Sorting method and device based on multi-stage viewpoint interaction and computer equipment

Abstract

The invention relates to a sorting method and device based on multi-stage viewpoint interaction and computer equipment. The method comprises the following steps: determining a tetrad of a user and an interactive learning object of the user in a social network, and calculating the score contribution of to-be-sorted articles of the user according to a preset learning rate and sorting in a current interaction round, calculating an aggregation scoring contribution of the interactive learning object set according to a preset learning rate, the familiarity of the interactive learning objects and a sequence of the interactive learning objects, and updating the sequence according to the scoring contribution and the aggregation scoring contribution to obtain a learning sequence; according to the preference and the preference fixity, performing preference updating on the learning sequence to obtain an updated sequence; and when the consistency proportion of the updated sorting of the user reaches a threshold value or reaches the maximum number of interactive iterations, outputting a consistency sorting result. By adopting the method, multi-user viewpoint fusion and sorting result output can be carried out more truly and flexibly.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a sorting method, apparatus and computer device based on multi-stage opinion interaction. Background Art

[0002] Group decision making (GDM) involves a collaborative process in which multiple individuals come together to collect a wide range of ideas, integrate various opinions, and reach a unified decision. By integrating the opinions and knowledge of multiple individuals, GDM can achieve more accurate and robust decision results in complex environments. The GDM method usually includes two steps: the consensus reaching process (CRP) and the solution selection process. In the early stages of group decision making, users' opinions often conflict or diverge significantly. Therefore, users need to adjust their views through the updating and iterative process in CRP to improve the consensus level of the group. Therefore, CRP is necessary to achieve decision consistency for most users. In CRP, users are not completely independent, especially due to the rapid development of instant messaging tools (such as WeChat, WhatsApp, Facebook) and social media platforms (such as Instagram, Weibo, Twitter). These platforms have significantly increased the frequency of social interactions and profoundly affected individuals' social networks and relationship structures.

[0003] However, the existing rules on the evolution of group decision-making opinions are mostly concentrated in the field of opinion dynamics, with less consideration of sociological and psychological knowledge. Although the opinion interaction rules designed in this way can reach consensus more quickly, they are difficult to reflect the actual object selection process and opinion interaction evolution rules of decision-making in CRP. Summary of the invention

[0004] Based on this, it is necessary to provide a sorting method, device and computer equipment based on multi-stage opinion interaction to address the above technical problems.

[0005] A ranking method based on multi-stage opinion interaction, the method comprising: Determine a quadruple of users in a social network and interactive learning objects of the users; wherein the quadruple includes: familiarity, preference, preference stubbornness and ranking of items to be sorted; the interactive learning objects are determined according to the similarity of users in the social network; In the current interaction round, the score contribution of the user's to-be-sorted items is calculated according to the preset learning rate and the sorting, and the aggregate score contribution of the set of interactive learning objects is calculated according to the preset learning rate, the familiarity of the interactive learning objects, and the sorting of the interactive learning objects; updating the ranking according to the score contribution and the aggregate score contribution to obtain a learned ranking; According to the preference and the preference stubbornness, updating the preference degree of the learned ranking to obtain an updated ranking; When the consistency ratio of all users' update rankings reaches a threshold, or the maximum number of interaction iterations is reached, the consistency ranking result is output.

[0006] In one embodiment, the method further includes: obtaining the user's ranking and the ranking of users to be evaluated ,in M is the total number of items to be sorted; Calculate the sort order separately and sort The normalized probability distributions P and Q of are: ; ; in, ; Calculate the probability distribution P relative to the average probability distribution KL divergence and probability distribution Q relative to the mean probability distribution KL divergence : ; ; in, ; According to KL divergence and KL divergence Calculate the JS divergence as: .

[0007] Convert JS divergence to JS distance: .

[0008] When the JS distance between the user and the user to be evaluated is less than a threshold, the user to be evaluated is determined to be an interactive learning object.

[0009] In one embodiment, the step further includes: calculating the score contribution of the user's items to be sorted according to the preset learning rate and the sorting as follows: ; in, represents the score contribution, represents the position of item i in the order, M is the total number of items in the sort, Represents the learning rate.

[0010] In one embodiment, the method further includes: calculating the aggregated score contribution of the set of interactive learning objects according to the preset learning rate, the familiarity of the interactive learning objects, and the ranking of the interactive learning objects as follows: ; in, Represents an interactive learning object The familiarity weight of , K is the total number of interactive learning objects, Represents an interactive learning object The familiarity of Represents item i for interactive learning objects ranking.

