CN116304352A

CN116304352A - Message pushing method, device, equipment and storage medium

Info

Publication number: CN116304352A
Application number: CN202310323866.5A
Authority: CN
Inventors: 房志瑞; 英继越
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2023-03-29
Filing date: 2023-03-29
Publication date: 2023-06-23

Abstract

The disclosure provides a message pushing method, a device, equipment and a storage medium, which can be applied to the technical field of artificial intelligence or the technical field of finance. The method comprises the following steps: acquiring message content to be pushed and a user identifier of a target user, wherein the target user is a user for receiving the message content; according to the user identification, acquiring a historical interaction data set of the target user and the application terminal from a buried data base of the application terminal; performing cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set to determine a target pushing moment and a target pushing channel from the historical interaction data set; and pushing the message content to the target user according to the target pushing moment and the target pushing channel.

Description

Message pushing method, device, equipment and storage medium

Technical Field

The present disclosure relates to the field of artificial intelligence technology or financial technology, and in particular, to a method, an apparatus, a device, and a storage medium for pushing a message.

Background

With the development of internet technology, accurate pushing of messages has gradually become a main requirement for improving user experience. In the process of implementing the inventive concept of the present disclosure, the inventors found that the following problems generally exist in the related art: the existing message pushing mode is generally to set pushing content, pushing channels, pushing time and the like manually and subjectively by service personnel, so that a large number of messages are not really touched to clients, the requirement of accurate pushing of the messages cannot be met, and finally the messages are low in touch rate, low in clicking rate and low in conversion rate.

Disclosure of Invention

In view of the foregoing, the present disclosure provides a message pushing method, apparatus, device, storage medium, and program product.

One aspect of the present disclosure provides a message pushing method, including: acquiring message content to be pushed and a user identifier of a target user, wherein the target user is a user for receiving the message content; acquiring a historical interaction data set of the target user and the application terminal from a buried point database of the application terminal according to the user identification; based on the data density and the data distance of the historical interaction data set, carrying out cluster analysis on the historical interaction data set so as to determine a target pushing moment and a target pushing channel from the historical interaction data set; and pushing the message content to the target user according to the target pushing time and the target pushing channel.

According to an embodiment of the present disclosure, the performing cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set to determine a target pushing time and a target pushing channel from the historical interaction data set includes: and inputting the historical interaction data set into a clustering model, and outputting the target pushing moment and the target pushing channel, wherein the clustering model is constructed based on the data density and the data distance.

According to the embodiment of the disclosure, the application terminal is configured with a plurality of interaction channels; the inputting the historical interaction data set into a clustering model, and outputting the target pushing time and the target pushing channel includes: extracting the interaction time, the interaction time length and the interaction times of the target user and each interaction channel from the historical interaction data; processing the interaction time, the interaction time and the interaction times to obtain feature data for mapping the interaction characteristics of the target user; and inputting the characteristic data into the cluster model, and outputting the target pushing channel and the target pushing moment which have corresponding relations.

According to an embodiment of the present disclosure, inputting the feature data into the cluster model, and outputting the target push channel and the target push time having a correspondence relationship includes: inputting the characteristic data into the cluster model, and outputting an initial pushing channel and an initial pushing moment with corresponding relations; and screening and outputting the target pushing channel and the target pushing moment from the initial pushing channel and the initial pushing moment according to the expected activity index and the expected stability index.

According to an embodiment of the present disclosure, the above method further includes: after pushing the message content to the target user according to the target pushing moment and the target pushing channel, collecting interaction data of the target user and the message content; evaluating the clustering model according to the actual liveness index and the actual stability index of the interaction data; and under the condition that the evaluation passes, assigning a preset model identifier to the clustering model.

According to an embodiment of the present disclosure, the pushing the message content to the target user according to the target pushing moment and the target pushing channel includes: generating a combined message according to the target pushing moment and the target pushing channel; splicing the combined message and the message content to obtain a final push message; and pushing the target user to the target user through the target pushing channel at the target pushing time.

Another aspect of the present disclosure further provides a training method of a cluster model, including: according to the sample user identification, a sample historical interaction data set of a sample user and the application terminal is obtained from a buried data base of the application terminal, wherein the sample historical interaction data set comprises a plurality of sample historical interaction data; inputting the sample history interaction data set, a preset cluster number K and a neighborhood radius of a cluster into an initial cluster model, wherein K is a positive integer; in the initial clustering model, selecting initial clustering points based on a preset rule, and generating a cluster center point set based on the initial clustering points; iteratively selecting the rest cluster points except the initial cluster points based on the sample data density and the sample data distance until the number of the cluster points in the cluster center point set is K; and circularly calculating the average value of the sample history interaction data in each clustering cluster, and adjusting the parameters of the initial clustering model according to the average value of the sample history interaction data until the average value of the training sample data is not changed, so as to obtain the clustering model.

According to an embodiment of the present disclosure, in the foregoing initial clustering model, selecting an initial clustering point based on a preset rule includes: in the initial clustering model, calculating the sample data density of the sample history interaction data based on a density function to obtain a sample density queue of the sample history interaction data set; and (3) adjusting the neighborhood radius, and taking at least one clustering point meeting a density threshold in the sample density queue as an initial clustering point.

According to an embodiment of the present disclosure, the iteratively selecting the remaining cluster points other than the initial cluster point based on the sample data density and the sample data distance includes: randomly selecting sample points in the sample history interaction data set; generating a pre-clustering cluster according to the sample points; constructing a target value function according to the shortest sample data distance between the sample points and the initial clustering points and the sample data density of the pre-clustering clusters; and under the condition that the target value function meets the preset condition, confirming the sample points as the rest clustering points.

Another aspect of the present disclosure also provides a message pushing apparatus, including: the first acquisition module is used for acquiring message content to be pushed and a user identifier of a target user, wherein the target user is a user for receiving the message content; the second acquisition module is used for acquiring a historical interaction data set of the target user and the application terminal from a buried point database of the application terminal according to the user identification; the analysis module is used for carrying out cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set so as to determine a target pushing moment and a target pushing channel from the historical interaction data set; and the pushing module is used for pushing the message content to the target user according to the target pushing moment and the target pushing channel.

