A ranking information recommendation method based on distribution entropy

Abstract

This invention discloses a ranking information recommendation method based on distribution entropy, belonging to the field of recommendation strategy technology, including the following steps: S1, obtaining ranking information: each ranking entity ranks all or a portion of the evaluated objects separately; S2, networking the ranking data: constructing an undirected weighted network; S3, the ranking information recommendation method based on distribution entropy: introducing global edge weight distribution entropy, quantifying the uniformity of edge weight distribution, and designing a recommendation scoring mechanism based on the quantification results. This invention, employing the above-mentioned ranking information recommendation method based on distribution entropy, not only provides an intuitive and scientific means to objectively evaluate the quality of ranking data, but also, through a low-computational-cost, high-generalization-ability mechanism, suppresses the Matthew effect in the evaluation system from the source, ensuring that the final obtained ranking set has extremely high coverage and fairness, providing solid data support for scientific decision-making in the field of computer and auxiliary equipment repair.

Description

Technical Field

[0001] This invention relates to the field of recommendation strategy technology, and in particular to a ranking information recommendation method based on distribution entropy. Background Technology

[0002] With the continuous progress of human society, the rapid increase in information volume necessitates the integration of ranking or evaluation results from multiple sources when facing complex decisions. In scenarios such as academic evaluation, product recommendation, search result fusion, computer and auxiliary equipment repair, and multi-indicator comprehensive evaluation, different sources or methods often produce their own ranking results. How to rationally integrate these heterogeneous and potentially conflicting rankings to obtain a comprehensive ranking that fully reflects overall preferences has become an important issue. Ranking aggregation has emerged to address this need. It is a technique for integrating ranking results from multiple sources or methods. Its core idea lies in fusing the preference information contained in different rankings to generate a final ranking that comprehensively reflects overall opinion. This method effectively improves the objectivity and stability of the integrated result by seeking optimal consistency among diverse, even conflicting, rankings.

[0003] To obtain the most effective and reliable aggregation results in different application scenarios, academia and industry have proposed various ranking aggregation methods, aiming to seek optimal consistency among multi-source ranking information. However, the effectiveness of ranking aggregation depends not only on the algorithm itself, but also to a large extent on the quality of the input ranking data. In other words, no matter how sophisticated the aggregation method is designed, if the input data is biased or incomplete, the final aggregation result may still be distorted, lack stability, or even fail to accurately reflect the overall preferences.

[0004] In past practice, the number of input rankings has often been considered the most intuitive and easily calculable indicator of ranking data quality, and is therefore frequently used to assess the representativeness and completeness of the data. Generally speaking, a larger number of rankings indicates a richer information source, which can reduce the bias caused by a single source to some extent. However, this indicator cannot quantify the distribution characteristics of rankings. If ranking information is largely concentrated on a few popular objects, even with a large number of rankings, it is difficult to obtain accurate and impartial aggregation results on a global scale. Therefore, in the evaluation and decision-making process related to information recommendation and computer and assistive device repair, it is necessary to proactively avoid uneven distribution of ranking information. Summary of the Invention

[0005] The purpose of this invention is to provide a ranking information recommendation method based on distribution entropy, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, this invention provides a ranking information recommendation method based on distribution entropy, comprising the following steps: S1. Obtain ranking information: For multiple rankers and multiple evaluated objects, each ranker ranks all or some of the evaluated objects separately, collects the corresponding ranking data, and obtains the ranking information. S2. Ranking data networking: Map each evaluated object to a network node, and map the pairwise comparison information between evaluated objects to undirected weighted edges between nodes to construct an undirected weighted network; S3. Ranking information recommendation method based on distribution entropy: In the constructed undirected weighted network, global edge weight distribution entropy is introduced to quantify the uniformity of edge weight distribution, and a recommendation scoring mechanism is designed based on the quantification results to achieve accurate recommendation of ranking information.

[0007] Preferably, the ranking data in S1 is specifically represented as follows: ; in, For ranking, Indicates the ranking For the evaluated subjects The ranking Indicates the ranking No object Ranking, Representation Object and They have the same ranking; i For the ranking index, 1≤ i ≤ m , m The total number of ranked individuals; j , k For the index of the evaluated object, 1≤ j , k ≤ n , n This represents the total number of individuals being evaluated.

[0008] Preferably, the specific steps of S2 are as follows: S21. Use an undirected weighted graph to represent the ranking information among all ranked objects; S22. The edge weight between any two nodes is defined as the number of rank values ​​in all input rank values ​​that simultaneously contain both nodes.

