Information push processing method and device based on historical behavior path, and medium

Abstract

The invention relates to the technical field of electric digital data processing, in particular to an information push processing method and device based on historical behavior paths and a medium. The method comprises the steps that a historical behavior path set of sample entities is obtained, the historical behavior path set comprises historical behavior paths of a plurality of sample entities, the historical behavior path of any sample entity comprises a plurality of nodes and edges connected with the nodes, and any node corresponds to one behavior; comparing the preset type attribute value of the target entity with the preset type attribute value of the sample entity to obtain similar entities of the target entity; acquiring a historical behavior path of the similar entity from a historical behavior path set of a sample entity, and acquiring a target directed graph according to the historical behavior path of the similar entity; and determining push information of the target entity according to a target behavior of the target entity and the target directed graph. According to the invention, the accuracy of pushing information to users or enterprises can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric digital data processing, in particular to an information push processing method and device based on historical behavior path and a medium. BACKGROUND

[0002] In the current era of information explosion, information push technology is widely used in many fields, aiming to provide accurate personalized services for users or enterprises and improve user or enterprise experience and service efficiency. Traditional information push methods are mostly based on static attributes (such as age, gender, region, and scale) of users or enterprises or simple behavior data (such as clicks and browsing records) for content recommendation. However, such methods have the following problems: on the one hand, static attributes are difficult to reflect the dynamic changes of user or enterprise demand, and push based on static attributes cannot be adapted in time; on the other hand, simple behavior data lacks deep mining of the behavior logic and time sequence relationship of users or enterprises. The above problems lead to difficulty in accurately and individually pushing users or enterprises, that is, the accuracy of the push information for users or enterprises is low, resulting in low satisfaction of users or enterprises with the push information. How to improve the accuracy of the push information for users or enterprises is a problem to be solved. SUMMARY

[0003] The present application aims to provide an information push processing method and device based on historical behavior path to improve the accuracy of the push information for users or enterprises.

[0004] According to a first aspect of the present application, an information push processing method based on historical behavior path is provided, which comprises the following steps: Obtaining a historical behavior path set of sample entities; the historical behavior path set of sample entities comprises historical behavior paths of a plurality of sample entities, any historical behavior path of a sample entity comprises a plurality of nodes and edges connecting the nodes, any node corresponds to one behavior, and the weight of any edge is the time interval between the occurrence of the two behaviors corresponding to the two nodes connected by the edge; the occurrence time of the behavior corresponding to the node located in front in any historical behavior path of a sample entity is earlier than the occurrence time of the behavior corresponding to the node located behind.

[0005] Comparing the preset type attribute value of the target entity with the preset type attribute value of the sample entity to obtain similar entities of the target entity; the attribute similarity between any similar entity and the target entity is greater than or equal to a preset attribute similarity threshold.

[0006] obtaining a historical behavior path of the similar entity from the historical behavior path set of the sample entity, and obtaining a target directed graph according to the historical behavior path of the similar entity; any node in the target directed graph corresponds to a behavior, and a first weight corresponding to any directed edge in the directed graph is a probability of occurrence from a behavior corresponding to a tail node of the directed edge to a behavior corresponding to a head node; and a second weight corresponding to any directed edge in the directed graph is a time interval of occurrence from the behavior corresponding to the tail node of the directed edge to the behavior corresponding to the head node.

[0007] determining push information of the target entity according to the target behavior of the target entity and the target directed graph.

[0008] Further, obtaining a target directed graph according to the historical behavior path of the similar entity comprises: traversing the historical behavior path of all similar entities, extracting all nodes in each historical behavior path of each similar entity, and constructing a node set of the directed graph.

[0009] constructing edges of the directed graph according to the historical behavior path of the similar entity, comprising: for each historical behavior path of each similar entity, sequentially processing adjacent node pairs in order of node position, adding a directed edge between two nodes corresponding to each adjacent node pair, and adding a directed edge between the two nodes corresponding to each adjacent node pair from the node with an earlier position in the adjacent node pair to the node with a later position in the adjacent node pair.

[0010] obtaining the first weight corresponding to each directed edge in the directed graph; the first weight corresponding to any directed edge is a probability of occurrence from a behavior corresponding to a tail node of the directed edge to a behavior corresponding to a head node; and the probability of occurrence from the behavior corresponding to the tail node of any directed edge to the behavior corresponding to the head node is a ratio of a number of occurrences of an adjacent node pair corresponding to the directed edge in the historical behavior path of the similar entity to a number of occurrences of the tail node of the directed edge as a tail node.