Another aspect of the present disclosure also provides an electronic device, including: one or more processors; and the storage device is used for storing one or more programs, wherein the one or more programs are executed by the one or more processors, and the one or more processors are caused to execute the message pushing method or the training method of the clustering model.

Another aspect of the present disclosure also provides a computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the above-described message pushing method or training method of a cluster model.

Another aspect of the disclosure also provides a computer program product, including a computer program, where the computer program, when executed by a processor, implements the above-mentioned message pushing method or training method of a cluster model.

According to the message pushing method, the device, the equipment and the storage medium provided by the embodiment of the disclosure, the message content to be pushed and the user identification of the target user are obtained; acquiring a historical interaction data set from a buried data base of the application terminal according to the user identification; based on the data density and the data distance of the historical interaction data set, carrying out cluster analysis on the historical interaction data set, and determining a target pushing moment and a target pushing channel from the historical interaction data set; and pushing the message content to the target user according to the target pushing moment and the target pushing channel. Because the historical interaction data of the target user are combined in the process of pushing the message content to the target user, and the historical interaction data set is subjected to clustering analysis based on the data density and the data distance, the target pushing time and the target pushing channel associated with the target user can be obtained, and the message content can be pushed through the target pushing channel at the target pushing time, so that the accurate pushing of the message can be realized, the problem that a large number of messages in the related technology are not really touched to the client is at least partially overcome, and the technical effects of improving the touch rate, the click rate and the conversion rate of the message are further achieved.

Drawings

The foregoing and other objects, features and advantages of the disclosure will be more apparent from the following description of embodiments of the disclosure with reference to the accompanying drawings, in which:

fig. 1 schematically illustrates an application scenario of a message pushing method and apparatus according to an embodiment of the present disclosure;

fig. 2 schematically illustrates a flow chart of a message pushing method according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates a flow chart of a training method of a cluster model implemented in accordance with the present disclosure;

fig. 4 schematically illustrates a frame diagram of a message pushing system according to an embodiment of the present disclosure;

FIG. 5 schematically illustrates a system architecture diagram of an intelligent routing module according to an embodiment of the present disclosure;

fig. 6 schematically illustrates a block diagram of a message pushing device according to an embodiment of the present disclosure;

FIG. 7 schematically illustrates a block diagram of a training apparatus of a cluster model, according to an embodiment of the disclosure; and

fig. 8 schematically illustrates a block diagram of an electronic device adapted to implement a message pushing method according to an embodiment of the disclosure.

Detailed Description

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the concepts of the present disclosure.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.

Where expressions like at least one of "A, B and C, etc. are used, the expressions should generally be interpreted in accordance with the meaning as commonly understood by those skilled in the art (e.g.," a system having at least one of A, B and C "shall include, but not be limited to, a system having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

In the technical scheme of the disclosure, the related data (such as including but not limited to personal information of a user) are collected, stored, used, processed, transmitted, provided, disclosed, applied and the like, all conform to the regulations of related laws and regulations, necessary security measures are adopted, and the public welcome is not violated.

The accurate pushing of various messages such as reminding and trading has become an important means and medium for realizing value conversion for client applications. Currently, message pushing is mainly classified into instant messaging triggered by clients and active business messaging configured by business personnel. Aiming at the message pushing of the active service class, the service personnel manually set the pushing content, pushing channel and pushing time; the subjectivity is strong in the message sending time and channel selection, so that a large number of messages are not really delivered to clients, and finally the problems of message low delivery, low clicking and low conversion are caused.

In view of this, the embodiment of the disclosure provides an intelligent routing system based on an improved clustering algorithm by performing cluster data analysis on the use behavior of the application terminal used by the user, which can automatically identify the active time period of using the application terminal and the channel of the application terminal most commonly used by the target user every day, then automatically push the message to the application terminal channel most commonly used by the client at the optimal time, further improve the success rate of the message reaching the target user, and effectively improve the click rate and conversion rate of the message.

The message pushing method comprises the steps of obtaining message content to be pushed and a user identifier of a target user, wherein the target user is a user receiving the message content; according to the user identification, acquiring a historical interaction data set of the target user and the application terminal from a buried data base of the application terminal; performing cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set to determine a target pushing moment and a target pushing channel from the historical interaction data set; and pushing the message content to the target user according to the target pushing moment and the target pushing channel.

It should be noted that, the method and the device for pushing the message determined by the embodiments of the present disclosure may be used in the field of artificial intelligence technology or the field of financial technology, and may also be used in any field other than the field of artificial intelligence technology or the field of financial technology, and the application field of the method and the device for pushing the message determined by the embodiments of the present disclosure is not limited.

In the technical scheme of the disclosure, the related data (such as including but not limited to user personal information, video recording, metadata extraction, personalized data extraction and the like) are collected, stored, used, processed, transmitted, provided, disclosed, applied and the like, all conform to the regulations of related laws and regulations, necessary security measures are adopted, and the public welcome is not violated.

Fig. 1 schematically illustrates an application scenario of a message pushing method and apparatus according to an embodiment of the present disclosure.

As shown in fig. 1, an application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is a medium used to provide a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user may interact with the server 105 via the network 104 using at least one of the first terminal device 101, the second terminal device 102, the third terminal device 103, to receive pushed messages etc. Various communication client applications, such as a mobile banking, a personal internet banking, a e-mail, an e-shopping, an e-living, etc. financial class application, a shopping class application, a web browser application, a search class application, an instant messaging tool, a mailbox client, social platform software, etc. (only examples) may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103.

The first terminal device 101, the second terminal device 102, the third terminal device 103 may be any of a variety of electronic devices with support for web browsing including, but not limited to, smartphones, tablets, laptop and desktop computers, and the like.

The server 105 may be a server providing various services, such as pushing messages to users using the first terminal device 101, the second terminal device 102, the third terminal device 103. Specifically, the server may obtain the message content to be pushed and a user identifier of a target user, where the target user is a user who receives the message content; according to the user identification, acquiring a historical interaction data set of the target user and the application terminal from a buried data base of the application terminal; performing cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set to determine a target pushing moment and a target pushing channel from the historical interaction data set; and pushing the message content to the target user according to the target pushing moment and the target pushing channel.

