out statistical calculation according to the user particles in each target user particle group, so that the data processing amount is reduced and the evaluation efficiency is improved by determining the representative user particles.
The user deviation degree may be determined directly according to the difference value of the index scores between the target user particle groups, or may be determined according to the statistical difference value such as variance, covariance, correlation, etc. of the index scores between the target user particle groups, which is not limited herein.
In an exemplary embodiment, according to a first attribute value PropValue_1, an initial user particle group group_k (k e [1,4 ]) corresponding to the first user data is determined, according to coordinates of each user particle group_i (i e [1,100 ]) after iterative updating, according to a preset coordinate distance threshold, each user particle group_i corresponds to 4 target user particle groups group_k (k e [1,4 ]), one user particle with the highest or lowest index score in each target user particle group group_k is determined as a representative user particle representing the target user particle group, the index score representing the user particle is determined as the index score of the target user particle group, and according to a difference value of index scores among the 4 target user particle groups group group_k, the user deviation degree of the object to be evaluated is determined.
According to the first attribute value, determining an initial user particle swarm corresponding to the first user data, wherein the initial user particle swarm is at least two preset user particle swarms, according to the coordinates of each user particle after iterative updating, determining a target user particle swarm corresponding to each user particle, and according to the difference value of index scores among the target user particle swarms, determining the user deviation degree of the object to be evaluated. The data volume of data processing is reduced, and the evaluation efficiency is improved.
Optionally, the change trend is monotonous change, the initial user particle swarm comprises a first direction user particle swarm and a second direction user particle swarm, wherein the first direction user particle swarm comprises first direction particles, and the second direction user particle swarm comprises second direction particles;
correspondingly, step 108 includes the following specific steps:
judging the first user particles as first direction particles or second direction particles according to the initial user particle group corresponding to the first user particles;
under the condition that the first user particles are particles in a first direction, iteratively updating the coordinates of the first user particles with the monotonically increasing index score as a target;
and under the condition that the first user particles are particles in the second direction, iteratively updating the coordinates of the first user particles with the aim of monotonically decreasing index scores.
By setting the change trend to be monotonous change, the user particles can be gathered according to a higher iteration speed, so that the iteration efficiency of the user particles is increased, and the concentrated and regular coordinate distribution of the user particles is obtained more quickly.
And setting particle motion directions for the initial user particle swarm in advance according to the target user attributes, wherein first direction particles in the first direction user particle swarm are user particles with index scores which are iteratively updated according to monotonically increasing directions, and second direction particles in the second direction user particle swarm are user particles with index scores which are iteratively updated according to monotonically decreasing directions. The first direction particles and the second direction particles are particles with opposite directions, and the user particles with different directions can be intensively distributed in each user particle group according to the higher index score and the lower index score by classifying the user particles.
And carrying out iterative updating on the coordinates of the first user particles by taking the monotonically increasing index score as a target when the first user particles are the first-direction particles, wherein the method is that the coordinates of the first user particles are iteratively updated according to the monotonically increasing first-direction iterative speed when the first user particles are the first-direction particles.
And (3) under the condition that the first user particles are particles in the second direction, taking the monotonic decrease of the index score as a target, carrying out iterative updating on the coordinates of the first user particles.
For example, according to the initial user particle group grouppat_le_k (k e [1,4 ]) corresponding to the first user particle pay_1, the first user particle pay_1 is determined to be the first direction particle or the second direction particle, the coordinates of the first user particle pay_1 are iteratively updated according to the monotonically increasing first direction iteration velocity v+ when the first user particle is the first direction particle, and the coordinates of the first user particle pay_1 are iteratively updated according to the monotonically decreasing second direction iteration velocity V-when the first user particle pay_1 is the second direction particle.
Judging whether the first user particle is a first-direction particle or a second-direction particle according to the initial user particle group corresponding to the first user particle, iteratively updating the coordinates of the first user particle with the monotonically increasing index score as a target when the first user particle is the first-direction particle, and iteratively updating the coordinates of the first user particle with the monotonically decreasing index score as a target when the first user particle is the second-direction particle. And simultaneously, the user particles are iteratively updated from two directions, so that the iteration updating efficiency and the coordinate distribution aggregation and regularity of the iteratively updated user particles are improved, and the efficiency and accuracy of the subsequent determination of the user deviation degree are improved.
Optionally, in the case that the first user particle is a first direction particle, iteratively updating the coordinates of the first user particle with the goal of monotonically increasing the index score, including the following specific steps:
under the condition that the first user particles are particles in a first direction, determining that the current coordinates of the first user particles are initial local coordinates, and determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles;
According to the initial global coordinates and the initial local coordinates, the coordinates of the first user particles are adjusted;
determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and determining the target local coordinates according to the updated index score and the historical index score;
if the preset iteration ending condition is not met, determining the target local coordinates as initial local coordinates, and returning to execute the step of determining initial global target coordinates according to index scores of all user particles in the initial user particle group corresponding to the first user particles until the preset iteration ending condition is met, so as to obtain the first user particles with complete iteration updating;
and/or the number of the groups of groups,
under the condition that the first user particles are particles in the second direction, taking monotonic decrease of the index score as a target, carrying out iterative updating on the coordinates of the first user particles, and comprising the following specific steps of:
under the condition that the first user particles are particles in the second direction, determining that the current coordinates of the first user particles are initial local coordinates, and determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles;
According to the initial global coordinates and the initial local coordinates, the coordinates of the first user particles are adjusted;
determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and determining the target local coordinates according to the updated index score and the historical index score;
if the preset iteration ending condition is not met, determining the target local coordinates as initial local coordinates, and returning to execute the step of determining the initial global target coordinates according to the index scores of all user particles in the initial user particle group corresponding to the first user particles until the preset iteration ending condition is met, so as to obtain the first user particles with complete iteration updating.
And determining target particles through index scores of the user particles, and moving each user particle in different directions towards the target particles in the corresponding directions, namely, iteratively updating coordinates, so that the coordinate distribution of the user particles in two directions is gathered around the target coordinates in the corresponding user particle groups, and centralized distribution in each user particle group is realized.
The global coordinates are coordinates of user particles having the highest index score or the lowest index score among the user particle groups, and when the first user particle is a first-direction particle, the global coordinates are coordinates of user particles having the highest index score among the user particle groups, and when the first user particle is a second-direction particle, the global coordinates are coordinates of user particles having the lowest index score among the user particle groups. The initial global coordinate is the global coordinate before coordinate adjustment in the current iteration, and the target global coordinate is the global coordinate after coordinate adjustment in the current iteration. For example, the initial user particle group to which the first user particle belongs includes 20 user particles, wherein coordinates of the user particle with the highest index score of the 20 user particles are determined as global coordinates of the initial user particle group, and subsequently, the coordinates of the user particles in the user particle group are updated, so that the updated user particles are gathered towards the global coordinates. The local coordinates are the historical coordinates with the highest or lowest historical index score of the first user particles, the local coordinates are the historical coordinates with the highest historical index score of the first user particles when the first user particles are the particles in the first direction, and the local coordinates are the historical coordinates with the lowest historical index score of the first user particles when the first user particles are the particles in the second direction. The initial local coordinate is the local coordinate of the first user particle before coordinate adjustment in the current iteration, and the target global coordinate is the local coordinate of the first user particle after coordinate adjustment in the current iteration. It should be noted that, with the iterative update of the coordinates of the user particles, the index score of each user particle is updated iteratively, and then both the global coordinates and the local coordinates are changed. For example, before coordinate adjustment in the current iteration, the historical index scores corresponding to the historical coordinates (P1, P2, P3, P4) of the first user particle are respectively: 0.72, 0.75, 0.68 and 0.79, determining P4 as an initial local coordinate, determining the current coordinate of the first user particle as P5 after coordinate adjustment in the current iteration, and determining the current coordinate as a target local coordinate with a corresponding index score of 0.85.
