A density clustering processing method, device and medium that does not rely on historical information

Abstract

This invention discloses a density clustering processing method, device, and medium that does not rely on historical information. The method includes: acquiring measurement parameters corresponding to a user; employing a noisy density-based clustering method based on the measurement parameters; calculating the set of samples connected by the highest density based on density reachability relationships; and using the obtained sample set as the clustering result. The measurement parameters are one or more combinations of regional attributes, registered capital of enterprises, establishment time, industry classification, capital type, enterprise type, shareholder information, number of employees, and enterprise movable property mortgage information; acquiring bidding information to be analyzed, including users participating in the current bidding activity; and determining whether there are any correlations among the users participating in the current bidding activity based on the clustering result. This invention's solution can promptly detect potential bid rigging and collusion even without historical bidding data through a density clustering algorithm, ensuring the fairness of the bidding process.

Description

[0001] This application is a divisional application of the invention patent application filed on January 20, 2021, with application number 202110072560.8 and invention title "A method, device and storage medium for processing bidding information based on density clustering". Technical Field

[0002] This invention relates to the field of computer technology, and specifically to a density clustering processing method, device, and storage medium that does not rely on historical information. Background Technology

[0003] The bidding and tendering system, which originated in the 1980s, has played a crucial role in improving project quality, strengthening safety management, and controlling project costs. However, over the past 30 years, problems such as fraudulent bidding, collusion, and deception in winning bids have consistently hindered its further development. Since 2012, "digitalization" and "networking" have become the development trends of bidding and tendering. Currently, massive amounts of electronic bidding and tendering data are generated daily in the national public resource transaction process, providing an important prerequisite for applying "big data" technology to bidding and tendering. By introducing "big data" technology to effectively collect, organize, process, and utilize the various data generated during the bidding and tendering process, and fully leveraging the fundamental, predictive, and guiding role of "data" in strengthening the supervision of bidding and tendering activities, the healthy and sustainable development of the bidding and tendering market can be promoted. However, the application of big data technology in the bidding and tendering field is not yet mature and still has many shortcomings, such as the untimely and inaccurate detection of bid rigging and collusion when the back-end server performs big data analysis. Summary of the Invention

[0004] Due to the aforementioned problems with existing methods, this invention proposes a density clustering processing method, device, and storage medium that does not rely on historical information.

[0005] In a first aspect, embodiments of the present invention propose a density clustering processing method that does not rely on historical information, including:

[0006] The measurement parameters corresponding to the user are obtained. Based on the measurement parameters, a density-based clustering method with noise is used to calculate the sample set with the highest density based on the density reachability relationship. The obtained sample set is used as the clustering result. The measurement parameters are one or more combinations of regional attributes, enterprise registered capital, establishment time, industry classification, capital type, enterprise type, shareholder information, number of employees, and enterprise movable property mortgage information.

[0007] Obtain the bidding information to be analyzed, including the users participating in the current bidding activity;

[0008] Based on the clustering results, determine whether there are any relationships among the users participating in the current bidding activity.

[0009] Optionally, the step of obtaining measurement parameters corresponding to user identity features, and performing cluster analysis on users using a density-based clustering algorithm based on the measurement parameters to obtain clustering results, specifically includes:

[0010] The user's measurement parameters are obtained as samples. If the number of samples is greater than a set threshold, the nearest neighbor is searched according to the K-space tree or ball tree. The sample distance in the clustering algorithm is determined by Euclidean distance based on the nearest neighbor. The user is then clustered according to the sample distance and the measurement parameters to obtain the clustering result.

[0011] Optionally, it also includes:

[0012] The clustering result includes a target set and noise points. The noise point data is corrected based on a predefined data set. The predefined data set includes known association data, which includes first association data to characterize the relationship between two users and second association data to characterize the absence of a relationship between two users.

[0013] The step of correcting the noise point data based on a predefined data set specifically involves: setting a default transfer step size and updating the target set according to the default transfer step size; the transfer step size is the number of transfer samples between two density-reachable samples in a density-reachable relationship.

[0014] Optionally, the propagation step size is increased based on the first association data in the predefined data set, and the target set is updated based on the propagation step size.

[0015] Optionally, the propagation step size can be reduced based on the second association data in the predefined data set, and the target set can be updated based on the changed propagation step size.