It should be noted that, the message pushing method provided by the embodiments of the present disclosure may be generally performed by the server 105. Accordingly, the message pushing device provided by the embodiments of the present disclosure may be generally disposed in the server 105. The message pushing method provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and is capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and/or the server 105. Accordingly, the message pushing apparatus provided by the embodiments of the present disclosure may also be provided in a server or a server cluster that is different from the server 105 and is capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and/or the server 105.

It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

The message pushing method of the disclosed embodiment will be described in detail with reference to fig. 2 to 4 based on the application scenario described in fig. 1.

Fig. 2 schematically illustrates a flow chart of a message pushing method according to an embodiment of the present disclosure.

As shown in fig. 2, the message pushing method of this embodiment includes operations S201 to S204.

In operation S201, the message content to be pushed and the user identification of the target user, which is the user who receives the message content, are acquired.

In operation S202, a historical interaction data set of the target user and the application terminal is obtained from a buried database of the application terminal according to the user identification.

In operation S203, cluster analysis is performed on the historical interaction data set based on the data density and the data distance of the historical interaction data set to determine a target pushing moment and a target pushing channel from the historical interaction data set.

In operation S204, message content is pushed to the target user according to the target push time and the target push channel.

According to the embodiment of the disclosure, the message content to be pushed may be reminding information for reminding the user, or article information, service information and the like which need to be recommended to the user.

According to embodiments of the present disclosure, the target user may be a registered user in various client applications, which may include mobile banking, personal internet banking, e-commerce, and the like. The target user may receive the pushed message content through the client application.

According to embodiments of the present disclosure, the user identification of the target user may refer to the number of the user. The user identification may be obtained by the server from a registered user list of the client application, or the user identification may be imported from outside into the server. It can be appreciated that the process of obtaining the user identification is performed after the authorization of the user, and necessary security measures are taken, without violating the public welfare.

According to an embodiment of the disclosure, the embedded point database may include interaction data of the user and the application terminal, which is acquired according to a pre-embedded point, for example, a time when the user opens a certain application terminal each time, a pushing time of a history message, a clicking time of the history message, an operation duration on a certain application terminal, a number of clicks of different pages, a retention time of different pages, and the like. It can be further understood that basic information of the target user can be acquired while the interactive data is acquired, so that the target pushing moment and the target pushing channel corresponding to the target user can be more accurately found.

According to embodiments of the present disclosure, data density may refer to a ratio of the number of data points within a cluster to the area of the cluster; the data distance may refer to the distance between the cluster point of one cluster and the cluster point of another cluster. Specifically, taking the click time of the history message as an example, according to the collected history click time of one or more target users, one history click time can be selected as an initial clustering point in a region with more concentrated time in a plurality of history click times, when the next clustering point is selected, the data distance from the initial clustering point should be selected as far as possible, and meanwhile, more history click times are still around the point, and under the condition that the two conditions are met, the point can be used as the next clustering point. Based on this, cluster analysis of historical click times and historical interaction datasets can be achieved.

According to the embodiment of the disclosure, the target pushing time may be the optimal pushing time, which may be understood that the probability that the target user clicks on the message at this time is higher. The target push channel may correspond to an optimal push time at which the probability of the user selecting this push channel is higher. After the cluster analysis is completed, the target pushing moment and the corresponding target pushing channel can be selected from the cluster clusters with larger density. Illustratively, the target user would click on the push message from the cell phone bank with a high probability at 17 points and the target user would click on the push message from the fuse link with a high probability at 16 points.

According to the embodiment of the disclosure, after the target pushing moment and the target pushing channel are determined, message content can be pushed to a target user through the target pushing channel at the target pushing moment.

According to an embodiment of the present disclosure, operation S203 may further include the following operations: and inputting the historical interaction data set into a clustering model, and outputting the target pushing moment and the target pushing channel, wherein the clustering model is constructed based on the data density and the data distance.

According to embodiments of the present disclosure, the cluster model may refer to a modified K-means (K-means clustering algorithm ) model. The traditional K-means model has the problems that the result is unstable, the local optimal solution is easy to generate and the optimal division cannot be obtained in the practical application. In order to improve the stability of information clustering of a K-means model on message pushing time and pushing channels, optimize a division effect and improve division efficiency, the embodiment of the disclosure provides a K-means algorithm based on time sample density-distance, which can optimize a center point selection process and solve a problem of local optimal solution to a certain extent, and is more efficient in efficiency than a traditional algorithm, and time and calculation overhead caused by randomly generating centers are avoided.

According to embodiments of the present disclosure, in density-based clustering algorithms, the density distribution of data samples may be considered as a factor of preference, the principle being that each cluster of clusters of the clustering result tends to have a dense set of samples. In the K-means model, the selection strategy of the cluster center to be determined is to select a position which is far away from the cluster center which is already determined as far as possible, so that the obtained clustering result can ensure that the outside has high enough distinction degree and the inside has enough similarity, thereby ensuring that the clustering result is more accurate. Embodiments of the present disclosure integrate two factors of data density (sample density) and data distance (distance of cluster center or distance of cluster point) for clustering, which will be described in detail below.

Fig. 3 schematically illustrates a flow chart of a training method of a cluster model implemented according to the present disclosure.

As shown in fig. 3, the training method of the cluster model may include operations S301 to S305.

In operation S301, a sample historical interaction data set of the sample user and the application terminal is obtained from a buried database of the application terminal according to the sample user identification, wherein the sample historical interaction data set includes a plurality of sample historical interaction data.

In operation S302, a sample history interaction dataset, a preset cluster number K, and a neighborhood radius of a cluster are input to an initial cluster model, where K is a positive integer.

In operation S303, in the initial cluster model, an initial cluster point is selected based on a preset rule, and a cluster center point set is generated based on the initial cluster point.