[0011] obtaining the second weight corresponding to each directed edge in the directed graph; the second weight corresponding to any directed edge is a time interval of occurrence from a behavior corresponding to a tail node of the directed edge to a behavior corresponding to a head node; and the time interval of occurrence from the behavior corresponding to the tail node of any directed edge to the behavior corresponding to the head node is a mean value of weights of edges corresponding to the directed edge in the historical behavior path of the similar entity.

[0012] determining the finally obtained directed graph as the target directed graph.

[0013] Further, determining push information of the target entity according to the target behavior of the target entity and the target directed graph comprises: locating a node matching the target behavior of the target entity in the target directed graph.

[0014] If the positioning succeeds, a head node set in which the matched node is located in the target directed graph is obtained.

[0015] For any head node in the head node set, a first weight of a directed edge between the matched node and the head node is obtained, and if the first weight is greater than or equal to a preset weight threshold, push information of the target entity is determined according to a behavior corresponding to the head node.

[0016] Further, the push information of the target entity determined according to the behavior corresponding to the head node includes the behavior corresponding to the head node and a second weight of a directed edge corresponding to the head node, the second weight of the directed edge corresponding to the head node being a second weight of a directed edge corresponding to the head node, the directed edge corresponding to the head node being a directed edge with the matched node as a tail node and the head node as a head node.

[0017] Further, comparing a preset type attribute value of the target entity with a preset type attribute value of the sample entity, the similar entity of the target entity includes: obtaining a similarity between each preset type attribute value of the target entity and a corresponding preset type attribute value of a specified sample entity, and if the similarity between a certain preset type attribute value of the target entity and a corresponding preset type attribute value of a specified sample entity is greater than or equal to a preset first similarity threshold, determining that the preset type attribute is a similar attribute of the target entity and the specified sample entity; the specified sample entity is any sample entity.

[0018] obtaining a number of similar attributes of the target entity and the specified sample entity, and determining a ratio of the number of similar attributes to a preset type number as an attribute similarity of the target entity and the specified sample entity.

[0019] If the attribute similarity of the target entity and the specified sample entity is greater than or equal to a preset attribute similarity threshold, the specified sample entity is determined as a similar entity of the target entity.

[0020] Further, any behavior includes a behavior action and an object name, and the construction process of the historical behavior path set of the sample entity includes: obtaining a text corresponding to each candidate object name in the candidate object name set; the text corresponding to any candidate object name includes a judgment condition for judging whether the behavior corresponding to the candidate object name occurs.

[0021] judging whether the behavior corresponding to the specified candidate object name occurs according to the text corresponding to the specified candidate object name, and if the behavior corresponding to the specified candidate object name occurs, appending a node corresponding to the behavior corresponding to the specified candidate object name to the historical behavior path of the specified sample entity according to a time when the behavior corresponding to the specified candidate object name occurs in the specified sample entity; the specified sample entity is any sample entity, and the specified candidate object name is any candidate object name.

[0022] Further, the candidate object name set obtaining process comprises: obtaining an initial object name set; the initial object name set comprises a plurality of initial object names.

[0023] For any initial object name, obtaining keywords and corresponding parts of speech in the initial object name.

[0024] For any two initial object names, obtaining a similarity of the two initial object names according to the keywords and the corresponding parts of speech of the two initial object names, and if the similarity of the two initial object names is greater than or equal to a preset object name similarity threshold, deleting one of the two initial object names in the initial object name set.

[0025] determining the updated initial object name set as the candidate object name set.

[0026] Further, the obtaining of the similarity of the two initial object names according to the keywords and the corresponding parts of speech of the two initial object names comprises: rearranging the keywords of a second initial object name according to a part of speech arrangement order of the keywords of a first initial object name; the first initial object name is one of the two initial object names, and the second initial object name is the other one of the two initial object names.

[0027] obtaining a semantic vector of the first initial object name according to the keyword order of the first initial object name; the semantic vector of the first initial object name is obtained by concatenating semantic vectors corresponding to the keywords of the first initial object name according to the keyword order.

[0028] obtaining a semantic vector of the second initial object name according to the rearranged keyword order of the keywords of the second initial object name; the semantic vector of the second initial object name is obtained by concatenating semantic vectors corresponding to the keywords of the second initial object name according to the rearranged keyword order.

[0029] determining a similarity of the semantic vector of the first initial object name and the semantic vector of the second initial object name as the similarity of the two initial object names.

[0030] According to a second aspect of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned information pushing processing method based on historical behavior path when executing the computer program.

[0031] According to a third aspect of the present application, a computer readable storage medium is provided, which stores a computer program, and the computer program, when executed by a processor, implements the above-mentioned information pushing processing method based on historical behavior path.