[0009] Preferably, the specific steps of S3 are as follows: S31. For multiple input rankings, sum all edge weights in the undirected weighted network to obtain the total information. S32. Calculate the proportion of ranking information between any two objects to the total information; S33. Calculate the global edge weight distribution entropy based on the edge weight probabilities of each node pair in an undirected weighted network; S34. Based on the information entropy theory, design a recommendation scoring mechanism by maximizing the global edge weight distribution entropy, and calculate the recommendation priority score.

[0010] Preferably, the calculation formula for S32 is as follows: ; in, For specific gravity, Let be the edge weight between any two nodes. For the total amount of information, and .

[0011] Preferably, the formula for calculating the global edge weight distribution entropy in S33 is as follows: ; in, This represents the global edge weight distribution entropy.

[0012] Preferably, the formula for calculating the recommendation priority score in S34 is as follows: ; in, Score the recommendation priority. This is the edge entropy gain.

[0013] Therefore, the ranking information recommendation method based on distribution entropy described above, as used in this invention, has the following beneficial effects: This invention not only provides an intuitive and scientific method to objectively evaluate the quality of ranking data, but also suppresses the "Matthew effect" in the evaluation system from the source through a mechanism with low computational cost and high generalization ability, ensuring that the final ranking set has extremely high coverage and fairness, and providing solid data support for subsequent scientific decision-making.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating an embodiment of a ranking information recommendation method based on distribution entropy according to the present invention. Figure 2 This is ranking data in an information distribution set, representing an embodiment of a ranking information recommendation method based on distribution entropy according to the present invention. Detailed Implementation

[0016] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Please see Figures 1-2 This invention provides a ranking information recommendation method based on distribution entropy. The method models the recommendation environment as an undirected ranking network, where nodes represent objects to be evaluated, edges between nodes represent the association of object pairs appearing together in the ranking sequence, and edge weights represent the total number of samples containing the ranking information of that pair of nodes.

[0019] To quantify the uniformity of information, this invention defines a global distribution probability: the weight between each pair of nodes is divided by the total weight of all ranking information in the network. Based on this, the distribution entropy of the entire graph is calculated to measure the disorder and balance of information within the system. According to the maximum entropy principle, the global entropy reaches its maximum value when the ranking information weights of all node pairs tend to be consistent, at which point the system's representativeness and resistance to bias are strongest. Figure 1 As shown, the method includes the following steps: S1. Obtain ranking.

[0020] Assume there is There are rankings, with Each participant can be evaluated, and each ranking person can... Rank some or all of the evaluated entities. Record users. The submitted rankings are: ; in, Indicates the ranking For the evaluated The ranking. This indicates the ranker No object Ranking. If in ranking medium object and If they have the same ranking, then .i For the ranking index, 1≤ i ≤ m , m The total number of ranked individuals; j , k For the index of the evaluated object, 1≤ j , k ≤ n , n This represents the total number of individuals being evaluated.

[0021] S2, Ranking data networking.

[0022] Each input ranking essentially contains pairwise comparisons between objects, while complex networks are modeled around nodes and their relationships. Based on this correspondence, the ranking problem can be mapped to network space: each ranked object is considered a node in the network, and the pairwise comparisons between objects are mapped to undirected weighted edges between nodes. Through this mapping, the originally linear or list-like ranking data can be transformed into a structured network representation, allowing the comparison relationships between objects to be presented and analyzed more intuitively at the topological level. Therefore, different structural features of complex networks can also reflect the distribution characteristics of different ranking data.

[0023] Specifically, this invention can use an undirected weighted graph. This represents the ranking information among all the evaluated objects in the ranking. Among them, This represents a collection of nodes, where each node corresponds to an object. This represents a set of undirected edges used to indicate pairwise comparisons between objects. For any two nodes... and Its edge weight is denoted as , This indicates that among all input rankings, including and The number of rankings that meet the requirements .

[0024] S3. Ranking information recommendation method based on distribution entropy.

[0025] In an undirected weighted network constructed from ranking information, the distribution characteristics of the ranking information are directly reflected in the distribution of edge weights between node pairs. When the ranking information is highly concentrated in a few object pairs, the edge weight distribution in the network will exhibit extreme imbalance; conversely, when the ranking information is evenly distributed among all possible node pairs, the network has the highest information coverage and representativeness. Therefore, global edge weight distribution entropy can be introduced to quantify the uniformity of this distribution, and recommendation mechanisms can be designed accordingly.

[0026] for Given rank inputs, the sum of all edge weights in the network (i.e., total information) is defined as . Based on this, any two objects and The proportion of ranking information between them to the total amount of information. for: ; Next, this invention calculates the global edge weight distribution entropy of the network based on the edge weight probability between each pair of nodes in the network. : ; In the recommendation process, to ensure that ranking information is distributed as evenly as possible, the goal of the recommendation system is to maximize the global edge weight distribution entropy. According to information entropy theory, a probability distribution is established if and only if all pairs of nodes have equal probability proportions. When they tend to be consistent, It has reached its maximum value.