[0016] Secondly, embodiments of the present invention also provide an electronic device, comprising:

[0017] At least one processor; and

[0018] At least one memory communicatively connected to the processor, wherein:

[0019] The memory stores program instructions that can be executed by the processor, and the processor can execute the aforementioned density clustering processing method that does not rely on historical information by calling the program instructions.

[0020] Thirdly, embodiments of the present invention also propose a non-transitory computer-readable storage medium storing a computer program that causes the computer to execute the aforementioned density clustering processing method that does not rely on historical information.

[0021] As can be seen from the above technical solution, the embodiments of the present invention, by combining a density clustering processing method that does not rely on historical information, avoid the problem of traditional community detection algorithms requiring historical bidding data of enterprises. It can promptly detect possible bid rigging and collusion even without historical bidding data through density-based clustering algorithms, thus ensuring the fairness of the bidding process. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 A schematic diagram illustrating the process of a community detection algorithm provided for an existing solution;

[0024] Figure 2 This is a schematic diagram of the bidding information processing steps based on density clustering.

[0025] Figure 3 This is a schematic diagram of a sample clustering process provided in an embodiment of the present invention;

[0026] Figure 4 This is a flowchart illustrating the steps of correcting clustering results for noisy point data based on a user-predefined dataset, according to an embodiment of the present invention.

[0027] Figure 5 This is a schematic diagram of sample clustering before clustering result correction according to an embodiment of the present invention;

[0028] Figure 6 This is a schematic diagram of sample clustering after the clustering results have been corrected according to an embodiment of the present invention;

[0029] Figure 7 This is a schematic diagram of another sample clustering after the clustering result is corrected according to an embodiment of the present invention;

[0030] Figure 8 This is a flowchart illustrating the steps of updating a target set based on data in a predefined data set according to an embodiment of the present invention.

[0031] Figure 9 This is a schematic diagram of sample clustering before clustering result correction according to an embodiment of the present invention;

[0032] Figure 10 This is a schematic diagram illustrating the steps of updating the target set according to the transmission step size in one embodiment of the present invention;

[0033] Figure 11 This is a block diagram of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0034] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. The following examples are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0035] First, the application of big data technology in the bidding and tendering field is not yet mature. For example, in detecting illegal activities such as bid rigging and collusion, the main method is to analyze historical bidding data to discover such information. Because participants in bid rigging and collusion naturally possess group attributes, traditional solutions often employ group detection algorithms. The main idea is to analyze historical data to determine whether participating companies belong to the same group, i.e., whether there is a connection. (Refer to...) Figure 1 The algorithm for detecting communities, as shown, is essentially based on graph clustering, where nodes are connected by edges. In bidding information processing, this can be visualized as establishing a connection (edge) between two companies (nodes), and then using the graph of nodes and edges as input for community detection analysis. Specifically, the algorithm uses the number of times companies participate in bidding together as the weight between two vertices, resulting in a weighted complex network of bidding relationships between companies (e.g., this network represents the frequency of companies participating in bidding together).

[0036] However, the problem with community-based bid-rigging detection algorithms is that their input data heavily relies on historical bidding information. Typically, frequent joint bidding by companies is required to detect potential strong connections between them. This makes it difficult to detect instances of bid-rigging that are infrequent. Furthermore, companies can engage in bid-rigging by constantly changing identities or using third parties. Even if companies frequently participate in bidding together, by the time detection occurs, they have already engaged in multiple illegal activities, leaving only post-event remedial measures and no preventative measures. This is a common problem with current big data detection methods for bid-rigging.

[0037] To address the aforementioned problems, this embodiment provides a bidding information processing method based on density clustering, referring to... Figure 2 A flowchart illustrating the method, which includes:

[0038] S1. Obtain the measurement parameters corresponding to the user, and perform cluster analysis on the user using a density-based clustering algorithm based on the measurement parameters to obtain the clustering results; the above-mentioned user refers to all bidding users in the sample database;

[0039] S2. Obtain the bidding information to be analyzed, which includes the participating users in the current bidding activity, where participating users refer to bidding users;

[0040] S3. Determine whether there are any relationships among the users participating in the current bidding activity based on the clustering results.

[0041] The measurement parameters include regional attributes, registered capital, establishment time, industry classification, capital type, enterprise type, shareholder information, number of employees, and / or enterprise movable property mortgage information.

[0042] The measurement parameters include various information related to the bidding users (i.e., the bidding companies), and are not limited to those described above. However, it is worth noting that the above measurement parameters do not mention bidding data, i.e., historical bidding information mentioned above. This embodiment uses a density-based clustering algorithm, which does not rely on historical bidding information; therefore, historical bidding information is not necessarily required in the measurement parameters.