The preset iteration end condition may be a preset iteration number, or may be a preset coordinate distribution condition of the user particles (convergence condition of the user particle swarm), for example, the distance between any two user particles is smaller than a preset threshold, and for example, N user particles are included within the radius R of any user particle, which is not limited herein.
And adjusting the coordinates of the first user particles according to the initial global coordinates and the initial local coordinates by utilizing a preset iteration speed formula according to the initial global coordinates and the initial local coordinates.
The preset iteration speed formula is shown in formula 1:
wherein, the liquid crystal display device comprises a liquid crystal display device,
representing the iteration speed of the ith user particle in the (k+1) th iteration updating process, f () represents the calculation formula of the iteration speed,>
characterizing the initial local coordinates of the ith user particle in the kth iteration update procedure, a->
Characterizing the initial global coordinates of the initial user particle swarm of the ith user particle in the kth iterative updating process,/for>
Characterizing the initial local coordinates of the ith user particle in the kth iteration update procedure, a->
And (3) representing the initial global coordinates of the initial user particle swarm of the ith user particle in the kth iteration updating process. The i-th user particle carries out iterative updating of the coordinates under the combined action of the initial global coordinates and the initial local coordinates, and the updated target local coordinates and the target global coordinates are achieved.
Further, if only the actions of the global coordinates and the local coordinates are considered, the user particles are caused to move directly towards the global coordinates, the characteristics of the user particles themselves are lost, the final user particles are gathered in the global coordinates, and by considering the characteristics of the user particles themselves, the final user particle coordinates are distributed in a gathered and regular form similar to a 'star-planetary' form. Correspondingly, the preset iteration speed formula is adjusted to formula 2 in consideration of the characteristics of the user particles:
Wherein, the liquid crystal display device comprises a liquid crystal display device,
and (3) representing the iteration speed of the ith user particle in the kth iteration updating process.
And determining the local coordinates of the target according to the updated index score and the historical index score by comparing the updated index score with the historical index score.
In an exemplary embodiment, when the first user particle pair_1 is a first direction particle, determining that the current coordinate of the first user particle is an initial local coordinate p_best, determining an initial global coordinate g_best according to the index score of each user particle in an initial user particle group grouppaticle_k (k e [1,4 ]) corresponding to the first user particle, adjusting the coordinate of the first user particle according to the initial global coordinate and the initial local coordinate by using a formula 2, determining the matching degree Fit between the attribute value PropValue corresponding to the adjusted coordinate of the first user particle and the target object as an updated index score of the first user particle, comparing the updated index score with the historical index score, determining a target local coordinate p_best', if the convergence condition of the preset user particle group is not met, determining the target local coordinate as the initial local coordinate, and returning to execute the step of determining the initial global target according to the index score of each user particle in the initial user particle group corresponding to the first user particle until the convergence condition of the preset user particle is met, thereby obtaining the first iterative particle group; under the condition that the first user particle Paticle_1 is the second direction particle, determining the current coordinate of the first user particle as an initial local coordinate P_best, determining an initial global coordinate G_best according to the index score of each user particle in an initial user particle group GroupPaticle_k (k E [1,4 ]) corresponding to the first user particle, adjusting the coordinate of the first user particle according to the initial global coordinate and the initial local coordinate by utilizing a formula 2, determining the matching degree Fit between an attribute value PropValue corresponding to the adjusted coordinate of the first user particle and a target object as an updated index score of the first user particle, comparing the updated index score with a historical index score, determining a target local coordinate P_best', if the convergence condition of the preset user particle group is not met, determining the target local coordinate as the initial local coordinate, and returning to execute the steps of determining the initial global target coordinate according to the index score of each user particle in the initial user particle group corresponding to the first user particle until the preset convergence condition of the first user particle group is met, and obtaining the first user particle iterative updated.
In the embodiment of the specification, the coordinates of the user particles are iteratively updated from two directions at the same time, so that the iterative updating of the coordinates is performed by utilizing the global coordinates and the local coordinates on the basis of improving the iterative updating efficiency, the accuracy of the iterative updating of the coordinates of the first user particles is improved, the accuracy of index scores of the updated user particle swarms is ensured, and the accuracy of the determined user deviation degree is ensured.
Optionally, adjusting the coordinates of the first user particle according to the initial global coordinates and the initial local coordinates, including the following specific steps:
and adjusting the coordinates of the first user particles according to the initial global coordinates and the initial local coordinates and a preset random inertia parameter, wherein the random inertia parameter is a variable parameter meeting random distribution.
The inertia parameter is a weight parameter of an acting factor representing the iteration speed. From equation 2, in the embodiment of the present specification, the iteration speed of the ith user particle for the (k+1) th iteration
It is affected by three contributing factors: iteration speed of ith user particle in kth iteration update procedure +.>
The initial local coordinates of the user particle group where the ith user particle is in the kth iterative updating process and the initial global coordinates of the kth iterative updating process. The final coordinates of the ith user particle can be ensured by setting reasonable inertial parameters, and the final coordinates are not excessively dependent on global coordinates The local coordinates are not excessively relied on, and the global coordinates and the local coordinates are not excessively far away, so that the coordinate distribution of the final user particles is more regular.
The random inertia parameter is a weight variable parameter satisfying a random distribution, wherein the random distribution may be a 0-1 distribution, a normal distribution, a poisson distribution, a binomial distribution, etc., and is not limited herein. By setting random inertia parameters, the randomness of iterative updating of the coordinates of the user particles is improved, the capability of the user particles to get rid of local coordinates is improved, and the aggregation and regularity of the coordinate distribution of the updated user particles are further improved.
And adjusting the coordinates of the first user particles according to the initial global coordinates and the initial local coordinates and a preset random inertia parameter, wherein the random inertia parameter is a variable parameter meeting random distribution, and specifically, the coordinates of the first user particles are adjusted according to the initial global coordinates and the initial local coordinates and an iteration speed formula determined according to the preset random inertia parameter.
The iteration speed formula determined according to the preset random inertia parameter is shown in formula 3:
Wherein, the liquid crystal display device comprises a liquid crystal display device,
for the current coordinates of the ith user particle after the completion of the kth iteration update,/th user particle>
For the iteration speed of the ith user particle in the preset kth iteration update process +.>
Alpha is a constant weight parameter, +.>
C is a random parameter on a standard normal distribution 1 ·r 1 C, updating random inertia parameters of initial local coordinates of the ith user particle in the process of the preset kth iteration 1 Is a constant weight parameter, r 1 Is a random parameter on the 0-1 distribution, c 2 ·r 2 C is a random inertial parameter of an initial global coordinate in a kth iterative updating process 2 Is a constant weight parameter, r 2 Is a random parameter on a 0-1 distribution.