[0032] Compared with the prior art, the present application has at least the following advantages: The present application obtains a historical behavior path set of a sample entity, and each historical behavior path of the sample entity can reflect the behavior information of the sample entity and the time interval between behaviors. Based on the historical behavior path of the sample entity, the present application can obtain the time sequence feature and the logical relationship of the behavior of the sample entity. On this basis, the present application compares the preset type attribute value of the target entity with the sample entity, obtains a similar entity, and then constructs a target directed graph suitable for the target entity based on the historical behavior path of the similar entity. Based on the directed graph, the occurrence order relationship and the time interval relationship between different behaviors of the similar entity can be obtained, and then the target entity can be pushed more matched and more accurate personalized information through the directed graph and the target behavior of the target entity, which can improve the satisfaction of the target entity to the pushed information. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0034] Figure 1 The flowchart of the historical behavior path information pushing processing method provided by the first embodiment of the present application is shown in the figure. Figure 2 The flowchart of the construction process of the historical behavior path set of the sample entity provided by the first embodiment of the present application is shown in the figure. Figure 3 The flowchart of the acquisition process of the candidate object name set provided by the first embodiment of the present application is shown in the figure. Figure 4 The flowchart of obtaining the similarity of the two initial object names provided by the first embodiment of the present application is shown in the figure. Figure 5 The flowchart of obtaining the similar entity of the target entity provided by the first embodiment of the present application is shown in the figure. Figure 6 The flowchart of obtaining the target directed graph provided by the first embodiment of the present application is shown in the figure. Figure 7 The flowchart of determining the pushed information of the target entity provided by the first embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of the present application.

[0036] Embodiment one: According to the present embodiment, as shown in Figure 1 , a method for information pushing of historical behavior path is provided, and the method comprises the following steps: S100, obtaining a historical behavior path set of a sample entity; the historical behavior path set of the sample entity comprises historical behavior paths of a plurality of sample entities, any historical behavior path of a sample entity comprises a plurality of nodes and edges connecting the nodes, any node corresponds to a behavior, and the weight of any edge is the time interval between the occurrence of the two behaviors corresponding to the two nodes connected by the edge; the occurrence time of the behavior corresponding to the node located in front in any historical behavior path of a sample entity is earlier than the occurrence time of the behavior corresponding to the node located behind.

[0037] In the present embodiment, the sample entity is an entity (such as a user, an enterprise, etc.) participating in the collection of historical behavior data, the historical behavior path can reflect the behavior of the sample entity in the time dimension, one node in the historical behavior path represents one behavior, and any behavior comprises a behavior action and an object name, for example, purchasing a toothbrush, wherein purchasing is the behavior action and toothbrush is the object name; for example, applying for a high-tech enterprise, wherein applying is the behavior action and high-tech enterprise is the object name. As a specific implementation, the edge in the historical behavior path is a directed edge connecting adjacent nodes, and the weight is the time interval between the occurrence of the two behaviors (for example, the time interval between purchasing a toothbrush and purchasing toothpaste is 3 days). As a specific implementation, the node in the historical behavior path is the object name in the corresponding behavior, for example, the behavior corresponding to a node is purchasing a toothbrush, and the node is represented by the character toothbrush.

[0038] In the present embodiment, the historical behavior path set of the sample entity is pre-constructed. As a specific implementation, as shown in Figure 2 , the construction process of the historical behavior path set of the sample entity comprises: S110, obtaining a text corresponding to each candidate object name in a candidate object name set; the text corresponding to any candidate object name comprises a judgment condition for judging whether the behavior corresponding to the candidate object name occurs.

[0039] As a specific implementation, as shown in Figure 3 , the obtaining process of the candidate object name set comprises: S111, obtaining an initial object name set; the initial object name set comprises a plurality of initial object names.

[0040] As a specific embodiment, the initial object names are object names collected from the Internet; it should be understood that the initial object names are used to distinguish the candidate object names; both the initial object names and the candidate object names are object names, the difference is that the candidate object names are processed by deleting the repeated semantic names, and the initial object names are not processed by deleting the repeated semantic names, therefore, there may be object names with the same semantics in the initial object name set, and the number of initial object names is greater than or equal to the number of candidate object names. For example, the two initial object names with the same semantics are 2025 model intelligent robot and intelligent robot 2025 model.

[0041] S112, for any initial object name, obtaining the keywords and corresponding parts of speech in the initial object name.

[0042] Those skilled in the art know that the process of obtaining keywords and corresponding parts of speech is prior art, which will not be described here. For example, the initial object name is 2025 model intelligent robot, wherein the keywords are 2025 model, intelligent and robot, 2025 model is a time noun phrase, intelligent is an adjective, and robot is a noun.

[0043] S113, for any two initial object names, obtaining the similarity of the two initial object names according to the keywords and corresponding parts of speech of the two initial object names, and if the similarity of the two initial object names is greater than or equal to a preset object name similarity threshold, deleting one of the two initial object names in the initial object name set.