[0043] The clustering result is a target set of users that meets the expected criteria after analyzing bidding users based on measurement parameters. Specifically, it is a set of users with high relevance in bidding activities, obtained by analyzing user-specific measurement parameters such as regional attributes, registered capital, and establishment time. The user characteristics of the target set are the common relevant characteristics of the users in the target set.

[0044] For example, bidding activities involve a wide variety of companies, each with its own characteristics. Without access to relevant user (bidding company) information, it's impossible to accurately analyze the correlation between companies. This embodiment can utilize, for example, a credit reporting agency's data platform to obtain basic user data, such as the aforementioned regional attributes, registered capital, and establishment date.

[0045] Finally, based on the clustering results, it is determined whether there are two users in the same target set among the participants in the current bidding activity. If so, there are users with a high degree of correlation in the current bidding activity, indicating that there is likely an anomaly.

[0046] Furthermore, based on the above method embodiments, the density-based clustering algorithm is a density-based clustering method with noise (DBSCAN), which calculates the set of samples connected by the highest density according to the density reachability relationship, and the obtained set of samples is used as the clustering result.

[0047] Specifically, by analyzing the measurement parameters, users can be divided into different groups, each group corresponding to a set, which facilitates further analysis of users in each set.

[0048] The density-based clustering algorithm with noise (DBSCAN) defines a cluster as the largest set of density-connected points. It can divide regions with sufficiently high density into clusters and can discover clusters of arbitrary shapes in noisy spatial databases. DBSCAN describes the compactness of a sample set based on a set of neighborhoods, reflecting the relevance of different regional attributes, enterprise registered capital, and other dimensions to bidding information. The parameters (∈, MinPts) describe the compactness of the sample distribution within the neighborhood, for example, whether the relevance of bidding information affecting a specified dimension exhibits significant clustering (higher clustering indicates greater parameter relevance). Here, ∈ describes the neighborhood distance threshold for a given sample, and MinPts describes the threshold for the number of samples in the neighborhood of a given sample at distance ∈.

[0049] For example, suppose the sample set is D = (x1, x2, ..., x...). m (Equivalent to the company's registered capital, establishment time, industry classification, capital type, company type, shareholder information, etc.), then the specific density description definition of DBSCAN is as follows:

[0050] ∈-neighborhood: for x j ∈D, its ∈-neighborhood contains samples in the sample set D that are the same as x. j A subset of samples whose distance is not greater than ∈ is called a subset of samples. For example, a user sample whose registered capital (represented by the letter M) is between 500,000 and 1,000,000 can be represented as D containing the dimension of registered capital of enterprises in N∈(D)={M∈D|distance(30.0,50.0)≤∈}. The number of this subset is denoted as |N∈(M)|.

[0051] Core object: For any sample x jIf (M)∈D, and its ∈-neighborhood corresponds to N∈(M) containing at least MinPts samples, that is, if |N∈(M)|≥MinPts, then M is a core object.

[0052] Density direct access: If the data X of a single user in the sample is located at x j In the ∈-neighborhood of (M), and x j If (M) is the core object, then X is said to be composed of x. j (M) Density directly reaches the target. The converse is not necessarily true; in this case, we cannot say x... j (M) is directly reached by the density of X, unless X is also a core object.

[0053] Density achievable: for X and x j (M), if a sample sequence p1, p2, ..., p exists. T Satisfying p1=X,p T =x j (M), and p t+1 By p t If the density is directly reached, then it is called x. j (M) is reachable by the density of X, and the density reachability satisfies transitivity. At this point, the transitive samples p1, p2, ..., p in the sequence... T-1 All are core objects because only core objects allow direct density access to other samples. Density accessibility also does not satisfy symmetry, which can be derived from the asymmetry of density accessibility. Among them, samples p1 and p... T Samples (p1, p2, ..., p) are passed between each other. T-1 Let the number of samples be denoted as T-1. The number of samples transmitted in the shortest transmission path between two samples is defined as the transmission step size, then the transmission step size is T-1.

[0054] Density connected: for X and x j (M), if a core object sample x exists k (M), such that X and x j (M) are all composed of x k If the density of (M) is achievable, then X and x are called X and x. j (M) Density connection. The density connection relationship satisfies symmetry.