In operation S304, based on the sample data density and the sample data distance, the rest of cluster points except the initial cluster point are iteratively selected until the number of cluster points in the cluster center point set is K.

In operation S305, the average value of the sample history interaction data in each cluster is calculated in a circulating manner, and the parameters of the initial cluster model are adjusted according to the average value of the sample history interaction data until the average value of the training sample data is not changed, so as to obtain the cluster model.

According to embodiments of the present disclosure, a sample historical interaction data set, sample historical interaction data, sample data distance, sample data density, etc. may be understood as a historical interaction data set, historical interaction data, data distance, data density, etc. described above, so a "sample" is utilized to distinguish in order to facilitate the training process of training a cluster model.

According to an embodiment of the present disclosure, operation S303 may further include the following operations: in the initial clustering model, calculating sample data density of sample history interaction data based on a density function to obtain a sample density queue of a sample history interaction data set; and (3) taking at least one clustering point meeting the density threshold value in the sample density queue as an initial clustering point by adjusting the neighborhood radius. For example, the cluster points with higher density are selected as the initial cluster points, so that the time cost and the calculation cost caused by randomly generating the cluster points of the traditional model can be effectively reduced. The density threshold can be adaptively adjusted according to actual needs.

According to an embodiment of the present disclosure, operation S304 may further include the following operations: randomly selecting sample points in a sample history interaction data set; generating a pre-clustering cluster according to the sample points; constructing a target value function according to the shortest sample data distance between the sample point and the initial clustering point and the sample data density of the pre-clustering cluster; and under the condition that the target value function meets the preset condition, confirming the sample points as the rest clustering points. The target value function may be as shown in equation (1).

obj＝α×min{dist(p,q)}×β×|N _ε (q)| (1)

Wherein p is C _i ，C _i ∈{C ₁ ,C ₂ ,...C _l }，q∈C _j ，

C can represent a cluster point set, alpha and beta are model parameters, dist (p, q) is a data distance of p, q, and min { dist (p, q) } can represent the shortest data distance and the minimum value of the data distance; p may be a cluster point determined, for example, an initial cluster point; q may be the next cluster point to be determined, e.g. a sample point; n (N) _ε (q) may represent the density of points q to be determined. The purpose of the arrangement of equation (1) is to integrate the density and distance as core factors when selecting the next cluster point. The above-described determination of the target value function can be understood as the product of the minimum value of the distances from the points to be determined and the density of the points to be determined among all the divided points.

According to the embodiment of the disclosure, the preset conditions can be adaptively adjusted according to actual needs. For example, when the target value obtained from the target value function is the maximum value, the sample point corresponding to the maximum target value may be set as the next cluster point.

The model training method provided by the embodiment of the present disclosure is described below through an exemplary embodiment, but the embodiment of the present disclosure is not limited thereto.

According to the embodiment of the disclosure, eps and epsilon are set as the neighborhood radius; minPts is a threshold value of the number of samples contained in the neighborhood of the core point when the cluster core point is defined, and can be understood as a density threshold value; the epsilon neighborhood is for any one sample point,

Wherein dist (p, q) is the correlation distance of p, q, p can be a determined cluster point, q can be the next cluster point to be determined, and D is a training sample set; sample density is for

|N _ε (x) I, is the density of point x in the epsilon neighborhood; core points (cluster points satisfying the density threshold) refer to the values for +.>

If |N _ε (x)|>MinPts, then x is a core point.

The specific training process for the improved K-means model is as follows.

In Euclidean space R ^m Input sample set of

And clustering the cluster number K, calculating a distance matrix of the sample set through a formula (2):

D _n×n ＝{dist(x _i ,x _j )|1≤i≤n,1≤j≤n,i≠j} (2)

wherein D represents a distance matrix, n represents the number of sample data, n is a positive integer, x _i 、x _j Different sample data may be represented separately.

The neighborhood radius parameter epsilon of the algorithm is input, and sample points meeting the density threshold are used as initial clustering points by adjusting epsilon neighborhood, for example, sample points with higher density can be selected as initial first clustering points.

Calculating density information of samples in the data set according to a density formula to generate a sample density queue Que _ρ Wherein, the density formula can be according to the ratio of the number of sample data in a single cluster to the area of the cluster.

A cluster center point set V is set to store core points, which can be understood as storing cluster points satisfying a density threshold.

Random selection of density queue Que _ρ Meeting |N _ε (x)|>Element x of MinPts _v Put into the cluster center point set V.

Will belong to x _v Sample points within epsilon neighborhood from Que _ρ Deleting; at Que _ρ Selecting a target value function

Maximum sample point x _v Adding the cluster points into V as the next cluster points; and iteratively executing the two steps until the number of elements in the cluster center point set V is K.

For the obtained cluster center point set V, using K-means algorithm to perform Que _ρ The rest elements in the tree are all divided, and Euclidean distance is used as a dividing standard; and (3) for the divided K class cluster results, averaging the samples in each class cluster to serve as a new class cluster center, and updating the new class cluster center into the cluster center point set V. The two steps are repeatedly executed until the average value of the samples in each cluster is no longer changed, i.e. the cluster center point is no longer changed, and a cluster center point set V can be output.

According to the embodiment of the disclosure, global density information exploration is performed on an input data set in a training process, and a epsilon neighborhood is used to select a sample point with higher density as an initial first clustering point, so that time expenditure and calculation expenditure caused by the traditional randomly generated clustering points can be effectively reduced; by adjusting the parameter epsilon, all global data can be considered, and only local data is not focused any more, so that the problem of optimal solution localization caused by focusing only on the local data can be effectively solved.

According to the embodiment of the disclosure, for selecting other cluster centers, a dual factor determination strategy of data distance and data density is adopted, and model parameters alpha and beta are set at the same time, so that the tendency of the strategy to the two factors can be adjusted according to a data set. And after the centers of all the class clusters are determined, classifying the rest non-center samples into class clusters. And finally, correcting the cluster center in an average value mode to obtain a clustering result. The correction method can refer to a traditional K-means model, and is not described herein.