[0011] In one embodiment, the step further includes: adding the score contribution and the aggregate score contribution to obtain a comprehensive score of: .

[0012] According to the comprehensive score, the items in the ranking are arranged in descending order to obtain a learning ranking.

[0013] In one of the embodiments, the preference includes: an absolute preference and a relative preference; wherein the absolute preference represents an absolute constraint on the ordering of items in the ordering, and the relative preference represents a constraint on the relative ordering between items in the ordering.

[0014] In one of the embodiments, the preference stubbornness is initially 1, and when execution needs to be performed according to the preference stubbornness, the preference stubbornness is reduced after execution.

[0015] A sorting device based on multi-stage opinion interaction, the device comprising: An initialization module, used to determine a user's quadruple in a social network and an interactive learning object of the user; wherein the quadruple includes: familiarity, preference, preference stubbornness and ranking of items to be sorted; the interactive learning object is determined according to the similarity of users in the social network; A score calculation module, used to calculate the score contribution of the user's to-be-sorted items in the current interaction round according to a preset learning rate and the sorting, and to calculate the aggregate score contribution of the set of interactive learning objects according to a preset learning rate, the familiarity of the interactive learning objects, and the sorting of the interactive learning objects; A learning update module, used for updating the ranking according to the score contribution and the aggregate score contribution to obtain a learning ranking; A preference setting module, used for updating the preference of the learning ranking according to the preference and the preference stubbornness, to obtain an updated ranking; The sorting output module is used to output the consistent sorting results when the consistency ratio of the updated sorting of all users reaches a threshold or reaches the maximum number of interaction iterations.

[0016] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented: Determine a quadruple of users in a social network and interactive learning objects of the users; wherein the quadruple includes: familiarity, preference, preference stubbornness and ranking of items to be sorted; the interactive learning objects are determined according to the similarity of users in the social network; In the current interaction round, the score contribution of the user's to-be-sorted items is calculated according to the preset learning rate and the sorting, and the aggregate score contribution of the set of interactive learning objects is calculated according to the preset learning rate, the familiarity of the interactive learning objects, and the sorting of the interactive learning objects; updating the ranking according to the score contribution and the aggregate score contribution to obtain a learned ranking; According to the preference and the preference stubbornness, updating the preference degree of the learned ranking to obtain an updated ranking; When the consistency ratio of all users' update rankings reaches a threshold, or the maximum number of interaction iterations is reached, the consistency ranking result is output.

[0017] The above-mentioned sorting method, device and computer equipment based on multi-stage opinion interaction first determine the user's quadruple and the user's interactive learning object in the social network. Familiarity is introduced in the quadruple to reflect the user's choice and learning process at different familiarities in reaching a real consensus. At the same time, preferences and preference stubbornness are introduced. When opinions are fused, the user's preferences can be executed to ensure the rationality and consistency of the sorting results. Based on the above-mentioned initialization design, when interacting, the score of the current sorting and the score of the interactive learning object sorting are first calculated. When scoring, the concept of familiarity is introduced, so that the opinions of users with higher familiarity occupy a larger proportion in the final opinion update. Finally, the sorting is updated according to the preference and preference stubbornness, and a consistent sorting result is achieved through multiple rounds of iteration. Therefore, the above-mentioned scheme of the present application can more realistically and flexibly perform multi-user opinion fusion and sorting result output. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of a sorting method based on multi-stage viewpoint interaction in one embodiment; Figure 2 is a structural block diagram of a sorting device based on multi-stage viewpoint interaction in one embodiment; Figure 3FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0020] In one embodiment, Figure 1 As shown, a ranking method based on multi-stage opinion interaction is provided, comprising the following steps: Step 102: Determine the user's quadruple in the social network and the user's interactive learning objects.

[0021] A social network can be modeled as an undirected graph ,in represents the set of edges, Represents a collection of nodes. The edges in the undirected graph represent the connections between users, while the nodes correspond to users. The adjacency matrix of ,in: ; Indicates user and There are connections between them, and they can interact in the decision-making process.

[0022] In an undirected graph, users have four attributes, which are represented by four tuples, including familiarity, preference, preference persistence, and ranking of the items to be sorted. Specifically, users with higher familiarity are given higher weights in the interactive iteration process. Preference describes the user's inclination towards items, and preference persistence measures the degree to which the user adheres to his or her preferences. Ranking represents the user's ranking of multiple items, which is updated after each iteration.

[0023] Interactive learning objects are determined based on the similarity of users in the social network.