[0055] To illustrate the algorithm with a more intuitive example, such as... Figure 3 As shown in the diagram, the points connected by the arrows with MinPts = 5 are all core objects because their ∈- neighborhood has at least 5 samples. Points other than those connected by arrows are non-core objects. All density-reachable samples of core objects are within a sphere centered on the core object; if a sample is not within the sphere, it is not density-reachable. The core objects connected by arrows in the diagram form density-reachable sample sequences. Within the ∈- neighborhood of these density-reachable sample sequences, all samples are density-connected to each other.

[0056] Specifically, DBSCAN clustering is straightforward: the set of samples connected by the highest density, derived from density reachability relations, yields the final category or cluster, which serves as the clustering result. A DBSCAN cluster can contain one or more core objects. If there is only one core object, all other non-core object samples in the cluster are within the ∈-neighborhood of this core object; if there are multiple core objects, then the ∈-neighborhood of any core object in the cluster must contain another core object, otherwise these two core objects are not density reachable. The set of all samples within the ∈-neighborhood of these core objects constitutes a DBSCAN cluster.

[0057] During the clustering process, an unclassified core dimension is arbitrarily selected as the seed, and then all sample sets that can be reached by density of this core dimension are found, which is a cluster (a group of users of a specific dimension); then another unclassified core object is selected to find a sample set that can be reached by density, thus obtaining another cluster; this process continues until all dimensions have found their own core user areas, and then remote dimensions or a small number of sample points outside the cluster are marked as noise points.

[0058] The DBSCAN clustering algorithm is specifically applied to dimension weights as follows:

[0059] Input: Sample set D = (regional attribute, registered capital of enterprise, establishment time, industry classification, capital type, enterprise type, shareholder information, number of employees, enterprise movable property mortgage, etc.), neighborhood parameter (∈, MinPts), sample distance metric.

[0060] Output: Cluster partition C.

[0061] The DBSCAN clustering algorithm, in practical applications, includes the following detailed steps:

[0062] (1) Initialize the core object collection Initialize the number of clusters k = 0, initialize the set of unvisited samples Γ = D, and then perform cluster partitioning.

[0063] (2) For j = 1, 2, ... m, find all core objects using the following steps:

[0064] Find sample x using a distance metric. j The ∈-neighborhood subset N∈(M); or, if the number of samples in the subset satisfies |N∈(M)|≥MinPts, add sample M to the core object sample set: Ω=Ω∪{M};

[0065] (3) If the core object collection The algorithm ends if the algorithm ends; otherwise, proceed to step (4).

[0066] (4) In the core object set Ω, randomly select the next core object o (regional attribute), initialize the current cluster core object queue Ωcur = {o (regional attribute)}, initialize the category index k = k + 1, and initialize the current cluster sample set C. k = {o(region attribute)}, update the unvisited sample set Γ = Γ - {o(region attribute)};

[0067] (5) If the current cluster core object queue Then the current cluster C k Once generated, update the cluster partition C = {Enterprise Registered Capital, Regional Attributes, ...}, and update the core object set Ω = Ω - C. k Proceed to step (3);

[0068] (6) Take a core object o′ from the current cluster core object queue Ωcur, find all ∈-neighborhood subsets N∈(o′) using the neighborhood distance threshold ∈, let Δ=N∈(o′)∩Γ, and update the current cluster sample set C. k =C k ∪Δ, update the unvisited sample set Γ=Γ-Δ, update Ωcur=Ωcur∪(Δ∩Ω)-o′, and proceed to step (5).

[0069] The output is: Cluster partition C = {C1, C2, ..., C} k}

[0070] Furthermore, based on the above method embodiments, step S1, which involves obtaining measurement parameters corresponding to user identity features and performing cluster analysis on users using a density-based clustering algorithm based on the measurement parameters to obtain clustering results, specifically includes:

[0071] The user's measurement parameters are obtained as samples. If the number of samples exceeds a set threshold, the nearest neighbors are searched using a K-space tree or ball tree. The sample distance in the clustering algorithm is determined using Euclidean distance based on the nearest neighbors. Clustering analysis is then performed on the users based on the sample distance and the measurement parameters to obtain the clustering result. This clustering result serves as the basis for correcting subsequent clustering results. Accordingly, when correcting clustering results based on a predefined dataset and updating the target set based on modified propagation step size, as described below, the data in the predefined dataset is not directly related to the bidding users and only reflects the data currently held by the bidding users (which may not be accurate) or their selection preferences. Therefore, updating the target set based on the predefined dataset (including the correction of noise points) does not serve as the basis for correcting clustering results in subsequent bidding activities.