Illustratively, in the case that the first user particle Paticle_1 is a first direction particle, adjusting the coordinates of the first user particle according to the initial global coordinates G_Best and the initial local coordinates P_Best by using the above formula 3; in the case that the first user particle paycle_1 is a second direction particle, the coordinates of the first user particle are adjusted according to the initial global coordinates g_best and the initial local coordinates p_best using the above formula 3.
In the embodiment of the specification, by setting random inertia parameters, the randomness of iterative updating of the coordinates of the user particles is improved, the capability of the user particles to get rid of local target particles is improved, the aggregation and regularity of the coordinate distribution of the updated user particles are further improved, and the accuracy of the obtained user deviation degree is further improved.
Optionally, before step 106, the method further includes the following specific steps:
judging whether the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target object is recorded in advance;
if not, calculating to obtain the matching degree between the attribute value and the target object according to the historical interaction data between the historical user corresponding to the attribute value and the target object, and recording the matching degree;
if yes, obtaining the pre-recorded matching degree.
Because the user particles are the mass-free abstract particles corresponding to the user, after mapping each user particle to a preset coordinate space, the user particles can move according to a preset movement rule, each time a coordinate passes, the matching degree between the attribute value corresponding to the coordinate and the target object is determined, the matching degree is calculated by using a preset object recommendation algorithm, and when other user particles also pass the point, repeated calculation is caused, so that the data processing efficiency is reduced, after each preset iteration update, the matching degree is calculated to record the user particles, and the fact that the pre-recorded matching degree is directly obtained after the subsequent user particles reach the coordinate can be ensured, and repeated calculation is not needed.
The matching degree is calculated by the object recommendation algorithm according to the attribute value of the historical user data and the object data of the target object by utilizing the historical interaction data. The matching degree is dynamically recorded in a database, and correspondingly is dynamically obtained from the database.
The historical interaction data is the historical interaction data of the interaction between the historical user and the target object, for example, the target object is search information, the historical user obtains corresponding search information by inputting the keyword of 'ABC', and the keyword-search information (search) is the historical interaction data between the historical user and the search information.
The method includes the steps of judging whether the matching degree Fit between the attribute value corresponding to the current coordinate (X1, Y1) of the first user particle Patcle_1 and the target object is recorded in advance or not, if not, calculating to obtain the matching degree Fit between the attribute value and the target object according to the historical interaction data between the historical user corresponding to the attribute value and the target object, recording the matching degree Fit in a database, and if so, directly acquiring the pre-recorded matching degree Fit from the database.
Judging whether the matching degree between the attribute value corresponding to the current coordinate of the first user particle and the target object is recorded in advance, if not, calculating to obtain the matching degree between the attribute value and the target object according to the historical interaction data between the historical user corresponding to the attribute value and the target object, and recording the matching degree, if so, obtaining the pre-recorded matching degree. Repeated calculation is avoided, and evaluation efficiency is improved.
Optionally, determining whether the matching degree between the attribute value corresponding to the current coordinate of the first user particle and the target object is pre-recorded includes the following specific steps:
inquiring whether the matching degree between the attribute value corresponding to the current coordinate and the target object is recorded in a matching degree recording table in advance according to the current coordinate of the first user particle, wherein the matching degree recording table takes the coordinate as an index, and records the matching degree between the attribute value corresponding to each coordinate and the target object;
correspondingly, the matching degree is recorded, which comprises the following specific steps:
and recording the matching degree in a matching degree record table by taking the current coordinate as an index.
The matching degree record table is a data record table which is constructed in advance and recorded with matching degree, the matching degree record table is stored in a database, the matching degree record table takes coordinates as indexes, and the matching degree of attribute values corresponding to the coordinates and a target object is recorded.
For example, according to the current coordinates (X1, Y1) of the first user particle partial_1, inquiring whether the matching degree Fit between the attribute value corresponding to the current coordinates (X1, Y1) and the target object is recorded in the matching degree recording table Tab leFit in advance, if not, calculating to obtain the matching degree Fit between the attribute value and the target object according to the historical interaction data between the historical user corresponding to the attribute value and the target object, and recording the matching degree Fit in the matching degree recording table Tab leFit with the current coordinates (X1, Y1) as an index, if yes, directly obtaining the pre-recorded matching degree Fit from the matching degree recording table Tab leFit with the current coordinates (X1, Y1) as an index.
According to the current coordinates of the first user particles, inquiring whether the matching degree between the attribute values corresponding to the current coordinates and the target object is recorded in a matching degree record table in advance, wherein the matching degree record table takes the coordinates as indexes, records the matching degree between the attribute values corresponding to the coordinates and the target object, and records the matching degree in the matching degree record table by taking the current coordinates as indexes. Accuracy and efficiency of matching degree recording and obtaining are improved.
Optionally, step 102 includes the following specific steps:
providing a plurality of user data;
counting the distribution of attribute values of a plurality of user data, and screening the plurality of user data according to the counting result;
from the filtered user data, first user data is determined.
Since the user attributes of the object to be evaluated are not ideally uniformly distributed, there may be a case where the amounts of user data corresponding to different attribute values differ too much, for example, the amount of user data of young users and male users is greater than the amount of user data of elderly users and female users when applied to a commodity recommendation model of an e-commerce platform of a sports brand. The bias and deviation are usually aimed at users with smaller user data quantity, so that attribute distribution is required to be counted, a plurality of user data are screened according to the counted result, and the rationality of the coordinate distribution of the user particles after iterative updating is ensured.
The distribution of the attribute values of the plurality of user data is counted, the plurality of user data is screened according to the counted result, and the specific mode is that the distribution of the attribute values of the plurality of user data is counted, the attribute value distribution result is obtained, and the plurality of user data is screened according to the attribute value distribution result. Further, the screening may be performed on a plurality of user data for random sampling according to the attribute value distribution result, or may be performed on a plurality of user data for sampling satisfying the attribute value quantity constraint. The random sampling may be normal random sampling, poisson random sampling, etc. and satisfies the preset number of attribute values, that is, the sampling is performed according to the preset user data number threshold value of a certain attribute value.
Providing a plurality of user data, counting the distribution of attribute values of the plurality of user data, screening the plurality of user data according to the counting result, and determining the first user data from the screened user data. The reasonable selection of the user data is improved, the defect of insufficient acting force of the user data with small quantity is avoided, the coverage rate of the user particles constructed later in a preset coordinate space is improved, the rationality of the coordinate distribution of the user particles after iterative updating is improved, and the accuracy of user deviation degree is improved.
Referring to fig. 2, fig. 2 shows a flowchart of an evaluation method of a commodity recommendation system according to an embodiment of the present disclosure, including the following specific steps:
step 202: providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
step 204: according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
step 206: determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target commodity as the index score of the first user particles;
step 208: iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target;
step 210: and determining the user deviation degree of the commodity recommendation system according to the difference value of the index scores among the iteratively updated user particles.
