A user friend recommendation system and method with delay centralized distribution

Abstract

The application discloses a user friend recommendation system and method for delay centralized distribution, which comprises a client and a cloud end, the cloud end is provided with an allocation center and a decision engine; the client is used for real-time request initiation and data reporting of users, and communication between users; the allocation center centrally maintains a user list through a time window and retrieves and reports the latitude and longitude of the user, when the number of elements of the user list or the cumulative length of the time window meets the preset trigger condition, the decision engine is triggered to perform matching calculation; the decision engine receives the latitude and longitude of the user and performs matching calculation by using a Kuhn-Munkres algorithm, so that a globally optimal matching group is screened out and is issued to the corresponding client. The application adopts a strategy of delay centralized distribution and time selection matching, so that optimal matching of the to-be-matched party and the matched party in a global range is realized, the matching accuracy is improved, the utilization rate and conversion rate of social resources are improved, and the social experience of the user is optimized.

Description

Technical Field

[0001] This invention relates to the field of social recommendation system technology, and in particular to a user dating recommendation system and method with delayed centralized distribution. Background Technology

[0002] Current social recommendation systems primarily employ a Lambda architecture, combining offline training with real-time recommendation services. Real-time recommendation services can more quickly recommend users or products that users like. However, practical application issues still exist in paid social recommendation scenarios. For example, in social chat, a user A with a high chat response rate might be matched with multiple users B whose chat intentions differ. User A might not be able to reply to all users B in a timely manner. That is, user A might be chatting with one user B who has low chat intentions, while another user B with high chat intentions might wait for a period of time without receiving a reply. This not only wastes social resources but also negatively impacts the user's social experience. Summary of the Invention

[0003] To address the aforementioned problems, this invention provides a user-based friend recommendation system and method with delayed centralized distribution.

[0004] The present invention adopts the following technical solution:

[0005] A user-friendly social recommendation system with delayed centralized distribution, the system comprising a client and a cloud, wherein the cloud is equipped with a distribution center and a decision engine;

[0006] The client allows users to initiate requests and report data in real time, as well as communicate with each other.

[0007] The allocation center maintains the user list centrally through time windows and retrieves and reports the latitude and longitude of users. When the number of elements in the user list or the cumulative duration of the time window meets the preset triggering conditions, the decision engine is triggered to perform matching calculations.

[0008] The decision engine receives the user's latitude and longitude and performs matching calculations using the Kuhn-Munkres algorithm to select the globally optimal matching combination. The globally optimal matching combination is then sent to the corresponding client via the allocation center.

[0009] Furthermore, the users are divided into those to be matched and those to be matched, and the user list includes a list of those to be matched and a list of those to be matched.

[0010] Furthermore, the retrieval and reporting of the user's latitude and longitude specifically involves: calculating the latitude and longitude of the N nearest matching parties based on the latitude and longitude of the party to be matched using the GeoHash algorithm, and reporting them to the decision engine when the matching calculation is triggered.

[0011] Furthermore, the preset triggering conditions are as follows: when the number of elements in the matching user list or the user list to be matched within the time window exceeds 50 to 100, or when the cumulative duration of the time window exceeds 1 to 5 minutes, a matching request is sent to the decision engine to trigger the decision engine to perform matching calculations.

[0012] Furthermore, the calculation process of the Kuhn-Munkres algorithm is as follows:

[0013] S1. Use the nodes of the bipartite graph to represent the party to be matched X and the party to be matched Y respectively, and the line connecting the party to be matched X and the party to be matched Y represents the weight between the two.

[0014] S2. Calculate the weight value from each party to be matched X to each party to be matched Y;

[0015] S3. Use the Kuhn-Munkres algorithm to calculate the globally optimal matching combination between each party X to be matched and each party Y in the bipartite graph. Under the globally optimal matching combination, the sum of the weight values ​​of all the connections in the bipartite graph is maximized.