[0072] Compared to the traditional K-Means algorithm, the biggest difference of DBSCAN is that it does not require inputting the number of categories k, and its advantage lies in its ability to discover clusters of arbitrary shapes. Specifically, when solving the distance metric problem in DBSCAN, a nearest neighbor method is used based on user samples, employing a distance metric (such as Euclidean distance) to measure sample distances. When the number of user samples is relatively small, finding the nearest neighbor can directly calculate the distance between all samples; when the number of user samples is relatively large, a K-space tree or a ball tree can be used to quickly search for the nearest neighbor.

[0073] Furthermore, the clustering result includes a target set and noise points. Based on the clustering result obtained in step S1 of the above method embodiment, the clustering result is corrected for the noise point data based on a predefined data set. The predefined data set includes known association data, which includes first association data used to characterize the relationship between two users and second association data used to characterize the absence of a relationship between two users.

[0074] The predefined data set here refers to a data set predefined by the bidding user in the current bidding activity. It is defined by the current bidding user, and the data in the predefined data set is not directly related to the bidding user; it only reflects the data (not necessarily accurate) or selection preferences held by the current bidding user. For example, if the bidding user in the current bidding activity knows that users A and B have a relationship, then the information of users A and B will be stored in the predefined data set, and a first relationship between A and B will be marked. If it is known that users A and B do not have a relationship, then the information of users A and B will be stored in the predefined data set, and a second relationship between A and B will be marked. (See reference...) Figure 4 If the clustering results calculated in step S1 indicate that a certain sample point (bidder C) is a noise point, the sample point can be re-verified based on a predefined data set. If a sample point (bidder C) exists in the predefined data set and has a first association relationship with other sample points (bidder D), it means that the sample point should actually belong to a certain cluster sample set C. j Therefore, by using a predefined dataset, it is possible to prevent the omission of bid rigging and collusion due to incomplete data in large datasets, and to supplement or correct the big data clustering results using the information currently available to the user.

[0075] Furthermore, based on the clustering results obtained in step S1 of the above method embodiment, a default transfer step size is set, and the target set is updated according to the default transfer step size; the transfer step size has been defined above, which is the number of transfer samples between two density-reachable samples in the density reachability relationship.

[0076] In the specific context of correlation detection, the standards for identifying bid-rigging and collusion need to be relatively strict to avoid mistaking users with weak correlations for the strong correlations required by this invention. In a specific embodiment, a default propagation step size can be set to decompose the target set in the clustering results into multiple target sets. Sample points originally in the same target set will belong to different target sets after decomposition. This can be understood as requiring a stronger correlation to be identified as potentially involved in bid-rigging or collusion. Figure 5 For example, if the default transit step size is set to 3, since the transit step size from point p1 to point p5 in the original graph is 4, the original single cluster will be decomposed into two clusters, such as... Figure 6 and Figure 7 As shown.

[0077] Furthermore, based on the clustering results obtained in step S1 of the above method embodiment, the clustering results are corrected according to a predefined data set, referring to... Figure 8 The correction methods can include the following two:

[0078] (1) Based on the first association data in the predefined data set, increase the propagation step size and update the target set according to the propagation step size.

[0079] Reference Figure 8 First, it reads data from a predefined dataset that has a primary relationship, such as a primary relationship between users A and B. Then, it determines whether both A and B exist in a target set after decomposition based on the default propagation step size. If both A and B exist, the default propagation step size setting is not found to be unreasonable, and the target set does not need to be updated. If neither A nor B exists in any target set, the default propagation step size setting is unreasonable, and the propagation step size needs to be increased to expand the target set so that both A and B exist in the updated target set. Only then can the relationships between other users in the target set be more accurate.

[0080] Assuming that, based on the above embodiment, the default propagation step size is set to 3, since the original Figure 5 If the transit step size from midpoint p1 to point p5 is 4, then the original single cluster will be decomposed into two clusters, as follows: Figure 6 and Figure 7 As shown. At this point, a first association is found between points p1 and p5 in the predefined data set. Therefore, the propagation step size needs to be increased to 4 so that the updated target set contains both A and B. This is equivalent to needing to... Figure 6 and Figure 7 The two clusters shown are merged into the following: Figure 5 The image shows a cluster.