Compared with the traditional K-means model, the improved K-means model provided by the embodiment of the disclosure provides a more reasonable and efficient means in the initial determination process of the cluster center, solves the problem of high time complexity in the traditional mode, and solves the problem of local optimal solution to a certain extent.

According to the embodiment of the disclosure, the application terminal can be configured with a plurality of interaction channels, such as mobile banking, personal internet banking, e-commerce, e-life and the like. Inputting the historical interaction data set into the cluster model, outputting the target pushing moment and the target pushing channel may include the operations of: extracting the interaction time, the interaction duration and the interaction times of the target user and each interaction channel from the historical interaction data; processing the interaction time, the interaction time and the interaction times to obtain feature data for mapping the interaction characteristics of the target user; and inputting the characteristic data into the cluster model, and outputting a target pushing channel and target pushing time with corresponding relations.

Specifically, the process of obtaining the target push time may include the following operations: the imported list of the target user and the operation log file (such as a mobile phone log login information backup table) of the target user on the application terminal are obtained, the target user is matched with the operation log file of the target user on the application terminal through the unique identifier of the user number (user identifier), and basic information of the target user and interaction information of the target user are extracted. The interaction information may include: the method comprises the steps of carrying out data preprocessing and characteristic engineering processing on collected data, such as classifying text data and numerical data, carrying out numerical conversion processing on the text data, carrying out homogenization processing on all extracted or converted data and the like, so as to obtain interactive data with a uniform numerical format. The missing value and the abnormal value can be removed in the pretreatment and processing processes, so that the accuracy of obtaining the target pushing moment is improved. And inputting the sorted interaction data into the improved K-means model, and outputting the pushing moment by adjusting different cluster numbers K and parameters alpha and beta of the model. The verification set achieves acceptable prediction effect in the model by adjusting parameters. It should be noted that, the list of the target user, the operation log file of the target user on the application terminal, the extracted basic information and the interaction information of the target user are all the operations executed after the authorization of the target user is obtained, and corresponding security measures are adopted, and the public welfare is not violated.

The process of obtaining the target push channel may include the following operations: and acquiring the imported target user list and the operation log file of the target user on the application terminal, and matching the operation log file of the target user and the operation log file of the target user on the application terminal through the unique identification of the user. And extracting basic information and interaction information of the target user on each application terminal. The channel of the application terminal may include: the mobile banking, the personal online banking, the e-line melting, the e-purchase melting, the e-life melting and the like, so that the selection of a transmission channel is convenient for business personnel, parameterization processing can be performed on the channel of an application terminal on a system, for example, a parameter code is given to the application terminal: personal internet banking (302), mobile banking (303), e-commerce purchase (309), e-commerce link (312) and the like; and determining the channel of the application terminal which needs to be analyzed by the channel model through parameter configuration.

The obtained interactive data can be processed and feature data for mapping the interactive characteristics of the target user can be obtained, for example, average daily access times of the target user, average daily online time length of the target user, variance of access times of the target user, and the like, which are associated with each interactive channel, are obtained, the three indexes are given the same or different weights, the three indexes can be input into the improved K-means model, and push time and push channels with corresponding relations are output by adjusting different cluster numbers K and parameters alpha and beta of the model. By adjusting the parameters, the validation set achieves acceptable predictive effects in the present model, which may include to achieve a desired click rate, a desired touch rate, and the like. It should be noted that, the list of the target user, the operation log file of the target user on the application terminal, the extracted basic information and the interaction information of the target user are all the operations executed after the authorization of the target user is obtained, and corresponding security measures are adopted, and the public welfare is not violated.

According to an embodiment of the present disclosure, inputting feature data into a cluster model, outputting a target pushing channel and a target pushing time having a correspondence relationship may further include the operations of: inputting the characteristic data into a cluster model, and outputting an initial pushing channel and initial pushing time with corresponding relations; and screening and outputting the target pushing channel and the target pushing moment from the initial pushing channel and the initial pushing moment according to the expected activity index and the expected stability index.

According to the embodiment of the disclosure, the liveness index may be an overall click rate of the push message, and the stability index may be understood as whether the number of clicks of the target user in a certain period of time is stable. The expected activity index and the expected stability index can be set by a business person according to business requirements, and if the business needs tend to improve the overall click rate of the push message, the target push time and the target push channel can be selected based on the activity index, or the K-means model can be trained based on the activity index; if the business needs to tend to improve the stability of the click rate of the push message, the target push time and the target push channel can be selected based on the stability index, or the K-means model can be trained based on the stability index. The model may also be trained or the output result selected based on the average duty cycle of the desired activity index and the desired temperature index.

For example, the output in the K-means model may be an initial pushing time and a pushing channel, then, according to the service requirement, whether the activity index is dominant or the stability index is dominant is determined, and after the primary index is determined, one or more target pushing times and target pushing channels with corresponding relations may be selected from the initial pushing time and the initial pushing channel. Or training is directly carried out according to the proportion occupied by the expected activity index and the expected stability index in the model training process, so that the target pushing moment and the target pushing channel under the expected activity index and the expected stability index can be obtained.

According to an embodiment of the present disclosure, the above method may further include the following operations: after pushing the message content to the target user according to the target pushing moment and the target pushing channel, collecting interaction data of the target user and the message content; evaluating the clustering model according to the actual liveness index and the actual stability index of the interaction data; and under the condition that the evaluation passes, assigning a preset model identifier to the clustering model.

According to the embodiment of the disclosure, in order to verify the training effect of the K-means model, after the actual pushing is completed according to the output result of the model, the actual interaction data of the target user and the message content can be acquired, the actual liveness index and the actual stability index are calculated according to the actual interaction data, the actual liveness index and the actual stability index are respectively compared with the expected liveness index and the expected stability index, and when the comparison result shows that the actual result is consistent with the expected result, the situation that the reliability of the current K-means model is higher is shown, and the model identification with high reliability is given. And under the condition that the comparison result shows that the actual result does not accord with the expected result, the current K-means model needs to be continuously adjusted, and the model to be adjusted is given with the identification.