The embodiment of the present disclosure is applied to a server side or a cloud side device of an application for evaluating a user deviation degree of a commodity recommendation system, where the cloud side device is a distributed virtual network device, and data transmission is implemented with the end side device through a protocol interface, which is not limited herein. The commodity recommendation system is an artificial intelligent model with commodity recommendation function in the field of electronic commerce. The commodity recommendation system realizes commodity recommendation functions by utilizing corresponding commodity recommendation algorithms.
For the same inventive concept as the embodiment of fig. 1, the specific manner of steps 202 to 210 refers to steps 102 to 110 in the embodiment of fig. 1, and will not be described again here.
In this embodiment of the present disclosure, first user data is provided, where the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute, according to the target user attribute and the first attribute value, first user particles corresponding to the first user data are mapped to a preset coordinate space, a matching degree between the attribute value corresponding to a current coordinate of the first user particles and a target commodity is determined as an index score of the first user particles, a change trend of the index score is taken as a target, coordinates of the first user particles are iteratively updated, and a user deviation degree of the commodity recommendation system is determined according to a difference value of the index score between the iteratively updated user particles. The user deviation degree of the commodity recommendation system can be determined only by calculating the difference value of the index scores of the user particles after the iterative updating is completed, so that the user deviation degree can be determined after the indexes of all the user data are gradually calculated and analyzed, the accuracy of the user deviation degree is ensured, and the evaluation efficiency is improved.
Referring to fig. 3, fig. 3 shows a flowchart of an evaluation method of an information recommendation system according to an embodiment of the present disclosure, including the following specific steps:
step 302: providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
step 304: according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
step 306: determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target recommendation information as index scores of the first user particles;
step 308: iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target;
step 310: and determining the user deviation degree of the information recommendation system according to the difference value of the index scores among the iteratively updated user particles.
The embodiment of the present disclosure is applied to a server side or a cloud side device for evaluating an application of a user deviation degree of an information recommendation system, where the cloud side device is a distributed virtual network device, and implements data transmission with the end side device through a protocol interface, which is not limited herein. The information recommendation system is an artificial intelligent model with an information recommendation function for information searching. The information recommendation system utilizes a corresponding information recommendation algorithm to realize an information recommendation function.
For the same inventive concept as the embodiment of fig. 1, the specific manner of steps 302 to 310 refers to steps 102 to 110 in the embodiment of fig. 1, and will not be described again here.
In this embodiment of the present disclosure, first user data is provided, where the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute, according to the target user attribute and the first attribute value, first user particles corresponding to the first user data are mapped to a preset coordinate space, a matching degree between the attribute value corresponding to a current coordinate of the first user particles and target recommendation information is determined as an index score of the first user particles, a change trend of the index score is taken as a target, coordinates of the first user particles are iteratively updated, and a user deviation degree of the information recommendation system is determined according to a difference value of the index score between the iteratively updated user particles. The user deviation degree of the information recommendation system can be determined by constructing a preset coordinate space for any target user attribute, mapping user particles corresponding to user data under the coordinate space, and taking the change trend of index scores as a target, automatically and iteratively updating the user particle coordinates, and only calculating the difference value of the index scores of the user particles subjected to iterative updating, so that the user deviation degree can be determined after the indexes of all the user data are gradually calculated and analyzed, and the evaluation efficiency is improved while the accuracy of the user deviation degree is ensured.
The method for evaluating an object is further described below with reference to fig. 4 by taking an application of the method for evaluating an object provided in the present specification to an online social platform as an example. FIG. 4 is a flowchart illustrating a process of an object assessment method applied to an online social platform according to an embodiment of the present disclosure, including the following specific steps.
Step 402: providing a plurality of user data, target topic information and a plurality of historical interaction data of a user recommendation model of the online social platform;
the user data includes target user attributes of the academy and the region, and attribute values corresponding to the academy and the region, and the target topic information can be a plurality of topic information.
Step 404: counting the distribution of attribute values of a plurality of user data, and screening the plurality of user data according to the counting result;
step 406: determining first user data from the screened user data;
step 408: according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space, and according to the first attribute value, determining an initial user particle group corresponding to the first user;
The preset coordinate space is constructed in advance according to the object user attribute of the academic and the region.
Step 410: inquiring whether the matching degree between the attribute value corresponding to the current coordinate and the target topic information is recorded in a matching degree record table in advance by taking the current coordinate of the first user particle as an index;
the matching degree record table takes coordinates as indexes, and records the matching degree of attribute values corresponding to the coordinates and target topic information.
Step 412: if not, calculating to obtain the matching degree between the attribute value corresponding to the current coordinate and the target topic information by using a topic recommendation model according to the attribute value corresponding to the current coordinate of the first user particle and the object data of the target topic information, and recording the matching degree in a matching degree recording table by taking the current coordinate as an index;
the topic recommendation model is an object recommendation algorithm for calculating the matching degree according to historical interaction data.
Step 414: if yes, acquiring the matching degree from the matching degree record table;
step 416: determining the matching degree between the attribute value corresponding to the current coordinate and the target topic information as the index score of the first user particle;
step 418: judging the first user particles as first direction particles or second direction particles according to the initial user particle group corresponding to the first user particles;
Step 420: under the condition that the first user particles are particles in a first direction, determining that the current coordinates of the first user particles are initial local coordinates, determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles, and adjusting the coordinates of the first user particles according to an iteration speed formula determined by the initial global coordinates and the initial local coordinates and a preset random inertia parameter;
step 422: determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and comparing the updated index score with the historical index score to determine the target local coordinates;
step 424: if the preset iteration ending condition is not met, determining the target local coordinates as initial local coordinates, and returning to execute the step of determining initial global target coordinates according to index scores of all user particles in the initial user particle group corresponding to the first user particles until the preset iteration ending condition is met, so as to obtain the first user particles with complete iteration updating;
the iteration speed formula is shown in formula 3 in the embodiment of fig. 1.
Step 426: under the condition that the first user particles are particles in a first direction, determining that the current coordinates of the first user particles are initial local coordinates, determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles, and adjusting the coordinates of the first user particles according to an iteration speed formula determined by the initial global coordinates and the initial local coordinates and a preset random inertia parameter;
step 428: determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and comparing the updated index score with the historical index score to determine the target local coordinates;
step 430: if the preset iteration ending condition is not met, determining the target local coordinates as initial local coordinates, and returning to execute the step of determining initial global target coordinates according to index scores of all user particles in the initial user particle group corresponding to the first user particles until the preset iteration ending condition is met, so as to obtain the first user particles with complete iteration updating;
the iteration speed formula is shown in formula 3 in the embodiment of fig. 1.
Step 432: determining a target user particle group corresponding to each user particle according to the coordinates of each user particle after iterative updating;
step 434: and determining the user deviation degree of the user recommendation model according to the difference value of the index scores among the target user particle swarms.