[0081] In special cases, if A and B do not exist simultaneously in the clustering results calculated in the initial step S1 (without setting the propagation step size), then no matter how much the propagation step size is increased, the updated target set cannot contain both A and B. In this case, it is necessary to merge the cluster Cm where A is located with the cluster Cn where B is located. A and B can also be noise points.

[0082] like Figure 9 As shown, assuming that the default propagation step size is set to 3 based on the above embodiment, Figure 9 If the transfer step size from midpoint p1 to point p5 is 4, then the original single cluster will be decomposed into two clusters. The core object data points in these four clusters are {p1, p2, p3, p4} and {p2, p3, p4, p5}, respectively. Figure 6 and Figure 7 As shown; similarly, Figure 9 Midpoint p6 to point p 12 If the propagation step size is 6, then the original single cluster will be decomposed into four clusters, and the core points (sample sequences) of these four clusters will be {p6, p7, p8, p9}, {p7, p8, p9, p... ... 10}, {p8, p9, p 10 p 11}, {p9, p 10 p 11 p 12 Meanwhile, assume A is point p in the graph. 10 Let B be point p3 in the graph. Merging the target set containing A and the target set containing B, the core object data points in the corresponding clusters of the resulting target sets are {p1, p2, p3, p4, p7, p8, p9, p...}. 10}, {p1, p2, p3, p4, p8, p9, p 10 p 11}, {p1, p2, p3, p4, p9, p 10 p 11 p 12}, {p2, p3, p4, p5, p7, p8, p9, p 10}, {p2, p3, p4, p5, p8, p9, p 10 p 11}, {p2, p3, p4, p5, p9, p 10 p 11 p 12}, {p7, p8, p9, p 10 ,p1,p2,p3,p4},{p8,p9,p 10 p 11 ,p1,p2,p3,p4},{p9,p 10 p11 p 12 ,p1,p2,p3,p4},{p7,p8,p9,p 10 ,p2,p3,p4,p5},{p8,p9,p 10 p 11 ,p2,p3,p4,p5},{p9,p 10 p 11 p 12 After merging the target sets {p1, p2, p3, p4, p5}, no additional propagation step is needed to ensure that the updated target set simultaneously contains both A and B. Then, the merged target sets are decomposed according to the currently set propagation step of 4, finally yielding the corresponding updated target sets. Taking one of the merged target sets as an example, its corresponding cluster's core object data points are {p1, p2, p3, p4, p7, p8, p9, p1}. 10}, then after decomposition, the core object data points in the five corresponding clusters are {p1, p2, p3, p4}, {p2, p3, p4, p7}, {p3, p4, p7, p8}, {p4, p7, p8, p9}, {p7, p8, p9, p 10 Therefore, compared to the target set before the update, three new target sets have been added.

[0083] (2) Based on the second association data in the predefined data set, reduce the transmission step size and update the target set according to the changed transmission step size.

[0084] Reference Figure 7 First, it reads data from a predefined dataset that has a second association relationship, such as a second association relationship between users A and B. Then, it determines whether A and B both exist in a target set after decomposition based on the default propagation step size. If A and B both exist, it indicates that the default propagation step size setting is unreasonable and needs to be reduced to shrink the target set, ensuring that A and B do not both exist in the updated target set. Only then can the association relationships between other users in the target set be more accurate. If neither A nor B exists in any target set simultaneously, it indicates that the default propagation step size setting is reasonable and there is no need to update the target set.

[0085] In special cases, if the clustering results calculated in the initial step S1 contain both A and B, and the propagation step size is 1 (equivalent to A being directly reached by B's density or B being directly reached by A's density), then it is impossible to ensure that neither A nor B exists simultaneously in any of the updated target sets. In another special case, if the clustering results calculated in the initial step S1 contain both A and B, and the propagation step size is 2, then only by reducing the propagation step size to 1 can neither A nor B exist simultaneously in any of the updated target sets. However, reducing the propagation step size to 1 may not be reasonable. In both of these special cases, a lower limit for the propagation step size needs to be set, for example, setting the lower limit to 3, or an upper limit for the reduction in the propagation step size needs to be set, for example, limiting the propagation step size to a maximum reduction of 2, before updating the target sets. In this case, the reduction in the propagation step size reflects the correction of the clustering results, but it is still possible that A and B exist simultaneously in some target sets. Therefore, an exception needs to be added in step S3: even if A and B have a primary association relationship, it is not considered that there is a bid-rigging association relationship between the users represented by A and B.