[0027] Accordingly, this invention designs a recommendation scoring mechanism based on entropy increase potential. For candidate objects... and Its recommendation priority score The definition is as follows: ; This mechanism identifies "information-poor" node pairs with low weights in the network and assigns them higher recommendation weights. As recommendation information is continuously input, the originally scarce ranking information in the network is compensated for. This leads to a steady improvement, thereby ensuring the uniformity and representativeness of the input ranking data across the entire global scale.

[0028] Therefore, the recommendation mechanism designed in this invention no longer relies solely on traditional relevance prediction, but instead aims to maximize the overall network distribution entropy. The system dynamically monitors the weight distribution in the network, identifies information-poor edges with low weights, and prioritizes recommending these scarce object combinations to users. This method, from a mechanism design perspective, ensures a highly uniform distribution of ranking data across the topology, thus laying a data foundation for achieving more representative and fairer ranking aggregation.

[0029] Example 1 Suppose there exists a set of partial rankings for four subjects, provided by three ranking experts, such as... Figure 2 As shown, and assuming the actual ranking is .

[0030] After converting it to a network, it can be seen that the current ranking information is all concentrated on the object. a and object bHowever, the lack of ranking information for other objects will result in the inability to obtain effective aggregation results.

[0031] When collecting more ranking data, first calculate the priority score for each pair of nodes. as follows: ; ; ; ; ; ; It can be seen that, in addition to a and b Apart from that, all others received higher priority scores, making it easier to obtain scores during the subsequent information gathering process.

[0032] The current approach calculates the probability distribution of all node pairs in the entire network, aiming for a globally uniform distribution. It doesn't necessarily require all edges to have the same thickness; instead, it aims for the edge weights connecting each node to other nodes to be as uniform as possible.

[0033] Therefore, the above-mentioned ranking information recommendation method based on distribution entropy not only provides an intuitive and scientific means to objectively evaluate the quality of ranking data, but also suppresses the "Matthew effect" in the evaluation system from the source through a mechanism with low computational cost and high generalization ability, ensuring that the final ranking set has extremely high coverage and fairness, and providing solid data support for subsequent scientific decision-making.

[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A ranking information recommendation method based on distribution entropy, characterized in that, Includes the following steps: S1. Obtain ranking information: For multiple rankers and multiple evaluated objects, each ranker ranks all or some of the evaluated objects separately, collects the corresponding ranking data, and obtains the ranking information. S2. Ranking data networking: Map each evaluated object to a network node, and map the pairwise comparison information between evaluated objects to undirected weighted edges between nodes to construct an undirected weighted network; S3. Ranking information recommendation based on distribution entropy: In the constructed undirected weighted network, global edge weight distribution entropy is introduced to quantify the uniformity of edge weight distribution, and a recommendation scoring mechanism is designed based on the quantification results to achieve accurate recommendation of ranking information. The specific expression for the ranking data in S1 is as follows: ； in, For ranking, Indicates the ranking For the evaluated subjects The ranking Indicates the ranking No evaluation recipients were given Ranking, Indicates the person being evaluated and They have the same ranking; i For the ranking index, 1≤ i ≤ m , m The total number of ranked individuals; j , k For the index of the evaluated object, 1≤ j ≤ n ,1≤ k ≤ n , n The total number of respondents; The specific steps of S2 are as follows: S21. Use an undirected weighted graph to represent the ranking information among all ranked objects; S22. The edge weight between any two nodes is defined as the number of rank values ​​in all input rank values ​​that simultaneously contain both nodes. The specific steps of S3 are as follows: S31. For multiple input rankings, sum all edge weights in the undirected weighted network to obtain the total information. S32. Calculate the proportion of ranking information between any two evaluated objects to the total information; S33. Calculate the global edge weight distribution entropy based on the edge weight probabilities of each node pair in an undirected weighted network; S34. Based on the information entropy theory, design a recommendation scoring mechanism by maximizing the global edge weight distribution entropy and calculate the recommendation priority score. The calculation formula for S32 is as follows: ； in, For specific gravity, Let be the edge weight between any two nodes. For the total amount of information, and ,in, i < j ; The formula for calculating the global edge weight distribution entropy in S33 is as follows: ， i ＜ j ； in, This represents the global edge weight distribution entropy.

2. The ranking information recommendation method based on distribution entropy according to claim 1, characterized in that, The formula for calculating the recommendation priority score in S34 is as follows: ； in, Score the recommendation priority. This is the edge entropy gain.

Images