[0044] Optionally, the preset object name similarity threshold is an empirical value, for example, the preset object name similarity threshold is 0.9.

[0045] As a specific embodiment, if the number of keywords included in the two initial object names is the same, the similarity of the two initial object names is determined according to S1131-S1134; otherwise, the similarity of the two initial object names is determined as 0. As a specific embodiment, as shown in Figure 4 obtaining the similarity of the two initial object names according to the keywords and corresponding parts of speech of the two initial object names comprises: S1131, rearranging the keywords of a second initial object name according to the part-of-speech arrangement order of the keywords of a first initial object name; the first initial object name is one of the two initial object names, and the second initial object name is the other one of the two initial object names except the first initial object name.

[0046] For example, the first initial object name is a 2025 model intelligent robot, the keywords of the first initial object name are 2025 model, intelligent and robot, and the part-of-speech arrangement order of the keywords of the first initial object name is time noun phrase, adjective and noun; the second initial object name is intelligent robot 2025 model, the keywords of the second initial object name are intelligent, robot and 2025 model, the part-of-speech arrangement order of the keywords of the second initial object name is adjective, noun and time noun phrase, and the result of rearranging the keywords of the second initial object name according to the part-of-speech arrangement order of the keywords of the first initial object name is 2025 model, intelligent and robot.

[0047] In S1132, a semantic vector of the first initial object name is obtained according to the keyword order of the first initial object name; the semantic vector of the first initial object name is obtained by concatenating the semantic vectors corresponding to the keywords of the first initial object name according to the keyword order.

[0048] Those skilled in the art know that the process of converting keywords into corresponding semantic vectors is prior art, which will not be described here. The dimensions of the semantic vectors obtained after converting any keyword are the same.

[0049] In S1133, a semantic vector of the second initial object name is obtained according to the keyword order after rearranging the keywords of the second initial object name; the semantic vector of the second initial object name is obtained by concatenating the semantic vectors corresponding to the keywords of the second initial object name according to the keyword order after rearranging.

[0050] In S1134, the similarity between the semantic vector of the first initial object name and the semantic vector of the second initial object name is determined as the similarity between the two initial object names.

[0051] In this embodiment, the dimensions of the semantic vector of the first initial object name and the semantic vector of the second initial object name are the same. Those skilled in the art know that the process of determining the similarity between two vectors is prior art, which will not be described here.

[0052] Based on S1131-S1134, in this embodiment, the keywords of one of the two initial object names are reordered according to the part-of-speech order of the keywords of the other initial object name, and the similarity between the two initial object names is obtained by concatenating the semantic vectors corresponding to the keywords. Thus, the problem of inaccurate judgment of whether the two initial object names are similar due to different keyword orders is solved, and the accuracy of the judgment result of whether the two initial object names are similar is improved.

[0053] In S114, the updated initial object name set is determined as the candidate object name set.

[0054] In this embodiment, the similarity between any two initial object names in the updated initial object name set is less than the preset object name similarity threshold.

[0055] Based on S111-S114, the embodiment performs the repeated semantic name deletion processing on the initial object name set, so that the obtained candidate object name set is relatively concise, ensuring the uniqueness and semantic accuracy of the node, and enabling the historical behavior path to be more concise and accurate, thereby providing high-quality data for subsequent analysis.

[0056] S120, according to the text corresponding to the specified candidate object name, judging whether the specified sample entity has the behavior corresponding to the specified candidate object name, if so, according to the time when the specified sample entity has the behavior corresponding to the specified candidate object name, appending the node corresponding to the behavior to the historical behavior path of the specified sample entity; the specified sample entity is any sample entity, and the specified candidate object name is any candidate object name.

[0057] In this embodiment, the historical behavior path is constructed in time sequence, and the time sequence relationship of the behavior is preserved.

[0058] Based on S110-S120, the obtained historical behavior path set of the sample entity is more concise and accurate.

[0059] S200, comparing the preset type attribute value of the target entity with the preset type attribute value of the sample entity, obtaining the similar entity of the target entity; the attribute similarity between any similar entity and the target entity is greater than or equal to a preset attribute similarity threshold.

[0060] In this embodiment, the target entity is an entity that needs to push information (such as a current user or a target enterprise), and the preset type attribute is a key attribute of the entity (such as the age and region of the user, and the industry and scale of the enterprise). The preset type attribute is determined in advance and is known.