[0086] The main drawback of the DBSCAN clustering algorithm is that its parameter tuning is slightly more complex than traditional clustering algorithms like K-Means. It primarily requires joint tuning of the distance threshold ∈ and the neighborhood sample number threshold MinPts, and different parameter combinations have a significant impact on the final clustering results. The above embodiment uses a step-by-step approach, which avoids the need for joint parameter tuning after calculating the clustering results due to data correction, thus avoiding extensive recalculation and improving clustering accuracy. In particular, this embodiment introduces a re-verification of noisy points, avoiding the need for re-joint parameter tuning every time noise points are corrected. It also facilitates the setting of preferences by bidding users, ensuring that the preferences of a single bidding user do not affect those of other bidding users.

[0087] The above embodiments detail the steps for correction based on a predefined dataset, primarily involving automatic correction of the clustering results based on the association information held by the bidding users. However, in some cases, this automatic correction method may not achieve the desired results. For example, in practical applications, it has been found that most companies, to prevent fraud during the bidding process, require a certain number of participating users (enterprises) before the bidding process can begin. However, the clustering results obtained in the above specific embodiments may show an insufficient number of qualified bidding users. Therefore, in addition to allowing bidding users to input predefined data, it is also possible to allow them to input parameters for adjusting the transmission step size.

[0088] Furthermore, based on the above embodiments, it also includes:

[0089] The step to obtain the step size adjustment parameter is to obtain the step size adjustment parameter input by the bidding user in the current bidding activity.

[0090] The target set update step involves calculating a new transfer step based on the parameters adjusted according to the transfer step, and updating the target set based on the transfer step.

[0091] The recommended user generation step involves removing the target set containing the users participating in the current bidding activity from the updated target set, and then selecting one bidding user from each of the remaining target sets as the recommended user.

[0092] After the bidding user inputs data from a predefined dataset, they can modify the propagation step size to correct the clustering results. However, the predefined dataset cannot directly control the increase or decrease of the propagation step size. Therefore, in some embodiments of this invention, the propagation step size adjustment parameter input by the bidding user of the current bidding activity is directly obtained and adjusted. For example, after judging the bidding information (including participating users) to be analyzed based on the clustering results obtained in step S1 and deleting participating users with correlations, if the number of qualified participating (bidding) users is insufficient, the propagation step size is reduced according to the propagation step size adjustment parameter, thereby increasing the number of the modified target set and the corresponding number of qualified participating (bidding) users. The target set containing the participating users of the current bidding activity is removed from the updated target set, and one bidding user is selected from each of the remaining target sets as a recommended user to ensure that there are no correlations among the recommended users. For the recommended users, the bidding user of the current bidding activity can send bidding invitation information (because these recommended users are newly added, they are not in the list of participating bidding users of the current bidding activity). Therefore, the bidding user can obtain recommendation information according to the above steps, which can be used to solve the problem of insufficient number of qualified bidding users.

[0093] The step size adjustment parameter can be one of the following two types: (1) a specific step size adjustment value; (2) a step size adjustment direction.

[0094] Reference Figure 10 In the first scenario, when the step size adjustment parameter is a specific step size adjustment value, the adjustment is made directly based on this parameter. For example, a step size adjustment value of step = 1 indicates an increase of 1 in the step size, and a step size adjustment value of step = -1 indicates a decrease of 1 in the step size. The target set is then updated based on this specific step size adjustment value. The number of participating users meeting the bidding requirements can then be output and displayed. If the bidding user notices that the number of bidding users still does not meet expectations, they can re-enter the step size adjustment parameter and repeat the step size acquisition and target set update steps until the result meets expectations.

[0095] Reference Figure 10In the second case, when the transfer step size adjustment parameter is a specific long adjustment direction, the transfer step size adjustment parameter only indicates whether the transfer step size is increased (dir = 1) or decreased (dir = -1). Taking the transfer step size adjustment parameter only indicating that the transfer step size is decreased as an example, refer to the foregoing embodiments and appendices. Figure 5-7 It can be seen that the updated target set increases in size, while the number of data points coexisting in a single target set decreases. Therefore, the number of data points indicating a correlation decreases, which in turn indicates an increase in the number of eligible bidding users. A specific update method can be implemented using the minimum propagation step size modification value (step). min And setting a threshold for the number of bidding users. th In combination, for example, setting the transfer step size adjustment parameter dir = -1 reduces the transfer step size, and the minimum transfer step size modification value is step. min =1, then before each update of the target set, the step value will be decremented by 1. After the update, it is determined whether the number of qualified bidding users is greater than the set threshold for the number of bidding users. If yes, the correction is completed. If no, the step value will be decremented by 1 again, and the number of bidding users will be determined after the update. This process is repeated until the number of qualified bidding users after the update is greater than the set threshold for the number of bidding users.