According to an embodiment of the present disclosure, operation S204 may further include the following operations: generating a combined message according to the target pushing moment and the target pushing channel; splicing the combined message and the message content to obtain a final push message; and pushing the target user to the target user through the target pushing channel at the target pushing time.

According to the embodiment of the disclosure, a target pushing moment output by a K-means model and a target pushing channel are spliced, a combined message about a pushing mode is generated, and the combined message and the message content are spliced to obtain a final complete pushing message; and finally, according to the push message, completing the push of the message content through the target push channel at the target push time.

According to the embodiment of the disclosure, on the basis of the existing message sending link, the judgment on the user sending time and the sending channel is increased, so that the message is automatically pushed to the most commonly used application terminal channel of the client at the optimal time, and the efficient access of the message is realized.

Fig. 4 schematically illustrates a frame diagram of a message pushing system according to an embodiment of the present disclosure.

As shown in fig. 4, the message pushing system 400 includes a message maintenance module 401, an intelligent routing module 402, a message processing module 403, and a sending module 404, wherein the modified K-means model may be provided in the intelligent routing module 402. When the message pushing is completed according to the embodiment of the present disclosure, a target user group file and an original message content file corresponding to the pushed message may be obtained from the message maintenance module 401; processing the target user information through the intelligent routing module 402; the K-means model of the intelligent routing module 402 may output a corresponding optimal push time message and an optimal push channel message; outputting an optimal push time message and an optimal push channel message aiming at each user, and combining the optimal push time message and the optimal push channel message to obtain a combined message; the message processing module 403 splices the combined message obtained by the intelligent routing module 402 with the original message content file, generates and outputs a final message, and pushes the final message to the sending module 404; the sending module 404 may send the corresponding message to the most appropriate receiving channel for the client at the best push time.

Fig. 5 schematically illustrates a system architecture diagram of an intelligent routing module according to an embodiment of the present disclosure.

The intelligent routing module 402 may further include a data import submodule 4021, a model processing submodule 4022, and a model output submodule 4023. Specifically, the intelligent routing module 402 mainly includes, when processing: the intelligent routing module 402 acquires a target user information file and an operation log file of a user on an application terminal; positioning operation information stored in an operation log of the user in the application terminal according to the user number, and extracting related indexes; the operation behavior data of the user at the application terminal is imported into a channel selection model and a time pushing model of the model processing sub-module 4022 to be subjected to model processing; the model processing submodule 4022 can output an optimal pushing channel message and an optimal pushing time message through the model output submodule 4023 after model processing; and carrying out combination matching on the two messages according to the user to generate an intelligent push combined file finally associated with the target user.

According to the embodiment of the present disclosure, the content of the message pushing system 400 may refer to the relevant content in operations S201 to S204 or operations S301 to S305, which are not described herein.

The message pushing system provided by the embodiment of the disclosure overcomes subjectivity in message sending time and sending channel selection during manual sending, and solves the problem of invalid triggering of the message. After confirming push targets and push contents, business personnel perform cluster analysis on the using behavior habit of target users in application terminals by an intelligent routing module, calculate the most commonly accessed time period in a day and the most commonly accessed application terminals in the day of the users, push messages to channels with highest contact rate at optimal time according to the output result of a k-means model, and improve the contact power of the messages.

It should be noted that, unless there is an execution sequence between different operations or an execution sequence between different operations in technical implementation, the execution sequence between multiple operations may be different, and multiple operations may also be executed simultaneously in the embodiment of the disclosure.

Based on the message pushing method, the disclosure further provides a message pushing device. The device will be described in detail below in connection with fig. 6.

Fig. 6 schematically shows a block diagram of a message pushing device according to an embodiment of the present disclosure.

As shown in fig. 6, the message pushing device 600 of this embodiment includes a first acquisition module 610, a second acquisition module 620, an analysis module 630, and a pushing module 640.

The first obtaining module 610 is configured to obtain the message content to be pushed and a user identifier of a target user, where the target user is a user who receives the message content.

And the second obtaining module 620 is configured to obtain, according to the user identifier, a historical interaction data set of the target user and the application terminal from the buried data base of the application terminal.

The analysis module 630 is configured to perform cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set, so as to determine the target pushing moment and the target pushing channel from the historical interaction data set.

And the pushing module 640 is used for pushing the message content to the target user according to the target pushing moment and the target pushing channel.

According to an embodiment of the present disclosure, the analysis module may further include an input sub-module.

And the input sub-module is used for inputting the historical interaction data set into a clustering model and outputting the target pushing moment and the target pushing channel, wherein the clustering model is constructed based on the data density and the data distance.

According to an embodiment of the present disclosure, the input sub-module may include an extraction unit, a processing unit, an input unit.

The extraction unit is used for extracting the interaction time, the interaction duration and the interaction times of the target user and each interaction channel from the historical interaction data.

And the processing unit is used for processing the interaction time, the interaction time and the interaction times to obtain feature data for mapping the interaction characteristics of the target user.

And the input unit is used for inputting the characteristic data into the cluster model and outputting a target pushing channel and target pushing time with corresponding relations.

According to an embodiment of the present disclosure, the input unit may further include an input subunit, a filtering subunit.

And the input subunit is used for inputting the characteristic data into the cluster model and outputting an initial pushing channel and an initial pushing moment which have corresponding relations.

And the screening subunit is used for screening and outputting the target pushing channel and the target pushing moment from the initial pushing channel and the initial pushing moment according to the expected activity index and the expected stability index.

According to an embodiment of the disclosure, the message pushing device may further include an acquisition module, an evaluation module, and an assignment module.

And the acquisition module is used for acquiring interaction data of the target user and the message content after pushing the message content to the target user according to the target pushing moment and the target pushing channel.

And the evaluation module is used for evaluating the clustering model according to the actual activity index and the actual stable index of the interaction data.

And the giving module is used for giving a preset model identifier to the clustering model under the condition that the evaluation passes.

According to an embodiment of the present disclosure, the pushing module may further include a generating sub-module, a splicing sub-module, and a pushing sub-module.

The generation sub-module is used for generating a combined message according to the target pushing moment and the target pushing channel.

And the splicing sub-module is used for splicing the combined message and the message content to obtain a final push message.