[0024] Step 104, in the current interaction round, the score contribution of the user's items to be sorted is calculated according to the preset learning rate and sorting, and the aggregate score contribution of the set of interactive learning objects is calculated according to the preset learning rate, the familiarity of the interactive learning objects, and the sorting of the interactive learning objects.

[0025] Step 106, updating the ranking according to the score contribution and the aggregate score contribution to obtain a learned ranking.

[0026] Step 108, updating the preference of the learned ranking according to the preference and the preference stubbornness to obtain an updated ranking.

[0027] Step 110: When the consistency ratio of update rankings of all users reaches a threshold, or reaches a maximum number of interaction iterations, a consistency ranking result is output.

[0028] In the above-mentioned sorting method based on multi-stage opinion interaction, first, the quadruple of users in the social network and the interactive learning objects of the users are determined. Familiarity is introduced in the quadruple to reflect the selection and learning process of users at different familiarities in reaching a real consensus. At the same time, preferences and preference stubbornness are introduced. When opinions are fused, the user's preferences can be executed, thereby ensuring the rationality and consistency of the sorting results. Based on the above-mentioned initialization design, when interacting, the score of the current ranking and the score of the ranking of the interactive learning object are first calculated. When scoring, the concept of familiarity is introduced, so that the opinions of users with higher familiarity occupy a larger proportion in the final opinion update. Finally, the ranking is updated according to the preference and preference stubbornness, and a consistent sorting result is achieved through multiple rounds of iteration. Therefore, the above-mentioned scheme of the present application can more realistically and flexibly perform multi-user opinion fusion and sorting result output.

[0029] In one embodiment, the interactive learning object is determined based on the similarity of users in the social network, specifically including: S1, get the user's ranking and the ranking of users to be evaluated ,in M is the total number of items to be sorted.

[0030] S2, calculate the sorting and sort The normalized probability distributions P and Q of are: ; ; in, .

[0031] S3, respectively calculate the probability distribution P relative to the average probability distribution KL divergence and probability distribution Q relative to the mean probability distribution KL divergence : ; ; in, ; S4, according to KL divergence and KL divergence Calculate the JS divergence as: .

[0032] S5, convert JS divergence to JS distance: .

[0033] S6: When the JS distance between the user and the user to be evaluated is less than a threshold, the user to be evaluated is determined to be an interactive learning object.

[0034] In this embodiment, since the JS divergence itself is a divergence measure in information theory, its value range is To convert it into a more intuitive distance metric, after taking the square root, the range of the JS distance is .

[0035] After the user initializes the quadruple, the user is placed in a preset social network. The user can interact with the neighbors who have established connections in the network. However, considering the interaction process in reality, when selecting the interaction object, the user will give priority to the consistency of the neighbor's opinion with his own opinion, and tend to communicate with users with smaller differences in opinions. Based on this, a method for selecting interactive neighbors based on label similarity is proposed. When selecting interactive neighbors, the user first evaluates the similarity distance with each neighbor according to the user ranking label similarity calculation method based on JS distance proposed in . After completing the distance calculation, the distance is compared with the preset threshold similar_threshold. Those neighbor users whose JS distance with themselves is less than the threshold similar_threshold are determined as interactive learning objects, and the selected users among these neighbors are stored in the user person_similar set.

[0036] After determining the interactive object set user person_similar, the user's familiarity label needs to be considered at this time. The familiarity label F is used to measure the user's familiarity with a specific item or field, and is a key factor affecting the weight of the neighbor's opinion being learned. The reason for introducing familiarity is that users have a deeper understanding and more reliable judgment in their familiar fields, so their opinions should contribute more to consensus. Familiarity F is usually evaluated and determined through data in multiple dimensions, including the user's historical performance, professional background, number of projects participated in, and frequency of interaction with specific items. For example, indicators such as the number of papers published by a user in a related field, the number of research projects participated in, and the degree of activity in the user's social network can all be used to quantify their familiarity. In addition, familiarity can also be supplemented by methods such as questionnaires or user self-evaluation to ensure that it can accurately reflect the user's actual familiarity. In this embodiment, the familiarity label F will be used to weight the neighbors' opinions so that the opinions of users with higher familiarity occupy a larger proportion in the final opinion update, thereby improving the accuracy of opinion fusion and the efficiency of consensus formation.