[0096] Typically, when the number of qualified bidding users obtained from the clustering results in step S1 is insufficient, correction is achieved by setting the transmission step size adjustment parameter dir = -1 and decreasing the transmission step size. In rare cases, correction may also be necessary when the number of qualified bidding users obtained from the clustering results in step S1 is too large; in this case, the number of qualified bidding users is clearly greater than the set threshold for the number of bidding users. th Therefore, the transfer step size adjustment parameter dir = 1 is set, and the minimum transfer step size modification value is used in the actual calculation process as step. min The calculation is performed with a value of 1, which reduces the number of eligible bidding users accordingly, but the result must still be greater than the user threshold. th Typically, only one calculation is needed. Therefore, to avoid the calculation process falling into a positive feedback loop, the number of eligible bidding users must be less than the threshold user. th At this point, the process reverts to the state before updating the target set based on the propagation step size, and is directly set to complete the correction process. For example, if the clustering result obtained in step S1 shows that the number of qualified bidding users is 50, and the threshold for the number of bidding users is set to user... th =20, the propagation step size is 4, because the propagation step size adjustment parameter dir=1 is set, and the minimum propagation step size modification value is step. minIf the value is 1, the propagation step size is adjusted to 5 for calculation. Assuming the calculation result shows that the number of qualified bidders is 15, which is less than the threshold of 20, it is necessary to return to the state before updating the target set based on the propagation step size of 5, that is, return to the state of updating the target set based on the propagation step size of 4.

[0097] In specific embodiments of the present invention, the combination of step size adjustment parameters and visualization can be used to obtain the step size adjustment parameters input by the bidding user in the web client first, and then directly correct the clustering results to meet the needs of the bidding user based on the step size adjustment parameters. This eliminates the need to recalculate the clustering results from scratch, resulting in fast calculation speed. Furthermore, the corrected results can be used only for the calculation of the current bidding activity without affecting the judgment of other bidding activities. This increases the calculation speed while avoiding the influence of a single user's preference on other users.

[0098] Reference Figure 11 The electronic device includes a processor 101, a memory 102, and a bus 103; the processor 101 and the memory 102 communicate with each other through the bus 103; the processor 101 is used to call program instructions in the memory 102 to execute the methods provided in the above-described method embodiments.

[0099] This embodiment discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer can perform the methods provided in the above-described method embodiments.

[0100] This embodiment provides a non-transitory computer-readable storage medium that stores computer instructions that cause the computer to execute the methods provided in the above-described method embodiments.

[0101] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0102] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A density clustering method that does not rely on historical information, characterized in that, include: The system acquires measurement parameters corresponding to the user, and uses a density-based clustering method with noise based on the measurement parameters. It calculates the sample set with the highest density connection based on the density reachability relationship, and the obtained sample set is used as the clustering result. The clustering result includes the target set and noise points. The noise point data is corrected based on a predefined data set. The predefined data set includes known association data, which includes first association data to characterize the relationship between two users and second association data to characterize the absence of a relationship between two users. The measurement parameters are one or more combinations of regional attributes, registered capital, establishment time, industry classification, capital type, enterprise type, shareholder information, number of employees, and movable property mortgage information. The step of correcting the noise point data based on a predefined data set specifically involves: setting a default transfer step size and updating the target set according to the default transfer step size; the transfer step size is the number of transfer samples between two density-reachable samples in a density-reachable relationship. Obtain the bidding information to be analyzed, including the users participating in the current bidding activity; Based on the clustering results, determine whether there are any relationships among the users participating in the current bidding activity.

2. The density clustering processing method according to claim 1, which does not rely on historical information, is characterized in that, Based on the first association data in the predefined data set, increase the propagation step size and update the target set according to the propagation step size.

3. The density clustering processing method according to claim 1, which does not rely on historical information, is characterized in that, Based on the second association data in the predefined data set, reduce the propagation step size and update the target set according to the changed propagation step size.

4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a density clustering processing method that does not rely on historical information as described in any one of claims 1-3.

5. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a density clustering processing method that does not rely on historical information as described in any one of claims 1-3.

Images