[0016] Furthermore, the system also includes a message center, which is responsible for message transmission between the client and the cloud, between the cloud and the client, and between the client and the client.

[0017] Furthermore, the message transmission between the client and the cloud refers to the client reporting the latitude and longitude of the matching user and the user to be matched through the message center; the message transmission between the cloud and the client refers to the cloud pushing down the globally optimal matching combination result to the client; and the message transmission between clients refers to the communication between the matching user and the user to be matched.

[0018] A method for delayed centralized distribution of user dating recommendations, implemented through the user dating recommendation system described above, includes the following steps:

[0019] When the party to be matched logs in to the client, the client of the party to be matched automatically initiates a request and reports latitude and longitude. When the party to be matched logs in to the client, the client of the party to be matched automatically and continuously reports latitude and longitude.

[0020] The distribution center receives the latitude and longitude reported by the client, adds the parties to be matched and the parties to be matched to the list of parties to be matched and the list of parties to be matched to ...

[0021] When the number of elements in the list of parties to be matched or the list of parties to be matched, or the cumulative duration of the time window, meets the preset triggering conditions, the decision engine is triggered to perform matching calculations.

[0022] The decision engine receives the latitude and longitude of the parties to be matched and the parties to be matched, and uses the Kuhn-Munkres algorithm to perform matching calculations, thereby selecting the globally optimal matching combination and pairing the parties to be matched and the parties to be matched.

[0023] The globally optimal matching combination is distributed to the corresponding client via the allocation center.

[0024] By adopting the above technical solution, the present invention has the following advantages compared with the prior art:

[0025] This invention employs a strategy of delayed centralized distribution and time-selective matching. First, it centralizes the potential matches and those to be matched. Then, when a trigger condition is met, it performs optimal matching on the potential and those to be matched within a given time window, thus achieving the best match globally. Compared to traditional dating recommendation systems that only address the question of "who to assign to," this invention's dating recommendation system performs globally optimal recommendations from multiple perspectives, including "whether to assign, when to assign, and who to assign to," thereby improving matching accuracy, the utilization and conversion rates of social resources, and optimizing the user's social experience. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0027] Figure 2 This is an example diagram of the Kuhn-Munkres algorithm in this embodiment. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0029] Example

[0030] The first objective of this invention is to provide a user-friendly social recommendation system with delayed centralized distribution, such as... Figure 1 As shown, the system includes a client and a cloud, and the cloud is equipped with a distribution center and a decision engine;

[0031] (1) Client:

[0032] The client allows users to initiate requests and report data in real time, as well as communicate with each other.

[0033] (2) Distribution Center;

[0034] The allocation center maintains the user list centrally through time windows and retrieves and reports the latitude and longitude of users. When the number of elements in the user list or the cumulative duration of the time window meets the preset triggering conditions, the decision engine is triggered to perform matching calculations.

[0035] Specifically, the users are divided into those to be matched and those to be matched, and the user list includes a list of those to be matched and a list of those to be matched. When a user to be matched logs into the client, the client automatically initiates a request and reports the latitude and longitude to the allocation center via the message center. When a user to be matched logs into the client, the client automatically and continuously reports the latitude and longitude to the allocation center via the message center.

[0036] After receiving the data reported by the parties to be matched and the parties to be matched, the allocation center uses the GeoHash algorithm to calculate the latitude and longitude of the N nearest parties to the parties to be matched based on the latitude and longitude of the parties to be matched. When the matching calculation is triggered, the latitude and longitude of the parties to be matched and the parties to be matched are reported to the decision engine.

[0037] Specifically, the preset triggering conditions are as follows: when the number of elements in the matching user list or the user list to be matched within the time window exceeds 100, or when the cumulative duration of the time window exceeds 1 minute, a matching request is sent to the decision engine to trigger the decision engine to perform matching calculations.