[0061] In this embodiment, the group closest to the characteristics of the target entity is selected from the sample entities, ensuring the relevance of subsequent behavior analysis. As a specific implementation, as shown in Figure 5 The comparison of the preset type attribute value of the target entity with the preset type attribute value of the sample entity to obtain the similar entity of the target entity includes: S210, obtaining the similarity between each preset type attribute value of the target entity and the corresponding preset type attribute value of the specified sample entity, if the similarity between a certain preset type attribute value of the target entity and the corresponding preset type attribute value of the specified sample entity is greater than or equal to a preset first similarity threshold, determining that the preset type attribute is a similar attribute of the target entity and the specified sample entity; the specified sample entity is any sample entity.

[0062] Optionally, the first preset similarity threshold is an empirical value, for example, the first preset similarity threshold is 0.9. As a specific embodiment, the similarity between different attribute values corresponding to any preset type attribute is constructed in advance, or a large language model is used to determine the similarity between different attribute values corresponding to the same preset type attribute.

[0063] S220, obtaining the number of similar attributes of the target entity and the specified sample entity, and determining the ratio of the number of similar attributes to the preset number of types as the attribute similarity of the target entity and the specified sample entity.

[0064] S230, if the attribute similarity of the target entity and the specified sample entity is greater than or equal to a preset attribute similarity threshold, the specified sample entity is determined as a similar entity of the target entity.

[0065] Optionally, the preset attribute similarity threshold is an empirical value, for example, the attribute similarity threshold is 0.8.

[0066] Based on S210-S230, the similar entity of the target entity can be accurately obtained.

[0067] In this embodiment, based on the similar entity, the behavior data of irrelevant entities can be avoided to be pushed, the pertinence of subsequent directed graph construction is improved, and the accuracy of pushed information is improved.

[0068] S300, obtaining the historical behavior path of the similar entity from the historical behavior path set of the sample entity, and obtaining a target directed graph according to the historical behavior path of the similar entity; any node in the target directed graph corresponds to an action, the first weight corresponding to any directed edge in the directed graph is the occurrence probability from the action corresponding to the tail node of the directed edge to the action corresponding to the head node, and the second weight corresponding to any directed edge in the directed graph is the occurrence time interval from the action corresponding to the tail node of the directed edge to the action corresponding to the head node.

[0069] In this embodiment, the target directed graph is obtained according to the historical behavior path of the similar entity, one node in the target directed graph corresponds to an action, and different nodes correspond to different actions; the directed edge in the target directed graph includes the occurrence probability and the occurrence time interval from the action corresponding to the tail node of the directed edge to the action corresponding to the head node. It should be understood that the tail node of the directed edge is the node connected by the directed edge and corresponding to the action occurring earlier in time, and the head node of the directed edge is the node connected by the directed edge and corresponding to the action occurring later in time.

[0070] As a specific embodiment, as shown in Figure 6 obtaining the target directed graph according to the historical behavior path of the similar entity includes: S310, traversing the historical behavior paths of all similar entities, extracting all nodes in each historical behavior path of each similar entity, and constructing a node set of a directed graph.

[0071] S320, constructing edges of the directed graph according to the historical behavior paths of the similar entities, including: for each historical behavior path of each similar entity, sequentially processing adjacent node pairs in the order of node positions, adding a directed edge between two nodes corresponding to each adjacent node pair, and adding a directed edge between the two nodes corresponding to each adjacent node pair from the node with the earlier position in the adjacent node pair to the node with the later position in the adjacent node pair.

[0072] In this embodiment, there are at most 2 directed edges between two nodes in the directed graph, and when there are 2 directed edges between two nodes in the directed graph, the directions of the 2 directed edges are opposite.

[0073] S330, obtaining a first weight corresponding to each directed edge in the directed graph; the first weight corresponding to any directed edge is the occurrence probability of the behavior corresponding to the tail node of the directed edge to the behavior corresponding to the head node; and the occurrence probability of the behavior corresponding to the tail node of any directed edge to the behavior corresponding to the head node is the ratio of the number of occurrences of the adjacent node pair corresponding to the directed edge in the historical behavior path of the similar entity to the number of occurrences of the tail node of the directed edge as the tail node.

[0074] In this embodiment, the adjacent node pair corresponding to any directed edge is the adjacent node pair satisfying the first condition, and the first condition is that the node with the earlier position included is the tail node of the directed edge and the node with the later position included is the head node of the directed edge.

[0075] S340, obtaining a second weight corresponding to each directed edge in the directed graph; the second weight corresponding to any directed edge is the occurrence time interval of the behavior corresponding to the tail node of the directed edge to the behavior corresponding to the head node; and the occurrence time interval of the behavior corresponding to the tail node of any directed edge to the behavior corresponding to the head node is the mean of the weights of the edges corresponding to the directed edge in the historical behavior path of the similar entity.

[0076] S350, determining the finally obtained directed graph as the target directed graph.