[0037] In addition, when users interact with each other, there is a certain learning ratio and retention part between their original opinions and the opinions obtained from their neighbors. Therefore, this study introduces the learning rate To reflect the degree to which the user adopts the neighbor's opinions during the opinion update process. Learning rate The weight distribution of users when fusing their own opinions with those of their neighbors is controlled, so that users can dynamically adjust their acceptance of new information. , can balance the relationship between users retaining their original opinions and accepting new opinions from their neighbors, thereby improving the flexibility and accuracy of opinion updating.

[0038] The user's own ranking list is recorded as ,in represents the item number of the user's k-th ranking. The set of similar neighbors (interactive learning objects) is recorded as , where K is the total number of similar neighbors and each neighbor With familiarity , and its ranking list is recorded as ,in Indicates neighbors The number of the item ranked at the kth position. The learning rate is It represents the proportion of neighbor opinions in the final opinion learned by the user during the interaction process.

[0039] Specifically, the scoring contribution is calculated as follows: The score contribution of the user's items to be sorted is calculated based on the preset learning rate and sorting: ; in, represents the score contribution, represents the position of item i in the above sorting, M is the total number of items in the sort.

[0040] The aggregate score contribution is calculated as: The aggregate score contribution of the interactive learning object set is calculated based on the preset learning rate, the familiarity of the interactive learning object, and the ranking of the interactive learning objects: ; in, Represents an interactive learning object The familiarity weight of , K is the total number of interactive learning objects, Represents an interactive learning object The familiarity of Represents item i for interactive learning objects ranking.

[0041] When merging, the score contribution and the aggregate score contribution are added together to get the comprehensive score: ; The comprehensive score is specifically expanded as follows: According to the comprehensive score, the items in the sort are arranged in descending order to obtain the learning sort, which is as follows: , ; in .

[0042] In one embodiment, the preference includes: an absolute preference and a relative preference; wherein the absolute preference represents an absolute constraint on the order of items in the order, and the relative preference represents a constraint on the relative order between items in the order.

[0043] After the user combines his initial ranking with neighbor learning to obtain a new ranking, the user's preference attributes must be considered comprehensively. As mentioned above, the preference for the user to the item is divided into two forms: the first is absolute preference, that is, the user prefers that an item must be in the top C% of the ranking; the second is relative preference, that is, the user believes that item i must be ranked before item j. Specifically, absolute preference requires an item to be in the top C% of the total ranking. For example, in the case of five items, if a user prefers item 1 to be in the top 60%, the ranking of item 1 is limited to 1st, 2nd, and 3rd. Relative preference specifies the ranking order between items. For example, a user believes that item A must be ranked before item B.

[0044] In another embodiment, the initial value of the preference stubbornness B is set to 1, indicating that the user fully adheres to his or her preference in the initial stage. During the iteration process, when the new sorting result does not meet the user's preference requirements, preference enforcement needs to be performed and the preference stubbornness is adjusted accordingly. Preference stubbornness B, as a probability value, determines the way preference enforcement is performed. Initially, B=1, that is, preference enforcement is deterministic. Whenever preference enforcement is required, the user enforces it with the probability of preference stubbornness B, and reduces B according to a predetermined attenuation mechanism after each execution, thereby gradually transforming preference enforcement into probabilistic execution.

[0045] For users with absolute preferences, if the new ranking obtained after learning does not meet the preference requirements, the preferred items need to be randomly placed in the top C% of the ranking. For users with relative preferences, if the new ranking does not meet the preference requirements, the positions of the two items in the preference order need to be swapped to ensure that item i is ranked before item j.

[0046] Through the above mechanism, not only can the user's views be dynamically adjusted to integrate the opinions of neighbors, but the user's preferences can also be enforced when necessary, thereby ensuring the rationality and consistency of the sorting results. At the same time, as the stubbornness of preferences gradually decreases, the user's enforcement of preferences will gradually turn into probabilistic enforcement, which enhances the flexibility and adaptability of the model in the process of fusion of diverse opinions.

[0047] In the above way, in each round of iteration, the system traverses each user in turn, performs the steps of opinion update and ranking adjustment, and then updates its ranking attributes. Specifically, each user first generates a new ranking list based on the aggregated score and preference correction, and then checks whether the new ranking meets their preference requirements. If not, the preference is enforced and the preference stubbornness is adjusted. The entire iterative process continues until any of the following termination conditions is met: the consistency ratio of all users' ranking labels in the population reaches a preset threshold, or the predetermined maximum number of iterations is reached Through multiple rounds of iterations, opinions in the user social network gradually converge and eventually form a consensus conclusion.