In the embodiment of the specification, a preset coordinate space can be constructed for any target user attribute, user particles corresponding to user data are mapped under the coordinate space, the reasonable selection of the user data is improved, the shortage of acting force of a small number of user data is avoided, the coverage rate of the subsequently constructed user particles under the preset coordinate space is improved, the rationality of the coordinate distribution of the iteratively updated user particles is improved, the accuracy of the user deviation degree is improved, the iteratively updated user particles are simultaneously carried out from two directions, the iterative updating efficiency and the coordinate distribution aggregation and regularity of the iteratively updated user particles are improved, the efficiency and the accuracy of the subsequently determined user deviation degree are improved, the change trend of index scores is used as a target, the automatic iterative updating of the coordinates of the user particles is carried out, the user deviation degree of the target user attribute under the target information can be determined only according to the difference value of the index scores of the iteratively updated target user particles, the calculation and the random parameter distribution of the user particles is avoided, the user deviation degree of the user data is gradually calculated and the user data is further improved, the accuracy of the user deviation degree is further improved, the user model is further improved, the accuracy of the user deviation is further improved, the user deviation of the user deviation is further calculated, and the user parameter of the user deviation is further calculated, and the user deviation is further improved, and the user particle error is further estimated, and the user particle error is calculated.
Fig. 5 is a schematic flow chart of an object evaluation method according to an embodiment of the present disclosure.
As shown in fig. 5, the object recommendation model is an object recommendation algorithm in the object to be evaluated, and after the user data and the object data (the object 1 data and the object 2 data) are densely embedded and encoded by using the dense embedding layer, the corresponding factorization and feature vector processing are respectively performed by inputting the factorization layer and the hiding layer, and the target object is output through the output layer. For the user bias degree of the object recommendation model, the object evaluation method of the embodiment of fig. 1 is utilized, firstly, user data is provided, user particles of each user data are mapped to a preset coordinate space according to target user attributes of the user data and attribute values corresponding to the target user attributes, user particle groups corresponding to the user particles are determined, initialization of the user particles is achieved, then corresponding attribute values are determined according to current coordinates of the user particles, the matching degree of the attribute values and the target objects is determined according to the attribute values, the matching degree is determined to be an index parameter of the user particles, the change trend of the index parameter is taken as a target, coordinates of the user particles are iteratively updated, the user bias degree of the object to be evaluated is determined according to the difference value of index scores of the updated target user particle groups (the target user particle groups in the first direction and the target user particle groups in the second direction), and the object recommendation model is optimized according to the user bias degree.
Corresponding to the above method embodiments, the present disclosure further provides an object evaluation device embodiment, and fig. 6 shows a schematic structural diagram of an object evaluation device provided in one embodiment of the present disclosure. As shown in fig. 6, the apparatus includes:
a first providing module 602 configured to provide first user data, wherein the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute;
the first mapping module 604 is configured to map first user particles corresponding to the first user data into a preset coordinate space according to the target user attribute and the first attribute value;
a first index score determining module 606 configured to determine a degree of matching between an attribute value corresponding to a current coordinate of the first user particle and the target object as an index score of the first user particle;
a first updating module 608 configured to iteratively update the coordinates of the first user particle with respect to a trend of the index score;
the first deviation determining module 610 is configured to determine a user deviation degree of the object to be evaluated according to the iteratively updated difference value of the index score between the user particles.
Optionally, the apparatus further comprises:
The initial user particle swarm determination module is configured to determine an initial user particle swarm corresponding to the first user data according to the first attribute value, wherein the initial user particle swarm is at least two preset user particle swarms;
the first bias determination module 610 is further configured to:
determining a target user particle group corresponding to each user particle according to the coordinates of each user particle after iterative updating; and determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the target user particle swarms.
Optionally, the change trend is monotonous change, the initial user particle swarm comprises a first direction user particle swarm and a second direction user particle swarm, wherein the first direction user particle swarm comprises first direction particles, and the second direction user particle swarm comprises second direction particles;
correspondingly, the first update module 608 is further configured to:
judging the first user particles as first direction particles or second direction particles according to the initial user particle group corresponding to the first user particles; under the condition that the first user particles are particles in a first direction, iteratively updating the coordinates of the first user particles with the monotonically increasing index score as a target; and under the condition that the first user particles are particles in the second direction, iteratively updating the coordinates of the first user particles with the aim of monotonically decreasing index scores.
Optionally, the apparatus further comprises:
a target particle determination module configured to determine a global first direction target particle and a global second direction target particle according to the index score of each user particle; when the first user is a first-direction particle, determining a local first-direction target particle according to index scores of all user particles in an initial user particle group corresponding to the first user particle; under the condition that the first user is the second-direction particle, determining local second-direction target particles according to index scores of all user particles in an initial user particle group where the first user particle is located;
correspondingly, the first update module 608 is further configured to:
under the condition that the first user particles are particles in a first direction, determining that the current coordinates of the first user particles are initial local coordinates, and determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles; according to the initial global coordinates and the initial local coordinates, the coordinates of the first user particles are adjusted; determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and determining the target local coordinates according to the updated index score and the historical index score; if the preset iteration ending condition is not met, determining the target local coordinates as initial local coordinates, and returning to execute the step of determining initial global target coordinates according to index scores of all user particles in the initial user particle group corresponding to the first user particles until the preset iteration ending condition is met, so as to obtain the first user particles with complete iteration updating;
And/or the number of the groups of groups,
under the condition that the first user particles are particles in the second direction, determining that the current coordinates of the first user particles are initial local coordinates, and determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles; according to the initial global coordinates and the initial local coordinates, the coordinates of the first user particles are adjusted; determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and determining the target local coordinates according to the updated index score and the historical index score; if the preset iteration ending condition is not met, determining the target local coordinates as initial local coordinates, and returning to execute the step of determining the initial global target coordinates according to the index scores of all user particles in the initial user particle group corresponding to the first user particles until the preset iteration ending condition is met, so as to obtain the first user particles with complete iteration updating.
Optionally, the first update module 608 is further configured to:
and adjusting the coordinates of the first user particles according to the initial global coordinates and the initial local coordinates and a preset random inertia parameter, wherein the random inertia parameter is a variable parameter meeting random distribution.
Optionally, the apparatus further comprises:
the matching degree recording module is configured to judge whether the matching degree between the attribute value corresponding to the current coordinate of the first user particle and the target object is recorded in advance or not; if not, calculating to obtain the matching degree between the attribute value and the target object according to the historical interaction data between the historical user corresponding to the attribute value and the target object, and recording the matching degree; if yes, obtaining the pre-recorded matching degree.
Optionally, the matching degree recording module is further configured to:
inquiring whether the matching degree between the attribute value corresponding to the current coordinate and the target object is recorded in a matching degree recording table in advance according to the current coordinate of the first user particle, wherein the matching degree recording table takes the coordinate as an index, and records the matching degree between the attribute value corresponding to each coordinate and the target object; and recording the matching degree in a matching degree record table by taking the current coordinate as an index.
Optionally, the first providing module 602 is further configured to:
providing a plurality of user data; counting the distribution of attribute values of a plurality of user data, and screening the plurality of user data according to the counting result; from the filtered user data, first user data is determined.
In the embodiment of the present disclosure, first user data is provided, where the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute, and according to the target user attribute and the first attribute value, first user particles corresponding to the first user data are mapped into a preset coordinate space, and a matching degree between the attribute value corresponding to a current coordinate of the first user particles and a target object is determined as an index score of the first user particles; and iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target, and determining the user deviation degree of the object to be evaluated according to the difference value of the index score among the user particles after the iterative updating. The user deviation degree of the object to be evaluated can be determined by only carrying out difference value calculation on the index scores of the user particles after the iterative updating is completed, so that the user deviation degree can be determined after the index of all user data is gradually calculated and analyzed, and the evaluation efficiency is improved while the accuracy of the user deviation degree is ensured.