[0077] Based on S310-S350, this embodiment converts discrete behavior paths of different similar entities into a graph structure, quantifies the relevance and timeliness between behaviors through sending probability and occurrence time interval, and thus, based on the target directed graph, the correlation and time regularity between behaviors can be obtained, providing data support for subsequent accurate push information.

[0078] S400, determining push information of a target entity according to a target behavior of the target entity and the target directed graph.

[0079] In this embodiment, the target behavior is a behavior that has occurred to the target entity currently. For example, the target behavior is to purchase a toothbrush or to apply for a high-tech enterprise.

[0080] As a specific implementation, as shown in the figure, determining the push information of the target entity according to the target behavior of the target entity and the target directed graph comprises: Figure 7 S410, locating a node matching the target behavior of the target entity in the target directed graph.

[0081] In this embodiment, the node matching the target behavior of the target entity refers to a node whose corresponding behavior is the target behavior.

[0082] S420, if the locating is successful, obtaining a head node set of the matching node in the target directed graph.

[0083] In this embodiment, the head node set of the matching node in the target directed graph is all head nodes of directed edges in the target directed graph with the matching node as the tail node.

[0084] S430, for any head node in the head node set, obtaining a first weight of a directed edge between the matching node and the head node, and if the first weight is greater than or equal to a preset weight threshold, determining the push information of the target entity according to a behavior corresponding to the head node.

[0085] Optionally, the preset weight threshold is an empirical value, for example, the weight threshold is 0.7.

[0086] As an optional specific implementation, the push information of the target entity determined according to the behavior corresponding to the head node comprises an object name in the behavior corresponding to the head node. For example, in a user shopping scenario, the target behavior is to purchase a toothbrush, the behavior corresponding to the head node is to purchase toothpaste, and the push information of the target entity determined according to the behavior corresponding to the head node comprises toothpaste. The time for sending the push information to the target entity is a second weight of a directed edge from a node corresponding to the purchase of the toothbrush to a node corresponding to the purchase of the toothpaste.

[0087] ​As an optional implementation, the push information of the target entity determined according to the behavior corresponding to the head node includes the behavior corresponding to the head node and the occurrence time interval corresponding to the head node, the occurrence time interval corresponding to the head node being the second weight of the directed edge corresponding to the head node, the directed edge corresponding to the head node being the directed edge with the matched node as the tail node and the head node as the head node. For example, in the scenario of reporting a policy project of an enterprise, the target behavior is to report a high-tech enterprise, the behavior corresponding to the head node is to report a small giant enterprise, and the push information of the target entity determined according to the behavior corresponding to the head node includes reporting a small giant enterprise and a reporting time interval of 1 year.

[0088] As an implementation, if the first weight of the directed edge between the matched node and multiple head nodes in the head node set is greater than or equal to a preset weight threshold, the push information of the target entity is determined according to the behaviors corresponding to the multiple head nodes, or the push information of the target entity is determined according to the behaviors corresponding to the first largest weight of the head nodes in the multiple head nodes.

[0089] The embodiment obtains a historical behavior path set of a sample entity, and any historical behavior path of the sample entity can reflect behavior information of the sample entity and a time interval between behaviors. Based on the historical behavior path of the sample entity, the embodiment can obtain a time sequence feature and a logical relationship of the behavior of the sample entity. On this basis, the embodiment compares a preset type attribute value of a target entity with the sample entity, obtains a similar entity, and then constructs a target directed graph suitable for the target entity based on a historical behavior path of the similar entity. Based on the directed graph, an occurrence order relationship and a time interval relationship between different behaviors of the similar entity can be obtained, and then the target entity can be pushed more matched and more accurate personalized information through the directed graph and the target behavior of the target entity, which can improve the satisfaction of the target entity to the push information.

[0090] Embodiment two: The embodiment provides an electronic device, including a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the following steps when executing the computer program: obtain a historical behavior path set of a sample entity; the historical behavior path set of the sample entity includes historical behavior paths of a plurality of sample entities, any historical behavior path of the sample entity includes a plurality of nodes and edges connecting the nodes, any node corresponds to a behavior, and the weight of any edge is the occurrence time interval of the two behaviors corresponding to the two nodes connected by the edge; the occurrence time of the behavior corresponding to the node in the position in any historical behavior path of the sample entity is earlier than the occurrence time of the behavior corresponding to the node in the position.

[0091] The preset type attribute value of the target entity is compared with the preset type attribute value of the sample entity to obtain similar entities of the target entity; and the attribute similarity between any similar entity and the target entity is greater than or equal to a preset attribute similarity threshold.