[0038] (3) Decision Engine:

[0039] The decision engine receives the user's latitude and longitude and performs matching calculations using the Kuhn-Munkres algorithm to select the globally optimal matching combination. The globally optimal matching combination is then sent to the corresponding client via the allocation center and message center.

[0040] Specifically, the Kuhn-Munkres algorithm (i.e., the Hungarian algorithm) is a bipartite graph algorithm, and its calculation process is as follows:

[0041] S1. Use the nodes of the bipartite graph to represent the party to be matched X and the party to be matched Y respectively, and the line connecting the party to be matched X and the party to be matched Y represents the weight between the two.

[0042] S2. Calculate the weight value from each party to be matched X to each party to be matched Y;

[0043] S3. Use the Kuhn-Munkres algorithm to calculate the globally optimal matching combination between each party X to be matched and each party Y in the bipartite graph. Under the globally optimal matching combination, the sum of the weight values ​​of all the connections in the bipartite graph is maximized.

[0044] like Figure 2As shown, when the matching calculation is triggered, a portion of the parties to be matched and the parties to be matched have already been gathered. Let the parties to be matched be X1, X2, and X3, and the parties to be matched be Y1, Y2, and Y3. In the bipartite graph, X1, X2, and X3 and Y1, Y2, and Y3 are the nodes on both sides of the graph, respectively. Each party to be matched and each party to be matched forms a connection. However, depending on their latitude and longitude, the weights (i, j) of these connections are different. The Kuhn-Munkres algorithm can be used to calculate the optimal solution in the entire bipartite graph, obtaining the best pairing combination of parties to be matched and parties to be matched (X1 / Y3, X2 / Y1, X3 / Y2 in the figure), which maximizes the sum of the weights of all connections in the entire bipartite graph, thus considering the optimal solution of parties to be matched and parties to be matched globally.

[0045] (4) The message center is responsible for message transmission between the client and the cloud, between the cloud and the client, and between the client and the client.

[0046] The message transmission between the client and the cloud refers to the client reporting the latitude and longitude of the matching user and the user to be matched through the message center; the message transmission between the cloud and the client refers to the cloud pushing down the globally optimal matching combination result to the client; the message transmission between the client and the client refers to the communication between the matching user and the user to be matched.

[0047] Traditional dating recommendation systems provide real-time recommendations, where system items can be repeatedly distributed, meaning multiple matches are recommended to a user (one-to-many). In contrast, the dating recommendation system of this invention does not distribute items repeatedly within a time window; that is, users and matches are paired one-to-one within the time window, thus achieving globally optimal recommendations. Compared to traditional dating recommendation systems that only address the question of "who to assign to," this invention's system performs globally optimal recommendations from multiple perspectives, including "whether to match, when to match, and who to assign to," thereby improving matching accuracy, the utilization and conversion rate of social resources, and optimizing the user's social experience.

[0048] The second objective of this invention is to provide a delayed centralized distribution method for user dating recommendations, which is implemented through the user dating recommendation system described above, and includes the following steps:

[0049] When the party to be matched logs in to the client, the client of the party to be matched automatically initiates a request and reports latitude and longitude. When the party to be matched logs in to the client, the client of the party to be matched automatically and continuously reports latitude and longitude.

[0050] The distribution center receives the latitude and longitude reported by the client, adds the parties to be matched and the parties to be matched to the list of parties to be matched and the list of parties to be matched to ...

[0051] When the number of elements in the list of parties to be matched or the list of parties to be matched, or the cumulative duration of the time window, meets the preset triggering conditions, the decision engine is triggered to perform matching calculations.

[0052] The decision engine receives the latitude and longitude of the parties to be matched and the parties to be matched, and uses the Kuhn-Munkres algorithm to perform matching calculations, thereby selecting the globally optimal matching combination and pairing the parties to be matched and the parties to be matched.

[0053] The globally optimal matching combination is distributed to the corresponding client via the allocation center.