And the pushing sub-module is used for pushing the target users through the target pushing channels at the target pushing time.

Fig. 7 schematically illustrates a block diagram of a training apparatus of a cluster model according to an embodiment of the disclosure.

As shown in fig. 7, the training apparatus 700 of this embodiment may include a third acquisition module 710, an input module 720, a selection module 730, an iteration module 740, and a loop module 750.

And a third obtaining module 710, configured to obtain, from the embedded database of the application terminal, a sample historical interaction data set of the sample user and the application terminal according to the sample user identifier, where the sample historical interaction data set includes a plurality of sample historical interaction data.

The input module 720 is configured to input the sample history interaction dataset, the preset cluster number K, and the neighborhood radius of the cluster to the initial cluster model, where K is a positive integer.

The selecting module 730 is configured to select an initial cluster point based on a preset rule in the initial cluster model, and generate a cluster center point set based on the initial cluster point.

And the iteration module 740 is used for iteratively selecting the rest clustering points except the initial clustering points based on the sample data density and the sample data distance until the number of the clustering points in the clustering cluster center point set is K.

And the circulation module 750 is used for circularly calculating the average value of the sample history interaction data in each cluster, and adjusting the parameters of the initial cluster model according to the average value of the sample history interaction data until the average value of the training sample data is not changed any more, so as to obtain the cluster model.

According to an embodiment of the disclosure, the selection module may further include a calculation unit and an adjustment unit.

And the calculating unit is used for calculating the sample data density of the sample history interaction data based on the density function in the initial clustering model to obtain a sample density queue of the sample history interaction data set.

And the adjusting unit is used for taking at least one clustering point meeting the density threshold value in the sample density queue as an initial clustering point by adjusting the neighborhood radius.

According to an embodiment of the present disclosure, the iteration module may further include a selection unit, a generation unit, a construction unit, and a confirmation unit.

And the selecting unit is used for randomly selecting sample points in the sample history interaction data set.

And the generating unit is used for generating the pre-clustering cluster according to the sample points.

And the construction unit is used for constructing a target value function according to the shortest sample data distance between the sample point and the initial clustering point and the sample data density of the pre-clustering cluster.

And the confirming unit is used for confirming the sample points as the rest clustering points under the condition that the target value function meets the preset condition.

Any of the first acquisition module 610, the second acquisition module 620, the analysis module 630, and the push module 640, or any of the third acquisition module 710, the input module 720, the selection module 730, the iteration module 740, and the loop module 750 may be combined in one module to be implemented, or any of the modules may be split into a plurality of modules, according to embodiments of the present disclosure. Alternatively, at least some of the functionality of one or more of the modules may be combined with at least some of the functionality of other modules and implemented in one module. According to embodiments of the present disclosure, at least one of the first acquisition module 610, the second acquisition module 620, the analysis module 630, and the push module 640, or at least one of the third acquisition module 710, the input module 720, the selection module 730, the iteration module 740, and the loop module 750, may be implemented at least in part as hardware circuitry, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system-on-chip, a system-on-substrate, a system-on-package, an Application Specific Integrated Circuit (ASIC), or may be implemented in hardware or firmware in any other reasonable manner of integrating or packaging circuitry, or in any one of or a suitable combination of three implementations of software, hardware, and firmware. Alternatively, at least one of the first acquisition module 610, the second acquisition module 620, the analysis module 630 and the push module 640 or at least one of the third acquisition module 710, the input module 720, the selection module 730, the iteration module 740, the loop module 750 may be at least partially implemented as a computer program module, which may perform the respective functions when being executed.

It should be noted that, in the embodiment of the present disclosure, the message pushing device portion corresponds to the message pushing method portion in the embodiment of the present disclosure, and the description of the message pushing device portion specifically refers to the message pushing method portion, which is not described herein again.

As shown in fig. 8, an electronic device 800 according to an embodiment of the present disclosure includes a processor 801 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 802 or a program loaded from a storage section 808 into a Random Access Memory (RAM) 803. The processor 801 may include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or an associated chipset and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), or the like. The processor 801 may also include on-board memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing the different actions of the method flows according to embodiments of the disclosure.

In the RAM 803, various programs and data required for the operation of the electronic device 800 are stored. The processor 801, the ROM 802, and the RAM 803 are connected to each other by a bus 804. The processor 801 performs various operations of the method flow according to the embodiments of the present disclosure by executing programs in the ROM 802 and/or the RAM 803. Note that the program may be stored in one or more memories other than the ROM 802 and the RAM 803. The processor 801 may also perform various operations of the method flows according to embodiments of the present disclosure by executing programs stored in the one or more memories.

According to an embodiment of the present disclosure, the electronic device 800 may also include an input/output (I/O) interface 805, the input/output (I/O) interface 805 also being connected to the bus 804. The electronic device 800 may also include one or more of the following components connected to an input/output (I/O) interface 805: an input portion 806 including a keyboard, mouse, etc.; an output portion 807 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and a speaker; a storage section 808 including a hard disk or the like; and a communication section 809 including a network interface card such as a LAN card, a modem, or the like. The communication section 809 performs communication processing via a network such as the internet. The drive 810 is also connected to an input/output (I/O) interface 805 as needed. A removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 810 as needed so that a computer program read out therefrom is mounted into the storage section 808 as needed.

The present disclosure also provides a computer-readable storage medium that may be embodied in the apparatus/device/system described in the above embodiments; or may exist alone without being assembled into the apparatus/device/system. The computer-readable storage medium carries one or more programs which, when executed, implement methods in accordance with embodiments of the present disclosure.

According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example, but is not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to embodiments of the present disclosure, the computer-readable storage medium may include ROM 802 and/or RAM 803 and/or one or more memories other than ROM 802 and RAM 803 described above.

Embodiments of the present disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowcharts. The program code, when executed in a computer system, causes the computer system to implement the message pushing method provided by embodiments of the present disclosure.