[0048] In summary, this application can be applied in the following application scenarios. A megacity needs to determine the annual investment priorities of five types of new energy infrastructure. The five types of new energy infrastructure are: distributed photovoltaic power grid, electric vehicle fast charging network, hydrogen energy preparation center, intelligent energy storage system, and biomass power station. The five types of new energy infrastructure focus on environmental protection orientation, people's livelihood needs, technology foresight, and grid stability, respectively.

[0049] Nine experts used Indicates that five types of new energy infrastructure Indicates that the expert's familiarity is generated randomly , preference type and specific preferences , initial sort , preference stubbornness All are initialized to 1. The randomly initialized parameters in this example are shown in Table 1.

[0050] Table 1 Parameter design and iterative results

[0051] 9 experts are distributed in a two-dimensional On the grid, each expert needs to rank the five types of new energy infrastructure, reflecting his or her individual preferences and the adjustment of opinions influenced by his or her neighbors. The initial value is 1.0. Each time a preference is enforced, the stubbornness Decrease by a decay factor of 0.9. The initial ranking list of each expert It is generated randomly and adjusted according to its preferences during the initialization phase to meet its inherent preferences. Set to 0.9, the maximum number of iterations Set to 20 times, similarity threshold Set to 0.8, the learning rate The value is set to 0.9. Then the embodiment of the present application is adopted, and finally, after multiple rounds of discussion, information integration and feedback adjustment, the ranking result is output.

[0052] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0053] In one embodiment, Figure 2 As shown, a ranking device based on multi-stage opinion interaction is provided, comprising: an initialization module 202, a score calculation module 204, a learning update module 206, a preference setting module 208 and a ranking output module 210, wherein: Initialization module 202, used to determine the four-tuple of users in the social network and the interactive learning objects of the users; wherein the four-tuple includes: familiarity, preference, preference stubbornness and ranking of the items to be sorted; the interactive learning objects are determined according to the similarity of users in the social network; The score calculation module 204 is used to calculate the score contribution of the user's to-be-sorted items in the current interaction round according to the preset learning rate and the sorting, and calculate the aggregate score contribution of the set of interactive learning objects according to the preset learning rate, the familiarity of the interactive learning objects, and the sorting of the interactive learning objects; A learning update module 206, configured to update the ranking according to the score contribution and the aggregate score contribution to obtain a learning ranking; A preference setting module 208, configured to update the learning ranking according to the preference and the preference stubbornness, to obtain an updated ranking; The sorting output module 210 is used to output a consistent sorting result when the consistency ratio of the updated sorting of all users reaches a threshold or reaches a maximum number of interaction iterations.

[0054] In one embodiment, the initialization module 202 is also used to obtain the user's ranking and the ranking of users to be evaluated ,in M is the total number of items to be sorted; Calculate the sort order separately and sort The normalized probability distributions P and Q of are: ; ; in, ; Calculate the probability distribution P relative to the average probability distribution KL divergence and probability distribution Q relative to the mean probability distribution KL divergence : ; ; in, ; According to KL divergence and KL divergence Calculate the JS divergence as: ; Convert JS divergence to JS distance: .

[0055] When the JS distance between the user and the user to be evaluated is less than a threshold, the user to be evaluated is determined to be an interactive learning object.

[0056] In one embodiment, the score calculation module 204 is further configured to calculate the score contribution of the user's to-be-sorted items according to a preset learning rate and the sorting as follows: ; in, represents the score contribution, represents the position of item i in the above sorting, M is the total number of items in the sort, Represents the learning rate.

[0057] In one embodiment, the score calculation module 204 is further configured to calculate the aggregate score contribution of the set of interactive learning objects according to the preset learning rate, the familiarity of the interactive learning objects, and the ranking of the interactive learning objects as follows: ; in, Represents an interactive learning object The familiarity weight of , K is the total number of interactive learning objects, Represents an interactive learning object The familiarity of Represents item i for interactive learning objects ranking.

[0058] In one embodiment, the score calculation module 204 is further configured to add the score contribution and the aggregate score contribution to obtain a comprehensive score: .

[0059] According to the comprehensive score, the items in the ranking are arranged in descending order to obtain a learning ranking.

[0060] In one of the embodiments, the preference includes: an absolute preference and a relative preference; wherein the absolute preference represents an absolute constraint on the ordering of items in the ordering, and the relative preference represents a constraint on the relative ordering between items in the ordering.

[0061] In one embodiment, the preference stubbornness is initially 1, and when execution is required according to the preference stubbornness, the preference stubbornness decreases after execution.