[0054] The user dating recommendation method in this embodiment is based on the user dating recommendation system described above, and its implementation principle and strategy will not be elaborated here.

[0055] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A user-based social recommendation system with delayed centralized distribution, characterized in that: The system includes a client and a cloud, wherein the cloud is equipped with a distribution center and a decision engine; The client allows users to initiate requests and report data in real time, as well as communicate with each other. The allocation center maintains the user list centrally through time windows and retrieves and reports the latitude and longitude of users. When the number of elements in the user list or the cumulative duration of the time window meets the preset triggering conditions, the decision engine is triggered to perform matching calculations. The decision engine receives the user's latitude and longitude and performs matching calculations using the Kuhn-Munkres algorithm to select the globally optimal matching combination. The globally optimal matching combination is then sent to the corresponding client via the allocation center. The calculation process of the Kuhn-Munkres algorithm is as follows: S1. Use the nodes of the bipartite graph to represent the party to be matched X and the party to be matched Y respectively, and the line connecting the party to be matched X and the party to be matched Y represents the weight between the two. S2. Calculate the weight value from each party to be matched X to each party to be matched Y; S3. Use the Kuhn-Munkres algorithm to calculate the globally optimal matching combination between each party X to be matched and each party Y in the bipartite graph. Under the globally optimal matching combination, the sum of the weight values ​​of all the connections in the bipartite graph is maximized.

2. The user-friendly social recommendation system with delayed centralized distribution as described in claim 1, characterized in that: The users are divided into those to be matched and those to be matched, and the user list includes a list of those to be matched and a list of those to be matched.

3. The user-friendly social recommendation system with delayed centralized distribution as described in claim 2, characterized in that: The process of retrieving and reporting the latitude and longitude of users specifically involves: using the GeoHash algorithm to calculate the latitude and longitude of the N nearest matching parties based on the latitude and longitude of the party to be matched, and reporting them to the decision engine when the matching calculation is triggered.

4. The user-friendly social recommendation system with delayed centralized distribution as described in claim 3, characterized in that: The preset triggering conditions are as follows: when the number of elements in the matching user list or the user list to be matched within the time window exceeds 50 to 100, or when the cumulative duration of the time window exceeds 1 to 5 minutes, a matching request is sent to the decision engine to trigger the decision engine to perform matching calculations.

5. A user-friendly social recommendation system with delayed centralized distribution as described in claim 4, characterized in that: The system also includes a message center, which is responsible for message transmission between clients and the cloud, between the cloud and clients, and between clients.

6. The user-friendly social recommendation system with delayed centralized distribution as described in claim 5, characterized in that: The message transmission between the client and the cloud refers to the client reporting the latitude and longitude of the matching user and the user to be matched through the message center; the message transmission between the cloud and the client refers to the cloud pushing down the globally optimal matching combination result to the client; the message transmission between the client and the client refers to the communication between the matching user and the user to be matched.

7. A method for delayed centralized distribution of user-based friend recommendations, characterized in that: The method is implemented using a user-friendly recommendation system as described in any one of claims 1 to 6, and the method includes the following steps: When the party to be matched logs in to the client, the client of the party to be matched automatically initiates a request and reports latitude and longitude. When the party to be matched logs in to the client, the client of the party to be matched automatically and continuously reports latitude and longitude. The distribution center receives the latitude and longitude reported by the client, adds the parties to be matched and the parties to be matched to the list of parties to be matched and the list of parties to be matched to ... When the number of elements in the list of parties to be matched or the list of parties to be matched, or the cumulative duration of the time window, meets the preset triggering conditions, the decision engine is triggered to perform matching calculations. The decision engine receives the latitude and longitude of the parties to be matched and the parties to be matched, and uses the Kuhn-Munkres algorithm to perform matching calculations, thereby selecting the globally optimal matching combination and pairing the parties to be matched and the parties to be matched. The globally optimal matching combination is distributed to the corresponding client via the allocation center.

Images