[0092] The historical behavior path of the similar entity is obtained from the historical behavior path set of the sample entity, and a target directed graph is obtained according to the historical behavior path of the similar entity; any node in the target directed graph corresponds to an action, any directed edge in the directed graph corresponds to a first weight, which is the occurrence probability from the action corresponding to the tail node of the directed edge to the action corresponding to the head node, and any directed edge in the directed graph corresponds to a second weight, which is the occurrence time interval from the action corresponding to the tail node of the directed edge to the action corresponding to the head node.

[0093] The push information of the target entity is determined according to the target behavior of the target entity and the target directed graph.

[0094] Embodiment three: The embodiment provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the following steps: A historical behavior path set of a sample entity is obtained; the historical behavior path set of the sample entity includes historical behavior paths of a plurality of sample entities, any historical behavior path of a sample entity includes a plurality of nodes and edges connecting the nodes, any node corresponds to an action, and any edge has a weight, which is the occurrence time interval of two actions corresponding to two nodes connected by the edge; the occurrence time of the action corresponding to the node located in front in any historical behavior path of a sample entity is earlier than the occurrence time of the action corresponding to the node located behind.

[0095] The preset type attribute value of the target entity is compared with the preset type attribute value of the sample entity to obtain similar entities of the target entity; and the attribute similarity between any similar entity and the target entity is greater than or equal to a preset attribute similarity threshold.

[0096] The historical behavior path of the similar entity is obtained from the historical behavior path set of the sample entity, and a target directed graph is obtained according to the historical behavior path of the similar entity; any node in the target directed graph corresponds to an action, any directed edge in the directed graph corresponds to a first weight, which is the occurrence probability from the action corresponding to the tail node of the directed edge to the action corresponding to the head node, and any directed edge in the directed graph corresponds to a second weight, which is the occurrence time interval from the action corresponding to the tail node of the directed edge to the action corresponding to the head node.

[0097] The push information of the target entity is determined according to the target behavior of the target entity and the target directed graph.

[0098] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0099] Although some specific embodiments of the present application have been described in detail by way of example with reference to the accompanying drawings, it is to be understood that the above examples are intended to be illustrative only and are not intended to limit the scope of the present application. Those skilled in the art will understand that various modifications can be made to the embodiments without departing from the scope and spirit of the present application. The scope of the present application is defined by the appended claims.

Claims

1. An information push processing method based on historical behavior paths, characterized in that, The method includes the following steps: Obtain the historical behavior path set of the sample entity; the historical behavior path set of the sample entity includes the historical behavior paths of several sample entities, the historical behavior path of any sample entity includes several nodes and edges connecting the nodes, each node corresponds to one behavior, and the weight of any edge is the time interval between the occurrence of the two behaviors corresponding to the two nodes connected by the edge; in the historical behavior path of any sample entity, the behavior corresponding to the node that is earlier occurs earlier than the behavior corresponding to the node that is later. The target entity's preset type attribute value is compared with the sample entity's preset type attribute value to obtain similar entities to the target entity; the attribute similarity between any similar entity and the target entity is greater than or equal to the preset attribute similarity threshold. The historical behavior paths of similar entities are obtained from the historical behavior path set of sample entities, and a target directed graph is obtained based on the historical behavior paths of similar entities; any node in the target directed graph corresponds to one behavior, the first weight of any directed edge in the directed graph is the probability of occurrence from the behavior corresponding to the tail node of the directed edge to the behavior corresponding to the head node, and the second weight of any directed edge in the directed graph is the time interval from the occurrence of the behavior corresponding to the tail node of the directed edge to the occurrence of the behavior corresponding to the head node. The push information for the target entity is determined based on the target entity's target behavior and the target directed graph.

2. The information push processing method based on historical behavior paths according to claim 1, characterized in that, Obtaining the target directed graph based on the historical behavior paths of the similar entities includes: Traverse the historical behavior paths of all similar entities, extract all nodes in each historical behavior path of each similar entity, and construct a node set of a directed graph; The edges of a directed graph are constructed based on the historical behavior paths of similar entities, including: for each historical behavior path of each similar entity, adjacent node pairs are processed sequentially according to the order of node positions, so that a directed edge is added between the two nodes corresponding to each adjacent node pair, and the directed edge added between the two nodes corresponding to each adjacent node pair points from the node with the first position in the adjacent node pair to the node with the second position in the adjacent node pair. Obtain the first weight corresponding to each directed edge in the directed graph; the first weight corresponding to any directed edge is the probability of the behavior corresponding to the tail node of the directed edge occurring to the behavior corresponding to the head node; the probability of the behavior corresponding to the tail node of any directed edge occurring to the behavior corresponding to the head node is the ratio of the number of times the adjacent node pair corresponding to the directed edge appears in the historical behavior path of similar entities to the number of times the tail node of the directed edge appears as the tail node. Obtain the second weight corresponding to each directed edge in the directed graph; the second weight corresponding to any directed edge is the time interval between the occurrence of the behavior corresponding to the tail node of the directed edge and the occurrence of the behavior corresponding to the head node; the time interval between the occurrence of the behavior corresponding to the tail node of any directed edge and the occurrence of the behavior corresponding to the head node is the average weight of the edges corresponding to the directed edge in the historical behavior path of similar entities. The resulting directed graph is then identified as the target directed graph.