The above is a schematic scheme of an object evaluation apparatus of the present embodiment. It should be noted that, the technical solution of the object evaluation device and the technical solution of the object evaluation method belong to the same concept, and details of the technical solution of the object evaluation device, which are not described in detail, can be referred to the description of the technical solution of the object evaluation method.
Corresponding to the method embodiment, the present disclosure further provides an embodiment of an evaluation device of the commodity recommendation system, and fig. 7 shows a schematic structural diagram of an evaluation device of the commodity recommendation system according to one embodiment of the present disclosure. As shown in fig. 7, the apparatus includes:
a second providing module 702 configured to provide first user data, wherein the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute;
a second mapping module 704, configured to map, according to the target user attribute and the first attribute value, the first user particle corresponding to the first user data into a preset coordinate space;
a second index score determining module 706 configured to determine, as an index score of the first user particle, a degree of matching between an attribute value corresponding to the current coordinate of the first user particle and the target commodity;
A second updating module 708 configured to iteratively update the coordinates of the first user particles with respect to the trend of the index score
The second deviation determining module 710 is configured to determine a user deviation of the commodity recommendation system according to the iteratively updated difference value of the index score between the user particles.
In this embodiment of the present disclosure, first user data is provided, where the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute, according to the target user attribute and the first attribute value, first user particles corresponding to the first user are mapped to a preset coordinate space, a matching degree between the attribute value corresponding to a current coordinate of the first user particles and a target commodity is determined as an index score of the first user particles, a change trend of the index score is taken as a target, coordinates of the first user particles are iteratively updated, and a user deviation degree of the commodity recommendation system is determined according to a difference value of the index score between the iteratively updated user particles. The user deviation degree of the commodity recommendation system can be determined only by calculating the difference value of the index scores of the user particles after the iterative updating is completed, so that the user deviation degree can be determined after the indexes of all the user data are gradually calculated and analyzed, the accuracy of the user deviation degree is ensured, and the evaluation efficiency is improved.
The above is an exemplary embodiment of an evaluation device of a commodity recommendation system of the present embodiment. It should be noted that, the technical solution of the evaluation device of the commodity recommendation system and the technical solution of the evaluation method of the commodity recommendation system belong to the same concept, and the details of the technical solution of the evaluation device of the commodity recommendation system, which are not described in detail, can be referred to the description of the technical solution of the evaluation method of the commodity recommendation system.
Corresponding to the above method embodiments, the present disclosure further provides an embodiment of an evaluation device of an information recommendation system, and fig. 8 shows a schematic structural diagram of an evaluation device of an information recommendation system according to one embodiment of the present disclosure. As shown in fig. 8, the apparatus includes:
a third providing module 802 configured to provide first user data, wherein the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute;
a third mapping module 804, configured to map, according to the target user attribute and the first attribute value, the first user particle corresponding to the first user data into a preset coordinate space;
a third index score determining module 806 configured to determine, as an index score of the first user particle, a degree of matching between an attribute value corresponding to the current coordinate of the first user particle and the target recommendation information;
A third updating module 808 configured to iteratively update the coordinates of the first user particle with respect to a trend of the index score;
the third deviation determining module 810 is configured to determine the user deviation of the information recommendation system according to the iteratively updated difference value of the index score between the user particles.
In this embodiment of the present disclosure, first user data is provided, where the first user data includes a target user attribute and a first attribute value corresponding to the target user attribute, according to the target user attribute and the first attribute value, first user particles corresponding to the first user data are mapped to a preset coordinate space, a matching degree between the attribute value corresponding to a current coordinate of the first user particles and target recommendation information is determined as an index score of the first user particles, a change trend of the index score is taken as a target, coordinates of the first user particles are iteratively updated, and a user deviation degree of the information recommendation system is determined according to a difference value of the index score between the iteratively updated user particles. The user deviation degree of the information recommendation system can be determined by constructing a preset coordinate space for any target user attribute, mapping user particles corresponding to user data under the coordinate space, and taking the change trend of index scores as a target, automatically and iteratively updating the user particle coordinates, and only calculating the difference value of the index scores of the user particles subjected to iterative updating, so that the user deviation degree can be determined after the indexes of all the user data are gradually calculated and analyzed, and the evaluation efficiency is improved while the accuracy of the user deviation degree is ensured.
The above is an exemplary scheme of an evaluation device of an information recommendation system of the present embodiment. It should be noted that, the technical solution of the evaluation device of the information recommendation system and the technical solution of the evaluation method of the information recommendation system belong to the same concept, and details of the technical solution of the evaluation device of the information recommendation system, which are not described in detail, can be referred to the description of the technical solution of the evaluation method of the information recommendation system.
FIG. 9 illustrates a block diagram of a computing device provided by one embodiment of the present description. The components of computing device 900 include, but are not limited to, memory 910 and processor 920. Processor 920 is coupled to memory 910 via bus 930 with database 950 configured to hold data.
Computing device 900 also includes an access device 940, access device 940 enabling computing device 900 to communicate via one or more networks 960. Examples of such networks include PSTN (Public Switched Telephone Network ), LAN (Local Area Network, local area network), WAN (Wide Area Network ), PAN (Personal Area Network, personal area network), or a combination of communication networks such as the internet. The access device 940 may include one or more of any type of network interface, wired or wireless (e.g., NIC (Network Interface Controller, network interface card)), such as IEEE802.12WLAN (Wireless Local Area Networks, wireless local area network) wireless interface, wi-MAX (World Interoperability for Microwave Access, worldwide interoperability for microwave access) interface, ethernet interface, USB (Universal Serial Bus ) interface, cellular network interface, bluetooth interface, NFC (Near Field Communication ) interface, and so forth.
In one embodiment of the present description, the above-described components of computing device 900 and other components not shown in FIG. 9 may also be connected to each other, for example, by a bus. It should be understood that the block diagram of the computing device illustrated in FIG. 9 is for exemplary purposes only and is not intended to limit the scope of the present description. Those skilled in the art may add or replace other components as desired.
Computing device 900 may be any type of stationary or mobile computing device including a mobile computer or mobile computing device (e.g., tablet, personal digital assistant, laptop, notebook, netbook, etc.), mobile phone (e.g., smart phone), wearable computing device (e.g., smart watch, smart glasses, etc.), or other type of mobile device, or a stationary computing device such as a desktop computer or PC (Personal Computer ). Computing device 900 may also be a mobile or stationary server.
The processor 920 is configured to execute computer-executable instructions that, when executed by the processor, implement the above-described object assessment method, commodity recommendation system assessment method, or information recommendation system assessment method.
The foregoing is a schematic illustration of a computing device of this embodiment. It should be noted that, the technical solution of the computing device belongs to the same concept as the technical solution of the object evaluation method, the evaluation method of the commodity recommendation system and the evaluation method of the information recommendation system, and details of the technical solution of the computing device, which are not described in detail, can be referred to the description of the technical solution of the object evaluation method, the evaluation method of the commodity recommendation system or the evaluation method of the information recommendation system.
An embodiment of the present disclosure also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described object evaluation method, the commodity recommendation system evaluation method, or the information recommendation system evaluation method.