The above-described functions defined in the system/apparatus of the embodiments of the present disclosure are performed when the computer program is executed by the processor 801. The systems, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, or the like. In another embodiment, the computer program may also be transmitted, distributed, and downloaded and installed in the form of a signal on a network medium, and/or from a removable medium 811 via a communication portion 809. The computer program may include program code that may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.

In such an embodiment, the computer program may be downloaded and installed from a network via the communication section 809, and/or installed from the removable media 811. The above-described functions defined in the system of the embodiments of the present disclosure are performed when the computer program is executed by the processor 801. The systems, devices, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

According to embodiments of the present disclosure, program code for performing computer programs provided by embodiments of the present disclosure may be written in any combination of one or more programming languages, and in particular, such computer programs may be implemented in high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. Programming languages include, but are not limited to, such as Java, c++, python, "C" or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

Those skilled in the art will appreciate that the features recited in the various embodiments of the disclosure and/or in the claims may be provided in a variety of combinations and/or combinations, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, the features recited in the various embodiments of the present disclosure and/or the claims may be variously combined and/or combined without departing from the spirit and teachings of the present disclosure. All such combinations and/or combinations fall within the scope of the present disclosure.

The embodiments of the present disclosure are described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the embodiments cannot be used advantageously in combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be made by those skilled in the art without departing from the scope of the disclosure, and such alternatives and modifications are intended to fall within the scope of the disclosure.

Claims

1. A message pushing method, comprising:

acquiring message content to be pushed and a user identifier of a target user, wherein the target user is a user receiving the message content;

According to the user identification, acquiring a historical interaction data set of the target user and the application terminal from a buried data base of the application terminal;

performing cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set to determine a target pushing moment and a target pushing channel from the historical interaction data set; and

and pushing the message content to the target user according to the target pushing moment and the target pushing channel.

2. The method of claim 1, wherein the performing cluster analysis on the historical interaction dataset based on the data density and the data distance of the historical interaction dataset to determine target pushing moments and target pushing channels from the historical interaction dataset comprises:

and inputting the historical interaction data set into a clustering model, and outputting the target pushing moment and the target pushing channel, wherein the clustering model is constructed based on the data density and the data distance.

3. The method of claim 2, wherein the application terminal is configured with a plurality of interaction channels;

inputting the historical interaction data set into a clustering model, and outputting the target pushing moment and the target pushing channel comprises the following steps:

Extracting the interaction time, the interaction duration and the interaction times of the target user and each interaction channel from the historical interaction data;

processing the interaction time, the interaction time and the interaction times to obtain feature data for mapping the interaction characteristics of the target user;

and inputting the characteristic data into the cluster model, and outputting the target pushing channel and the target pushing moment which have corresponding relations.

4. The method of claim 3, wherein the inputting the feature data into the cluster model, outputting the target push channel and the target push time having a correspondence, comprises:

inputting the characteristic data into the cluster model, and outputting an initial pushing channel and an initial pushing moment with corresponding relations;

and screening and outputting the target pushing channel and the target pushing moment from the initial pushing channel and the initial pushing moment according to the expected activity index and the expected stability index.

5. The method of claim 2, further comprising:

after pushing the message content to the target user according to the target pushing moment and the target pushing channel, collecting interaction data of the target user and the message content;

Evaluating the clustering model according to the actual liveness index and the actual stability index of the interaction data; and

and under the condition that the evaluation passes, assigning a preset model identifier to the clustering model.

6. The method of claim 1, wherein the pushing the message content to the target user according to the target push time and the target push channel comprises:

generating a combined message according to the target pushing moment and the target pushing channel;

splicing the combined message and the message content to obtain a final push message; and

and pushing the target pushing channel to the target user at the target pushing moment.

7. A method of training a cluster model, comprising:

according to the sample user identification, a sample historical interaction data set of a sample user and an application terminal is obtained from a buried data base of the application terminal, wherein the sample historical interaction data set comprises a plurality of sample historical interaction data;

inputting the sample history interaction data set, a preset cluster number K and a neighborhood radius of a cluster to an initial cluster model, wherein K is a positive integer;

In the initial clustering model, selecting initial clustering points based on a preset rule, and generating a cluster center point set based on the initial clustering points;

iteratively selecting the rest clustering points except the initial clustering points based on the sample data density and the sample data distance until the number of the clustering points in the clustering cluster center point set is K;

and circularly calculating the average value of the sample history interaction data in each clustering cluster, and adjusting the parameters of the initial clustering model according to the average value of the sample history interaction data until the average value of the training sample data is not changed, so as to obtain the clustering model.

8. The method of claim 7, wherein the selecting initial cluster points based on a preset rule in the initial cluster model comprises:

in the initial clustering model, calculating sample data density of the sample historical interaction data based on a density function to obtain a sample density queue of the sample historical interaction data set;

and using at least one clustering point meeting a density threshold in the sample density queue as an initial clustering point by adjusting the neighborhood radius.

9. The method of claim 7, wherein iteratively selecting remaining cluster points other than the initial cluster point based on sample data density and sample data distance comprises:

Randomly selecting sample points in the sample history interaction data set;

generating a pre-clustering cluster according to the sample points;

constructing a target value function according to the shortest sample data distance between the sample point and the initial clustering point and the sample data density of the pre-clustering cluster;

and under the condition that the target value function meets a preset condition, the sample points are confirmed to be the rest clustering points.

10. A message pushing device, comprising:

the first acquisition module is used for acquiring message content to be pushed and a user identifier of a target user, wherein the target user is a user for receiving the message content;

the second acquisition module is used for acquiring a historical interaction data set of the target user and the application terminal from a buried data base of the application terminal according to the user identification;

the analysis module is used for carrying out cluster analysis on the historical interaction data set based on the data density and the data distance of the historical interaction data set so as to determine a target pushing moment and a target pushing channel from the historical interaction data set; and

and the pushing module is used for pushing the message content to the target user according to the target pushing moment and the target pushing channel.

11. An electronic device, comprising:

one or more processors;

storage means for storing one or more programs,

wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to perform the method of any of claims 1-6 or 7-9.

12. A computer readable storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to perform the method according to any of claims 1-6 or 7-9.

13. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 6 or 7 to 9.