3. The information push processing method based on historical behavior paths according to claim 2, characterized in that, The push information for the target entity is determined based on the target entity's target behavior and the target directed graph, including: Locate the node in the target directed graph that matches the target entity's target behavior; If the location is successful, obtain the set of head nodes of the matched node in the target directed graph; For any head node in the set of head nodes, obtain the first weight of the directed edge between the matched node and the head node. If the first weight is greater than or equal to a preset weight threshold, determine the push information of the target entity based on the behavior corresponding to the head node.

4. The information push processing method based on historical behavior paths according to claim 3, characterized in that, The push information of the target entity determined based on the behavior corresponding to the head node includes the behavior corresponding to the head node and the occurrence time interval corresponding to the head node. The occurrence time interval corresponding to the head node is the second weight of the directed edge corresponding to the head node. The directed edge corresponding to the head node is a directed edge with the matched node as the tail node and the head node as the head node.

5. The information push processing method based on historical behavior paths according to claim 1, characterized in that, The target entity's preset type attribute values ​​are compared with the sample entity's preset type attribute values ​​to obtain similar entities, including: Obtain the similarity between each preset type attribute value of the target entity and the corresponding preset type attribute value of the specified sample entity. If the similarity between a preset type attribute value of the target entity and the corresponding preset type attribute value of the specified sample entity is greater than or equal to a preset first similarity threshold, then the preset type attribute is determined to be a similarity attribute between the target entity and the specified sample entity; the specified sample entity is any sample entity. Obtain the number of similar attributes between the target entity and the specified sample entity, and determine the attribute similarity between the target entity and the specified sample entity by the ratio of the number of similar attributes to the number of preset types; If the attribute similarity between the target entity and the specified sample entity is greater than or equal to the preset attribute similarity threshold, then the specified sample entity is determined as a similar entity to the target entity.

6. The information push processing method based on historical behavior paths according to claim 1, characterized in that, Any behavior includes a behavior action and an object name. The process of constructing the historical behavior path set of a sample entity includes: Retrieve the text corresponding to each candidate object name in the candidate object name set; the text corresponding to any candidate object name includes a condition for determining whether the behavior corresponding to that candidate object name has occurred. Based on the text corresponding to the specified candidate object name, determine whether the specified sample entity has performed the behavior corresponding to the specified candidate object name. If it has, then based on the time when the specified sample entity performed the behavior corresponding to the specified candidate object name, append the node corresponding to the behavior corresponding to the specified candidate object name to the historical behavior path of the specified sample entity. The specified sample entity can be any sample entity, and the specified candidate object name can be any candidate object name.

7. The information push processing method based on historical behavior paths according to claim 6, characterized in that, The process of obtaining the candidate object name set includes: Obtain an initial object name set; the initial object name set includes several initial object names; For any initial object name, retrieve the keywords and their corresponding parts of speech in that initial object name; For any two initial object names, the similarity between the two initial object names is obtained based on their keywords and corresponding parts of speech. If the similarity between the two initial object names is greater than or equal to the preset object name similarity threshold, then one of the two initial object names in the initial object name set is deleted. The updated initial set of object names is used as the candidate set of object names.

8. The information push processing method based on historical behavior paths according to claim 7, characterized in that, The similarity between the two initial object names is obtained based on their keywords and corresponding parts of speech, including: The keywords of the second initial object name are rearranged according to the part-of-speech order of the keywords of the first initial object name; the first initial object name is one of the two initial object names, and the second initial object name is the other of the two initial object names besides the first initial object name; The semantic vector of the first initial object name is obtained according to the keyword order of the first initial object name; the semantic vector of the first initial object name is obtained by concatenating the semantic vectors corresponding to the keywords of the first initial object name according to the keyword order. The semantic vector of the second initial object name is obtained by rearranging the keywords of the second initial object name according to the keyword order; the semantic vector of the second initial object name is obtained by concatenating the semantic vectors corresponding to the keywords of the second initial object name according to the rearranged keyword order. The similarity between the semantic vectors of the first initial object name and the semantic vectors of the second initial object name is determined as the similarity between the two initial object names.

9. 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 computer program, it implements the information push processing method based on historical behavior paths as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the information push processing method based on historical behavior paths as described in any one of claims 1 to 8.

Images