The above is an exemplary version of a computer-readable storage medium of the present embodiment. It should be noted that, the technical solution of the storage medium and the technical solutions of the object evaluation method, the evaluation method of the commodity recommendation system and the evaluation method of the information recommendation system described above belong to the same concept, and the details of the technical solution of the storage medium which are not described in detail can be referred to the description of the technical solutions of the object evaluation method, the evaluation method of the commodity recommendation system or the evaluation method of the information recommendation system.
An embodiment of the present disclosure further provides a computer program, wherein the computer program when executed in a computer causes the computer to execute the steps of the above-described object evaluation method, the evaluation method of the commodity recommendation system, or the evaluation method of the information recommendation system.
The above is an exemplary version of a computer program of the present embodiment. It should be noted that, the technical solution of the computer program and the technical solutions of the object evaluation method, the evaluation method of the commodity recommendation system and the evaluation method of the information recommendation system described above belong to the same concept, and the detailed contents of the technical solution of the computer program which are not described in detail can be referred to the description of the technical solutions of the object evaluation method, the evaluation method of the commodity recommendation system or the evaluation method of the information recommendation system.
The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
The computer instructions include computer program code that may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier wave signal, a telecommunication signal, a software distribution medium, and so forth.
It should be noted that, for simplicity of description, the foregoing method embodiments are all expressed as a series of combinations of actions, but it should be understood by those skilled in the art that the embodiments are not limited by the order of actions described, as some steps may be performed in other order or simultaneously according to the embodiments of the present disclosure. Further, those skilled in the art will appreciate that the embodiments described in the specification are all preferred embodiments, and that the acts and modules referred to are not necessarily all required for the embodiments described in the specification.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to the related descriptions of other embodiments.
The preferred embodiments of the present specification disclosed above are merely used to help clarify the present specification. Alternative embodiments are not intended to be exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the teaching of the embodiments. The embodiments were chosen and described in order to best explain the principles of the embodiments and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention. This specification is to be limited only by the claims and the full scope and equivalents thereof.
Claims (12)
1. An object assessment method, comprising:
providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
Determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target object as the index score of the first user particles;
iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target;
and determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the user particles after iterative updating.
2. The method of claim 1, further comprising, after said mapping first user particles corresponding to said first user data into a preset coordinate space according to said target user attribute and said first attribute value:
determining an initial user particle swarm corresponding to the first user data according to the first attribute value, wherein the initial user particle swarm is at least two preset user particle swarms;
the determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the user particles after iterative updating comprises the following steps:
determining a target user particle group corresponding to each user particle according to the coordinates of each user particle after iterative updating;
and determining the user deviation degree of the object to be evaluated according to the difference value of the index scores among the target user particle swarms.
3. The method of claim 2, the trend of change being monotonic, the initial population of user particles comprising a first direction population of user particles and a second direction population of user particles, wherein the first direction population of user particles comprises first direction particles and the second direction population of user particles comprises second direction particles;
the iteratively updating the coordinates of the first user particle with the change trend of the index score as a target includes:
judging the first user particles as first direction particles or second direction particles according to the initial user particle group corresponding to the first user particles;
under the condition that the first user particles are particles in a first direction, iteratively updating the coordinates of the first user particles with the monotonically increasing index score as a target;
and under the condition that the first user particles are particles in the second direction, iteratively updating the coordinates of the first user particles by taking the monotonic decrease of the index score as a target.
4. A method according to claim 3, wherein, in the case that the first user particle is a first-direction particle, iteratively updating the coordinates of the first user particle with the goal of monotonically increasing the index score, comprises:
Under the condition that the first user particles are particles in a first direction, determining that the current coordinates of the first user particles are initial local coordinates, and determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles;
adjusting the coordinates of the first user particles according to the initial global coordinates and the initial local coordinates;
determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and determining the target local coordinates according to the updated index score and the historical index score;
if the preset iteration ending condition is not met, determining the target local coordinate as the initial local coordinate, and returning to the step of executing the index score of each user particle in the initial user particle group corresponding to the first user particle to determine the initial global target coordinate until the preset iteration ending condition is met, so as to obtain the first user particle with the completed iteration update;
and/or the number of the groups of groups,
and in the case that the first user particle is a particle in a second direction, iteratively updating the coordinates of the first user particle with the goal of monotonically decreasing index score, including:
Under the condition that the first user particles are particles in a second direction, determining that the current coordinates of the first user particles are initial local coordinates, and determining initial global coordinates according to index scores of all user particles in an initial user particle group corresponding to the first user particles;
adjusting the coordinates of the first user particles according to the initial global coordinates and the initial local coordinates;
determining the matching degree between the attribute value corresponding to the adjusted coordinates of the first user particles and the target object as the updated index score of the first user particles, and determining the target local coordinates according to the updated index score and the historical index score;
if the preset iteration ending condition is not met, determining the target local coordinate as the initial local coordinate, and returning to the step of determining the initial global target coordinate according to the index score of each user particle in the initial user particle group corresponding to the first user particle until the preset iteration ending condition is met, so as to obtain the first user particle with the completed iteration updating.
5. The method of claim 4, the adjusting the coordinates of the first user particle according to the initial global coordinates and the initial local coordinates, comprising:
And adjusting the coordinates of the first user particles according to the initial global coordinates and the initial local coordinates and a preset random inertia parameter, wherein the random inertia parameter is a variable parameter meeting random distribution.
6. The method of claim 1, further comprising:
judging whether the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target object is recorded in advance;
if not, calculating to obtain the matching degree between the attribute value and the target object according to the historical interaction data between the historical user corresponding to the attribute value and the target object, and recording the matching degree;
if yes, acquiring the pre-recorded matching degree.
7. The method of claim 6, wherein determining whether the degree of matching between the attribute value corresponding to the current coordinates of the first user particle and the target object is pre-recorded comprises:
inquiring whether the matching degree between the attribute value corresponding to the current coordinate and the target object is recorded in a matching degree recording table in advance according to the current coordinate of the first user particle, wherein the matching degree recording table takes the coordinate as an index and records the matching degree between the attribute value corresponding to each coordinate and the target object;
The recording of the matching degree comprises the following steps:
and recording the matching degree in the matching degree recording table by taking the current coordinate as an index.
8. The method of claim 1, the providing first user data comprising:
providing a plurality of user data;
counting the distribution of the attribute values of the plurality of user data, and screening the plurality of user data according to the counting result;
from the filtered user data, first user data is determined.
9. An evaluation method of a commodity recommendation system, comprising:
providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target commodity as the index score of the first user particles;
iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target;
And determining the user deviation degree of the commodity recommendation system according to the difference value of the index scores among the iteratively updated user particles.
10. An evaluation method of an information recommendation system, comprising:
providing first user data, wherein the first user data comprises a target user attribute and a first attribute value corresponding to the target user attribute;
according to the target user attribute and the first attribute value, mapping first user particles corresponding to the first user data into a preset coordinate space;
determining the matching degree between the attribute value corresponding to the current coordinates of the first user particles and the target recommendation information as the index score of the first user particles;
iteratively updating the coordinates of the first user particles by taking the change trend of the index score as a target;
and determining the user deviation degree of the information recommendation system according to the difference value of the index scores among the iteratively updated user particles.