[0062] For the specific definition of the sorting device based on multi-stage viewpoint interaction, please refer to the definition of the sorting method based on multi-stage viewpoint interaction above, which will not be repeated here. Each module in the above-mentioned sorting device based on multi-stage viewpoint interaction can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0063] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a sorting method based on multi-stage viewpoint interaction is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0064] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0065] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method in the above embodiment when executing the computer program.

[0066] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0067] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0068] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A ranking method based on multi-stage opinion interaction, characterized in that: The method comprises: Determine a quadruple of users in a social network and interactive learning objects of the users; wherein the quadruple includes: familiarity, preference, preference stubbornness and ranking of items to be sorted; the interactive learning objects are determined according to the similarity of users in the social network; In the current interaction round, the score contribution of the user's to-be-sorted items is calculated according to the preset learning rate and the sorting, and the aggregate score contribution of the set of interactive learning objects is calculated according to the preset learning rate, the familiarity of the interactive learning objects, and the sorting of the interactive learning objects; updating the ranking according to the score contribution and the aggregate score contribution to obtain a learned ranking; According to the preference and the preference stubbornness, updating the preference degree of the learned ranking to obtain an updated ranking; When the consistency ratio of all users' update rankings reaches a threshold, or the maximum number of interaction iterations is reached, the consistency ranking result is output.

2. The method according to claim 1, characterized in that Identify the user's interactive learning objects, including: Get the user's ranking and the ranking of users to be evaluated ,in M is the total number of items to be sorted; Calculate the sort order separately and sort The normalized probability distributions P and Q of are: in, ; Calculate the probability distribution P relative to the average probability distribution KL divergence and probability distribution Q relative to the mean probability distribution KL divergence : in, ; According to KL divergence and KL divergence Calculate the JS divergence as: Convert JS divergence to JS distance: When the JS distance between the user and the user to be evaluated is less than a threshold, the user to be evaluated is determined to be an interactive learning object.

3. The method according to claim 1, characterized in that: The score contribution of the user's items to be sorted is calculated according to the preset learning rate and the sorting, including: The score contribution of the user's items to be sorted is calculated based on the preset learning rate and the sorting: in, represents the score contribution, represents the position of item i in the order, M is the total number of items in the sort, Represents the learning rate.

4. The method according to claim 3, characterized in that The aggregate score contribution of the interactive learning object set is calculated according to the preset learning rate, the familiarity of the interactive learning object, and the ranking of the interactive learning object, including: The aggregate score contribution of the interactive learning object set is calculated based on the preset learning rate, the familiarity of the interactive learning object, and the ranking of the interactive learning objects: in, Represents an interactive learning object The familiarity weight of , K is the total number of interactive learning objects, Represents an interactive learning object The familiarity of Represents item i for interactive learning objects ranking.

5. The method according to claim 4, characterized in that The ranking is updated according to the score contribution and the aggregate score contribution to obtain a learning ranking, including: The score contribution and the aggregate score contribution are added together to obtain a comprehensive score of: According to the comprehensive score, the items in the ranking are arranged in descending order to obtain a learning ranking.

6. The method according to any one of claims 1 to 5, characterized in that: The preference includes: absolute preference and relative preference; wherein the absolute preference represents an absolute constraint on the order of items in the order, and the relative preference represents a constraint on the relative order between items in the order.

7. The method according to any one of claims 1 to 5, characterized in that: The preference stubbornness is initially 1. When execution needs to be performed according to the preference stubbornness, the preference stubbornness decreases after execution.

8. A sorting device based on multi-stage opinion interaction, characterized in that: The device comprises: An initialization module, used to determine a user's quadruple in a social network and an interactive learning object of the user; wherein the quadruple includes: familiarity, preference, preference stubbornness and ranking of items to be sorted; the interactive learning object is determined according to the similarity of users in the social network; A score calculation module, used to calculate the score contribution of the user's to-be-sorted items in the current interaction round according to a preset learning rate and the sorting, and to calculate the aggregate score contribution of the set of interactive learning objects according to a preset learning rate, the familiarity of the interactive learning objects, and the sorting of the interactive learning objects; A learning update module, used for updating the ranking according to the score contribution and the aggregate score contribution to obtain a learning ranking; A preference setting module, used for updating the preference of the learning ranking according to the preference and the preference stubbornness, to obtain an updated ranking; The sorting output module is used to output the consistent sorting results when the consistency ratio of the updated sorting of all users reaches a threshold or reaches the maximum number of interaction iterations.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images