11. A computing device, comprising:
a memory and a processor;
the memory is configured to store computer-executable instructions that, when executed by the processor, implement the object assessment method according to any one of claims 1 to 8, the merchandise recommendation assessment method according to claim 9, or the steps of the information recommendation assessment method according to claim 10.
12. A computer-readable storage medium storing computer-executable instructions which, when executed by a processor, implement the method of evaluating an object according to any one of claims 1 to 8, the method of evaluating a merchandise recommendation system according to claim 9, or the steps of the method of evaluating an information recommendation system according to claim 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310231352.7A CN116205536B (en) | 2023-03-06 | 2023-03-06 | Object evaluation method, computing device, and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310231352.7A CN116205536B (en) | 2023-03-06 | 2023-03-06 | Object evaluation method, computing device, and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116205536A true CN116205536A (en) | 2023-06-02 |
| CN116205536B CN116205536B (en) | 2023-10-13 |
Family
ID=86517128
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310231352.7A Active CN116205536B (en) | 2023-03-06 | 2023-03-06 | Object evaluation method, computing device, and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116205536B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108073672A (en) * | 2017-04-12 | 2018-05-25 | 北京市商汤科技开发有限公司 | Business object recommends method, apparatus and electronic equipment |
| CN109275094A (en) * | 2018-11-02 | 2019-01-25 | 北京邮电大学 | A kind of continuous covering method of high energy efficiency unmanned plane covering point and a device |
| CN110930182A (en) * | 2019-11-08 | 2020-03-27 | 中国农业大学 | Improved particle swarm optimization algorithm-based client classification method and device |
| CN111798336A (en) * | 2020-07-09 | 2020-10-20 | 山东大学 | Typical user-based resident electricity fee package recommendation method and system |
| CN113138848A (en) * | 2020-01-20 | 2021-07-20 | 中移(上海)信息通信科技有限公司 | Method, device, equipment and storage medium for acquiring resource allocation strategy |
| CN113268656A (en) * | 2021-04-15 | 2021-08-17 | 北京沃东天骏信息技术有限公司 | User recommendation method and device, electronic equipment and computer storage medium |
| CN113420067A (en) * | 2021-06-22 | 2021-09-21 | 北京房江湖科技有限公司 | Method and device for evaluating position credibility of target location |
| CN113470072A (en) * | 2021-07-06 | 2021-10-01 | 吉林省田车科技有限公司 | Particle swarm target tracking algorithm based on moving particles |
| CN114519135A (en) * | 2020-11-19 | 2022-05-20 | 上海汽车集团股份有限公司 | Interest point recommendation method based on simulated annealing particle swarm algorithm |
| CN115309985A (en) * | 2022-07-27 | 2022-11-08 | 阿里巴巴（中国）有限公司 | Fairness evaluation method and AI model selection method of recommendation algorithm |
| US20220405129A1 (en) * | 2021-06-22 | 2022-12-22 | Nanjing University Of Posts And Telecommunications | Workflow scheduling method and system based on multi-target particle swarm algorithm, and storage medium |
-
2023
- 2023-03-06 CN CN202310231352.7A patent/CN116205536B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108073672A (en) * | 2017-04-12 | 2018-05-25 | 北京市商汤科技开发有限公司 | Business object recommends method, apparatus and electronic equipment |
| CN109275094A (en) * | 2018-11-02 | 2019-01-25 | 北京邮电大学 | A kind of continuous covering method of high energy efficiency unmanned plane covering point and a device |
| CN110930182A (en) * | 2019-11-08 | 2020-03-27 | 中国农业大学 | Improved particle swarm optimization algorithm-based client classification method and device |
| CN113138848A (en) * | 2020-01-20 | 2021-07-20 | 中移(上海)信息通信科技有限公司 | Method, device, equipment and storage medium for acquiring resource allocation strategy |
| CN111798336A (en) * | 2020-07-09 | 2020-10-20 | 山东大学 | Typical user-based resident electricity fee package recommendation method and system |
| CN114519135A (en) * | 2020-11-19 | 2022-05-20 | 上海汽车集团股份有限公司 | Interest point recommendation method based on simulated annealing particle swarm algorithm |
| CN113268656A (en) * | 2021-04-15 | 2021-08-17 | 北京沃东天骏信息技术有限公司 | User recommendation method and device, electronic equipment and computer storage medium |
| CN113420067A (en) * | 2021-06-22 | 2021-09-21 | 北京房江湖科技有限公司 | Method and device for evaluating position credibility of target location |
| US20220405129A1 (en) * | 2021-06-22 | 2022-12-22 | Nanjing University Of Posts And Telecommunications | Workflow scheduling method and system based on multi-target particle swarm algorithm, and storage medium |
| CN113470072A (en) * | 2021-07-06 | 2021-10-01 | 吉林省田车科技有限公司 | Particle swarm target tracking algorithm based on moving particles |
| CN115309985A (en) * | 2022-07-27 | 2022-11-08 | 阿里巴巴（中国）有限公司 | Fairness evaluation method and AI model selection method of recommendation algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116205536B (en) | 2023-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hao et al. | Robust vehicle pre‐allocation with uncertain covariates | |
| Chatterjee et al. | Forest Type Classification: A hybrid NN-GA model based approach | |
| US11599393B2 (en) | Guaranteed quality of service in cloud computing environments | |
| JP2019057286A (en) | Systems, methods and apparatuses for implementing data upload, processing and predictive query api exposure | |
| CN106776928B (en) | Position recommendation method based on memory computing framework and fusing social contact and space-time data | |
| WO2021203854A1 (en) | User classification method and apparatus, computer device and storage medium | |
| CN109360097A (en) | Prediction of Stock Index method, apparatus, equipment and storage medium based on deep learning | |
| CN104951446A (en) | Big data processing method and platform | |
| CN109919202A (en) | Disaggregated model training method and device | |
| CN111695046B (en) | User portrait inference method and device based on spatio-temporal mobile data representation learning | |
| CN110119772B (en) | Three-dimensional model classification method based on geometric shape feature fusion | |
| CN109583949B (en) | User switching prediction method and system | |
| CN111931053A (en) | Item pushing method and device based on clustering and matrix decomposition | |
| CN108346067A (en) | Social networks advertisement sending method based on natural language processing | |
| CN111626767B (en) | Resource data issuing method, device and equipment | |
| CN114998712B (en) | Image recognition method, storage medium, and electronic device | |
| Unger et al. | Inferring contextual preferences using deep auto-encoding | |
| CN112528164A (en) | User collaborative filtering recall method and device | |
| CN112866486B (en) | Multi-source feature-based fraud telephone identification method, system and equipment | |
| CN116205536B (en) | Object evaluation method, computing device, and readable storage medium | |
| CN113409096B (en) | Target object identification method and device, computer equipment and storage medium | |
| CN115309985A (en) | Fairness evaluation method and AI model selection method of recommendation algorithm | |
| CN112784160B (en) | Interest point recommendation method based on tensor decomposition and considering context semantic information | |
| CN115438249A (en) | QoS prediction model construction method, device, equipment and readable storage medium | |
| CN114626459A (en) | Hyperspectral image space spectrum combined classification method and system using particle swarm optimization multi-core support vector machine |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |