WO2021259007A1 - 基于行为轨迹的数据推送方法、系统和计算机设备 - Google Patents

基于行为轨迹的数据推送方法、系统和计算机设备 Download PDF

Info

Publication number
WO2021259007A1
WO2021259007A1 PCT/CN2021/097278 CN2021097278W WO2021259007A1 WO 2021259007 A1 WO2021259007 A1 WO 2021259007A1 CN 2021097278 W CN2021097278 W CN 2021097278W WO 2021259007 A1 WO2021259007 A1 WO 2021259007A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
data
trajectory
feature vector
push
Prior art date
Application number
PCT/CN2021/097278
Other languages
English (en)
French (fr)
Inventor
张�杰
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021259007A1 publication Critical patent/WO2021259007A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9536Search customisation based on social or collaborative filtering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9537Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/01Social networking
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/50Network services
    • H04L67/55Push-based network services

Definitions

  • the embodiments of the present application relate to the field of data analysis, and in particular, to a data pushing method, system, computer device, and computer-readable storage medium based on behavior trajectory.
  • social network services have become one of the important Internet services.
  • Social network services enable people to communicate and interact with friends without leaving home, and at the same time make more friends to expand their circle of friends.
  • Internet service providers like Weibo, Twitter, and Facebook are all committed to providing better dating services.
  • an embodiment of the present application provides a data pushing method based on behavior trajectory, and the method steps include:
  • an embodiment of the present application also provides a data push system based on behavior trajectory, including:
  • the receiving module is used to receive the data push instruction triggered by the target user through the target user terminal;
  • An obtaining module configured to obtain target historical behavior data of the target user from the behavior log system according to the data push instruction
  • a construction module for constructing the target trajectory graph of the target user according to the target historical behavior data
  • An extraction module configured to extract the target trajectory feature vector of the target user according to the target trajectory graph
  • a recall module configured to input the target trajectory feature vector into a recall model, so as to recall multiple initial push data from multiple users through the recall model;
  • the classification module is used to input the multiple initial push data into the classification model
  • the sorting module is used to sort the plurality of initial push data in association relationship through the classification model, so as to obtain a plurality of target push data;
  • the push module is used to push the multiple target push data to the front end for display.
  • an embodiment of the present application further provides a computer device, the computer device including a memory, a processor, and computer-readable instructions stored on the memory and running on the processor, so When the computer-readable instructions are executed by the processor, the following steps are implemented:
  • an embodiment of the present application further provides a computer-readable storage medium having computer-readable instructions stored in the computer-readable storage medium, and the computer-readable instructions may be executed by at least one processor, So that the at least one processor executes the following steps:
  • the behavior trajectory-based data push method, system, computer equipment, and computer-readable storage medium provided in the embodiments of this application provide users with a more accurate and personalized data push method; Constructing a trajectory graph, and matching the acquaintance degree of the user's trajectory through the classification model and the regression model, enriches the user's data recommendation conditions, and improves the accuracy and efficiency of the data recommendation.
  • FIG. 1 is a schematic flowchart of a data pushing method based on behavior trajectory according to an embodiment of this application.
  • Figure 2 is a target trajectory diagram of the data pushing method based on behavior trajectory of the application.
  • Fig. 3 is a full trajectory diagram of the data pushing method based on behavior trajectory of the application.
  • Fig. 4 is a structural diagram of the Skip-gram model of the data push method based on behavior trajectory of the application.
  • Fig. 5 is a flowchart of the recall phase of the data push method based on behavior trajectory of the application.
  • FIG. 6 is a schematic diagram of program modules of Embodiment 2 of the data pushing system based on behavior trajectory of this application.
  • FIG. 7 is a schematic diagram of the hardware structure of the third embodiment of the computer equipment of this application.
  • the computer device 2 will be used as an execution subject for exemplary description.
  • FIG. 1 shows a flow chart of the steps of the method for pushing data based on behavior trajectory in an embodiment of the present application. It can be understood that the flowchart in this method embodiment is not used to limit the order of execution of the steps.
  • the following exemplarily describes the computer device 2 as the execution subject. details as follows.
  • Step S100 Receive a data push instruction triggered by the target user through the target user terminal.
  • the data push instruction may be triggered by a data push component in the data push system.
  • the target user can access the data push system through the target user terminal, and click the data push button in the data push system to trigger the data push component; wherein, when the data push component is triggered, all The data push system can receive a data push instruction.
  • Step S102 Obtain target historical behavior data of the target user from the behavior log system according to the data push instruction.
  • the target user needs to log in to the data push system before accessing the data push system.
  • the data push system can obtain the target user identity information that currently triggers the data push component according to the data push instruction. After being confirmed, the data push system can obtain the target historical behavior data of the target user from the behavior log system according to the target user identity information.
  • the target historical behavior data includes the target user’s historical visit and stay location (latitude and longitude) and Historical access time (working days/non-working days).
  • the data push system may also obtain customized behavior data of the target user according to the target user's identity information.
  • the customized behavior data includes the target user's customized visit location (latitude and longitude) and visit time. (Working days/non-working days).
  • Step S104 Construct a target trajectory graph of the target user according to the target historical behavior data.
  • the step S104 may further include:
  • Step S104a Obtain multiple target location data from the target historical behavior data: l 1 , l 2 , l 3 ... l n and multiple target time data corresponding to the multiple target location data: t 1 , t 2 , t 3 ... t n .
  • l 1 , l 2 and l n are the first target location data, the second target location data, and the nth target location data, respectively, t 1 , t 2 And t n are the target time data corresponding to the first target location data, the target time data corresponding to the second target location data, and the target time data corresponding to the nth target location data.
  • the target historical behavior data includes multiple target location data corresponding to multiple visit locations visited by the target user: l 1 , l 2 , l 3 ... l n and the multiple targets Multiple target time data corresponding to the location data: t 1 , t 2 , t 3 ... t n .
  • the nodes in Fig. 2 indicate the locations of visits, and the arrows indicate the order of visits. Therefore, the visit order of the user is l 1 , l 2 , l 1 , l 2 , l 3 .
  • all the trajectories of all people are put into the graph to form a full trajectory graph, as shown in FIG. 3.
  • Step S106 Extract the target trajectory feature vector of the target user according to the target trajectory graph.
  • step S106 may further include:
  • Step S106a extract the target location feature vector of each target location according to each target location data and the target time data corresponding to each target location data to obtain multiple target location feature vectors E 1 , E 2 , E 3 .. .E n; step S106b, based on the plurality of eigenvectors target site, the target user determines a target track feature vector u: And upload the target trajectory feature vector to the blockchain, where E i is the ith target location feature vector of the target user u, and E u is the target trajectory feature vector of the target user u.
  • uploading the target trajectory feature vector to the blockchain can ensure its security and fairness and transparency to users.
  • the blockchain referred to in this example is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain is essentially a decentralized database. It is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • Step S108 Input the target trajectory feature vector into a recall model, so as to recall multiple initial push data from multiple users through the recall model.
  • multiple initial push data are recalled from multiple users through the recall model according to the target trajectory vector.
  • the recall is to lock a small part of the candidate set from tens of millions of data, and the locked candidate set has been used as a preliminary screening of the recommendation list.
  • a common recall method is the collaborative filtering algorithm, and this method is only suitable for scenarios with fewer users, and is not suitable for scenarios with tens of millions of data.
  • the method of Graph embedding network representation learning, also known as graph embedding
  • graph embedding can be adopted first to convert each person into a vector, and the similarity of the two vectors is calculated through the recall model.
  • a high similarity indicates that the two people’s similarity is high. The higher the trajectory similarity is, the more likely it is a friend relationship.
  • the process of the recall phase is shown in Figure 5.
  • the trajectory similarity is judged. The higher the similarity, the more likely the two are to be related.
  • the multiple users with the highest similarity are used as the initial push According to the data, the number of users with the highest similarity is a preset number.
  • the initial push data may be users who are acquainted with the target user's behavior trajectory.
  • the step S108 further includes: step S108a, judging whether the target user is a special user; and step S108b, if the target user is a special user, calculating the non-appearance according to a predetermined rule Vector of locations.
  • the special user is a user whose part or all of the trajectory graph has not been trained by a Skip-gram (neural network structure) model.
  • POI Point of Information
  • POI is common location information in the geographic system, which represents the category of a certain location, which may be a gourmet store, a clothing store, a fitness store, and so on. Crawling Dianping’s data, POI is divided into 17 categories, including “food”, “education school”, “institutional group”, “car”, “entertainment and leisure”, “life service”, “sports and fitness”, and “real estate”.
  • Each location obtained through Skip-gram model training represents a certain type of POI, so the embedding (continuous vector) vectors of the 17 POIs are divided into different categories: Among them, when calculating the vector of a place that does not appear, only the POI type of the new place needs to be judged, and then the embedding vector of this type of POI is used as an approximate replacement.
  • the method may further include:
  • Step S200 Obtain multiple historical behavior data corresponding to multiple users;
  • Step S202 Extract the trajectory feature vector of each user based on the multiple historical behavior data;
  • Step S204 Use the multiple trajectory feature vectors as pre-acquired to be trained
  • the input of the recall model, and the trajectory feature vector that has a similar trajectory to each trajectory feature vector is used as the output of the recall model to be trained to train the recall model to be trained until the loss function converges;
  • step S206 The trajectory feature vector is used as the input of the trained recall model to obtain multiple acquaintance trajectory feature vectors corresponding to each trajectory feature vector;
  • step S208 determining the accuracy of the acquaintance trajectory feature vector recalled by each trajectory feature vector And whether the recall rate reaches the detection accuracy threshold and the detection recall threshold. If it reaches, the recall model to be trained after training is the initial recall model.
  • the recall model may determine the trajectory graphs of the multiple users according to the multiple historical behavior data: And extract the trajectory feature vector of each of the multiple users to obtain multiple trajectory feature vectors corresponding to the multiple users.
  • Step S110 Input the multiple initial push data into the classification model.
  • the multiple initial push data are only similar to the target user's trajectory.
  • this solution adopts The multiple initial push data are input into the classification model to obtain multiple target push data that have a strong association with the target user.
  • the target push data may be users who have a strong association with the target user. For example, if two users appear together in a subway station or a railway station, the similarity is higher. But not to prove that they are friends, it is possible that they will pass by. Therefore, further classification is required.
  • step S112 the multiple initial push data are sorted in association relationship through the classification model to obtain multiple target push data.
  • the classification model may sort the plurality of initial push data based on predetermined factors, and use the plurality of initial push data ranked higher as the target push data.
  • the preset factors include position entropy, time interval sequence, and time dimension.
  • the size of the association relationship between the target user and the plurality of initial push data may be judged according to the location entropy, and the size of the location entropy may be judged according to the popularity of the area, for example, a train station Many people have appeared together in such popular places; and if two people often appear in a non-popular area, such as a certain community, it means that they have a strong relationship. Therefore, in the classification process, it is necessary to determine the popularity of the places visited by the two persons together to determine the location entropy.
  • the numerator of p l (u) is the number of times that user u has been to location l, and the denominator is the number of times that all users have been to location l.
  • the locations with higher location entropy are as follows. It can be found that the higher location entropy is the railway station or subway station with a larger number of people. As shown in Table 1:
  • Table 2 lists 4 pairs of users, namely pair1, pair2, pair3, pair4. Each pair of users has visited 4 different addresses: l 1 , l 2 , l 3 , l 4 , and 4 of pair1 can be found Address location entropy is small, so "co-occurrence location location entropy" and "co-occurrence frequency location entropy" are large, and there may be a relationship between the two.
  • the addresses visited by the two users of pair4 are both relatively large, which may be a certain subway station or business district. Therefore, the "location entropy of co-occurrence location” and “location entropy of co-occurrence frequency" of the two users are relatively small. It might not matter.
  • the size of the association relationship between the target user and the plurality of initial push data may also be determined according to the time interval sequence, and the time interval sequence is used to determine that the two users visit the same place
  • the time interval of, where the time interval is 0 means that two people visit a certain place at the same time.
  • both users u and u′ have visited the place l, and the visit behavior of u is ⁇ u,t u,1 ,l>, ⁇ u,t u,2 ,l>,... ⁇ u,t u,m ,l>, the access behavior of u′ is ⁇ u,t u′,1 ,l>, ⁇ u,t u′,2 ,l >,... ⁇ u,t u′,n ,l>, which means that both u and u′ visit location l at multiple times. Then you can construct the maximum time interval sequence and the minimum time interval sequence of two persons, respectively
  • Lu , u′ are the locations visited by both users u and u′.
  • the size of the association relationship between the target user and the multiple initial push data can also be judged according to the time dimension.
  • features can be established at different time latitudes, and the time is divided into working days. , Non-working days and all time. Therefore, there are 15 features in total, as shown in Table 3:
  • Step S114 Push the multiple target push data to the front end for display.
  • the front end may be written in HTML and Javascript
  • the HTML is a standard markup language for creating web pages
  • the JavaScript is a scripting language belonging to the network.
  • Front-end and back-end interactions are written using Flask and Ajax. Flask is a lightweight web application framework. Ajax is a web development technology for creating interactive web applications. Update.
  • multiple target push data can be obtained based on the target trajectory feature vector. Specifically, by inputting the target trajectory feature vector into the trained recall model, and then input the result of the recall model output The multiple target push data can be obtained in the classification model.
  • Uploading the target trajectory feature vector to the blockchain can ensure its security and fairness and transparency to users.
  • the user equipment can download the target trajectory feature vector from the blockchain to verify whether the push data of multiple targets has been tampered with.
  • the blockchain referred to in this example is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain is essentially a decentralized database. It is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • FIG. 6 is a schematic diagram of program modules of Embodiment 2 of the data pushing system based on behavior trajectory of this application.
  • the data push system 20 based on the behavior trajectory may include or be divided into one or more program modules, and the one or more program modules are stored in a storage medium and executed by one or more processors to complete the application, And can realize the above-mentioned data push method based on behavior trajectory.
  • the program module referred to in the embodiments of the present application refers to a series of computer program instruction segments capable of completing specific functions, and is more suitable than the program itself to describe the execution process of the data push system 20 based on the behavior trajectory in the storage medium. The following description will specifically introduce the functions of each program module in this embodiment:
  • the receiving module 200 is used to receive a data push instruction triggered by a target user through the target user terminal.
  • the obtaining module 202 is configured to obtain target historical behavior data of the target user from the behavior log system according to the data push instruction.
  • the construction module 204 is configured to construct the target trajectory graph of the target user according to the target historical behavior data.
  • the extraction module 206 is configured to extract the target trajectory feature vector of the target user according to the target trajectory graph.
  • the extraction module 206 is further configured to: extract the target location feature vector of each target location according to each target location data and the target time data corresponding to each target location data to obtain multiple target location feature vectors E 1, E 2, E 3 ... E n; and
  • determine the target trajectory feature vector of the target user u And upload the target trajectory feature vector to the blockchain, where E i is the ith target location feature vector of the target user u, and E u is the target trajectory feature vector of the target user u.
  • the recall module 208 is configured to input the target trajectory feature vector into a recall model, so as to recall multiple initial push data from multiple users through the recall model.
  • the classification module 210 is configured to input the multiple initial push data into the classification model.
  • the sorting module 212 is configured to sort the plurality of initial push data by association relationship through the classification model to obtain a plurality of target push data.
  • the push module 214 is configured to push the multiple target push data to the front end for display.
  • the behavior trajectory-based data pushing system 20 further includes a training module, the training module is used to: obtain a plurality of historical behavior data corresponding to a plurality of users; extract each user according to the plurality of historical behavior data Trajectory feature vector; and using multiple trajectory feature vectors as the input of the pre-acquired recall model to be trained, and a trajectory feature vector having a similar trajectory with each trajectory feature vector as the output of the recall model to be trained, to The recall model to be trained is trained until the loss function converges; each trajectory feature vector is used as the input of the trained recall model to obtain multiple acquaintance trajectory feature vectors corresponding to each trajectory feature vector; each trajectory is judged Whether the accuracy and recall rate of the acquaintance trajectory feature vector recalled by the feature vector reaches the detection accuracy threshold and the detection recall threshold, if they reach, the recall model to be trained after training is the initial recall model.
  • the training module is used to: obtain a plurality of historical behavior data corresponding to a plurality of users; extract each user according to the plurality of historical behavior data
  • the computer device 2 is a device that can automatically perform numerical calculation and/or information processing in accordance with pre-set or stored instructions.
  • the computer device 2 may be a rack server, a blade server, a tower server, or a cabinet server (including an independent server or a server cluster composed of multiple servers).
  • the computer device 2 at least includes, but is not limited to, a memory 21, a processor 22, a network interface 23, and a data push system 20 based on a behavior trajectory that can communicate with each other through a system bus.
  • the memory 21 includes at least one type of computer-readable storage medium.
  • the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory ( RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disks, optical disks, etc.
  • the memory 21 may be an internal storage unit of the computer device 2, for example, a hard disk or a memory of the computer device 2.
  • the memory 21 may also be an external storage device of the computer device 2, such as a plug-in hard disk, a smart media card (SMC), and a secure digital (Secure Digital, SD) card, flash card (Flash Card), etc.
  • the memory 21 may also include both the internal storage unit of the computer device 2 and its external storage device.
  • the memory 21 is generally used to store the operating system and various application software installed in the computer device 2, for example, the program code of the behavior trajectory-based data pushing system 20 in the second embodiment.
  • the memory 21 can also be used to temporarily store various types of data that have been output or will be output.
  • the processor 22 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments.
  • the processor 22 is generally used to control the overall operation of the computer device 2.
  • the processor 22 is used to run the program code or process data stored in the memory 21, for example, to run the data push system 20 based on the behavior trajectory, so as to implement the data push method based on the behavior trajectory of the first embodiment.
  • the network interface 23 may include a wireless network interface or a wired network interface, and the network interface 23 is generally used to establish a communication connection between the computer device 2 and other electronic devices.
  • the network interface 23 is used to connect the computer device 2 with an external terminal through a network, and establish a data transmission channel and a communication connection between the computer device 2 and the external terminal.
  • the network may be an intranet, the Internet, a global system of mobile communication (GSM), a wideband code division multiple access (WCDMA), a 4G network, and a 5G Network, Bluetooth (Bluetooth), Wi-Fi and other wireless or wired networks.
  • GSM global system of mobile communication
  • WCDMA wideband code division multiple access
  • 4G 4G network
  • 5G Network Bluetooth (Bluetooth), Wi-Fi and other wireless or wired networks.
  • FIG. 7 only shows the computer device 2 with components 20-23, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.
  • the behavior track-based data pushing system 20 stored in the memory 21 can also be divided into one or more program modules, and the one or more program modules are stored in the memory 21 and consist of one Or executed by multiple processors (in this embodiment, the processor 22) to complete the application.
  • FIG. 6 shows a schematic diagram of program modules for implementing the data pushing system 20 based on behavior trajectory according to the second embodiment of the present application.
  • the data pushing system 20 based on behavior trajectory can be divided into receiving modules. 200.
  • the program module referred to in the present application refers to a series of computer program instruction segments capable of completing specific functions, and is more suitable than a program to describe the execution process of the behavior track-based data pushing system 20 in the computer device 2.
  • the specific functions of the program modules 200-214 have been described in detail in the second embodiment, and will not be repeated here.
  • the computer-readable storage medium may be non-volatile or volatile, such as flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX). Memory, etc.), random access memory (RAM), static random access memory (SRAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), magnetic memory , Magnetic disks, optical disks, servers, App application malls, etc., on which computer-readable instructions are stored, and the computer-readable instructions realize corresponding functions when executed.
  • the computer-readable storage medium of this embodiment is used in the data pushing system 20 based on the behavior trajectory, and the processor executes the following steps:

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Mathematical Physics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Evolutionary Computation (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种基于行为轨迹的数据推送方法,所述方法包括:接收目标用户通过目标用户终端触发的数据推送指令(S100);根据数据推送指令从行为日志系统中获取目标用户的目标历史行为数据(S102);根据目标历史行为数据的构建目标用户的目标轨迹图(S104);根据目标轨迹图提取目标用户的目标轨迹特征向量(S106);将目标轨迹特征向量输入到召回模型中以得到多个初始推送数据(S108);将多个初始推送数据输入到分类模型中以得到多个目标推送数据(S110,S112);及将多个目标推送数据推送到前端进行展示(S114)。所述方法根据用户的行为轨迹为用户构建轨迹图,并通过对分类模型和回归模型对用户的轨迹进行相识度匹配,提高了数据推荐的准确率和效率。

Description

基于行为轨迹的数据推送方法、系统和计算机设备
本申请申明2020年06月24日递交的申请号为202010588495.X、名称为“基于行为轨迹的数据推送方法、系统和计算机设备”的中国专利申请的优先权,该中国专利申请的整体内容以参考的方式结合在本申请中。
技术领域
本申请实施例涉及数据分析领域,尤其涉及一种基于行为轨迹的数据推送方法、系统、计算机设备及计算机可读存储介质。
背景技术
目前,社会网络服务已经成为重要的互联网服务之一。社会性网络服务使得人们足不出户也能和朋友交流、互动,同时也能结交更多的好友来扩大朋友圈。像微博、Twitter、facebook这样的网络服务提供商都在致力于提供更好的交友服务。
但是,发明人意识到,现有的好友推荐几乎均只关注两个用户的年龄、性别、兴趣爱好等人物基本属性,没有挖掘更多的人物信息。因此,如何挖掘更多的人物信息,从而进一步提升好友推荐的准确性,成为目前亟需解决的技术问题之一。
发明内容
有鉴于此,有必要提供一种基于行为轨迹的数据推送方法、系统、计算机设备及计算机可读存储介质,以解决当前好友推荐条件单一、没有挖掘更多的人物信息以及用户数据推荐的准确性低等技术问题。
为实现上述目的,本申请实施例提供了一种基于行为轨迹的数据推送方法,所述方法步骤包括:
接收目标用户通过目标用户终端触发的数据推送指令;
根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据;
将所述多个初始推送数据输入到所述分类模型中;
通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
将所述多个目标推送数据推送到前端进行展示。
为实现上述目的,本申请实施例还提供了一种基于行为轨迹的数据推送系统,包括:
接收模块,用于接收目标用户通过目标用户终端触发的数据推送指令;
获取模块,用于根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
构建模块,用于根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
提取模块,用于根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
召回模块,用于将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据;
分类模块,用于将所述多个初始推送数据输入到所述分类模型中;
排序模块,用于通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
推送模块,用于将所述多个目标推送数据推送到前端进行展示。
为实现上述目的,本申请实施例还提供了一种计算机设备,,所述计算机设备包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令,所述计算机可读指令被处理器执行时实现以下步骤:
接收目标用户通过目标用户终端触发的数据推送指令;
根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据;
将所述多个初始推送数据输入到所述分类模型中;
通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
将所述多个目标推送数据推送到前端进行展示。
为实现上述目的,本申请实施例还提供了一种计算机可读存储介质,所述计算机可读存储介质内存储有计算机可读指令,所述计算机可读指令可被至少一个处理器所执行,以使所述至少一个处理器执行如下步骤:
接收目标用户通过目标用户终端触发的数据推送指令;
根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回 多个初始推送数据;
将所述多个初始推送数据输入到所述分类模型中;
通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
将所述多个目标推送数据推送到前端进行展示。
本申请实施例提供的基于行为轨迹的数据推送方法、系统、计算机设备及计算机可读存储介质,为用户提供了更为精准更为个性化的数据推送方法;本申请根据用户的行为轨迹为用户构建轨迹图,并通过对分类模型和回归模型对用户的轨迹进行相识度匹配,丰富了用户的数据推荐条件,提高了数据推荐的准确率和效率。
附图说明
图1为本申请实施例基于行为轨迹的数据推送方法的流程示意图。
图2为本申请基于行为轨迹的数据推送方法的目标轨迹图。
图3为本申请基于行为轨迹的数据推送方法的轨迹全图。
图4为本申请基于行为轨迹的数据推送方法的Skip-gram模型的结构图。
图5为本申请基于行为轨迹的数据推送方法的召回阶段的流程图。
图6为本申请基于行为轨迹的数据推送系统实施例二的程序模块示意图。
图7为本申请计算机设备实施例三的硬件结构示意图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描述的具体实施例仅用以解释本申请,并不用于限定本申请。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
需要说明的是,在本申请中涉及“第一”、“第二”等的描述仅用于描述目的,而不能理解为指示或暗示其相对重要性或者隐含指明所指示的技术特征的数量。由此,限定有“第一”、“第二”的特征可以明示或者隐含地包括至少一个该特征。另外,各个实施例之间的技术方案可以相互结合,但是必须是以本领域普通技术人员能够实现为基础,当技术方案的结合出现相互矛盾或无法实现时应当认为这种技术方案的结合不存在,也不在本申请要求的保护范围之内。
以下实施例中,将以计算机设备2为执行主体进行示例性描述。
实施例一
参阅图1,示出了本申请实施例之基于行为轨迹的数据推送方法的步骤流程图。可以 理解,本方法实施例中的流程图不用于对执行步骤的顺序进行限定。下面以计算机设备2为执行主体进行示例性描述。具体如下。
步骤S100,接收目标用户通过目标用户终端触发的数据推送指令。
示例性的,所述数据推送指令可以通过数据推送系统中的数据推送组件触发。例如,所述目标用户可以通过目标用户终端访问数据推送系统,并点击所述数据推送系统中的数据推送按钮就可以触发所述数据推送组件;其中,在所述数据推送组件被触发时,所述数据推送系统可以接收到一个数据推送指令。
步骤S102,根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据。
所述目标用户在访问数据推送系统前需要先登录所述数据推送系统,所述数据推送系统可以根据所述数据推送指令获取当前触发数据推送组件的目标用户身份信息,在所述目标用户身份信息得到确认后,所述数据推送系统可以根据所述目标用户身份信息从所述行为日志系统中获取目标用户的目标历史行为数据,所述目标历史行为数据包括目标用户历史访问停留地点(经纬度)及历史访问时间(工作日/非工作日)。
在一些实施例中,所述数据推送系统还可以根据所述目标用户身份信息获取目标用户的自定义行为数据,所述自定义行为数据包括目标用户自定义的访问停留地点(经纬度)及访问时间(工作日/非工作日)。
步骤S104,根据所述目标历史行为数据的构建所述目标用户的目标轨迹图。
示例性的,所述步骤S104还可以进一步的包括:
步骤S104a,从所述目标历史行为数据中获取多个目标地点数据:l 1、l 2、l 3...l n和所述多个目标地点数据对应的多个目标时间数据:t 1、t 2、t 3...t n。步骤S104b,根据所述多个目标时间数据的时间顺序为所述目标用户u构建目标轨迹图:C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)},其中,l 1、l 2和l n分别为第一个目标地点数据、第二个目标地点数据和第n个目标地点数据,t 1、t 2和t n分别为第一个目标地点数据对应的目标时间数据、第二个目标地点数据对应的目标时间数据和第n个目标地点数据对应的目标时间数据。
示例性的,所述目标历史行为数据包括所述目标用户所访问过的多个访问地点对应的多个目标地点数据:l 1、l 2、l 3...l n和所述多个目标地点数据对应的多个目标时间数据:t 1、t 2、t 3...t n。所述目标用户的可以根据对所述多个访问地点的访问时间顺序构成一个目标轨迹图C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)}。
如图2所示,所述图2中的节点表示访问的地点,箭头表示访问的先后顺序,因此该用户的访问顺序为l 1、l 2、l 1、l 2、l 3。在一些实施例中,将所有人的轨迹全部放入图中,可构成轨迹全图,如图3所示。
步骤S106,根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量。
示例性的,所述步骤S106还可以进一步的包括:
步骤S106a,根据每个目标地点数据和每个目标地点数据所对应的目标时间数据提取每个目标地点的目标地点特征向量,以得到多个目标地点特征向量E 1、E 2、E 3...E n;步骤S106b,根据所述多个目标地点特征向量,确定所述目标用户u的目标轨迹特征向量:
Figure PCTCN2021097278-appb-000001
并将目标轨迹特征向量上传至区块链中,其中,E i为所述目标用户u的第i个目标地点特征向量,E u为所述目标用户u的目标轨迹特征向量。
示例性的,所述目标用户u的轨迹为C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)},表示所述目标用户u共访问过n个地点,通过Skip-gram模型训练得到为E 1、E 2、E 3...E n。则目标用户u代表的目标轨迹特征向量为:
Figure PCTCN2021097278-appb-000002
其中,所述Skip-gram是word2vec模型的一种,其结构如图4所示。
示例性的,将目标轨迹特征向量上传至区块链可保证其安全性和对用户的公正透明性。本示例所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
步骤S108,将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据。
示例性的,根据所述目标轨迹向量通过所述召回模型从多个用户中召回多个初始推送数据。所述召回是为了从千万级数据中锁定小部分的候选集,已将被锁定的候选集作为对推荐列表的初步筛选。常见的召回方法是协同过滤算法,而该方法只适合用于用户较少的场景,不适合用于千万级数据量的场景下。在本实施例中可以先采Graph embedding(网络表示学习,又称图嵌入)的方法,将每个人转化为一个向量,通过召回模型计算两个向量的相似度,相似度高则表示两人的轨迹相似度高,越有可能是好友关系。其中,召回阶段的流程如图5所示。
通过计算目标轨迹特征向量E u和其他轨迹特征向量E u′的余弦相似性,判断轨迹相似性,相似性越高表示两人越可能有关系,将相似性靠前的多个用户作为初始推送数据,所述相似性靠前的多个用户的数量为预设的数量。在一些实施例中,所述初始推送数据可以是与所述目标用户行为轨迹相识的用户。
示例性的,所述步骤S108还包括进一步的包括:步骤S108a,判断所述目标用户是否为特殊用户;及步骤S108b,如果所述目标用户为特殊用户时,则通过预定的规则计算未出现的地点的向量。
示例性的,所述特殊用户为轨迹图中有一部分或全部未被Skip-gram(神经网络结构)模型训练过的用户。例如,POI(Point of Information信息点)是地理系统中常见的地点信 息,表示某个地点的类别,可能是美食店、服装店、健身店等等。爬取大众点评的数据,POI一共分为17种,包括“美食”、“教育学校”、“机构团体”、“汽车”、“娱乐休闲”、“生活服务”、“运动健身”、“房产小区”、“基础设施”、“酒店宾馆”、“购物”、“医疗保健”、“旅游景点”、“文化场馆”、“公司企业”、“银行金融”和“地名地址”。通过Skip-gram模型训练得到的每个地点均表示某一种POI,因此17种POI的embedding(连续向量)向量分为别:
Figure PCTCN2021097278-appb-000003
其中,在计算未出现的地点的向量时,只需要判断新地点的POI类型,再用该类POI的embedding向量近似代替。
示例性的,所述方法还可以进一步的包括:
步骤S200,获取多个用户对应的多个历史行为数据;步骤S202,根据所述多个历史行为数据提取每个用户的轨迹特征向量;步骤S204,将多个轨迹特征向量作为预先获取的待训练召回模型的输入,与每个轨迹特征向量具有相似轨迹的轨迹特征向量作为所述待训练召回模型的输出,来对所述待训练召回模型进行训练,直至损失函数收敛;步骤S206,将每个轨迹特征向量作为训练后的待训练召回模型的输入,以获得与每个轨迹特征向量对应的多个相识轨迹特征向量;及步骤S208,判断每个轨迹特征向量召回的相识轨迹特征向量的准确率及召回率是否达到检测准确阈值及检测召回阈值,若达到,则训练后的待训练召回模型则为初始召回模型。
示例性的,所述召回模型可以根据所述多个历史行为数据确定所述多个用户的轨迹图:
Figure PCTCN2021097278-appb-000004
并提取多个用户中的每个用户的轨迹特征向量,以得到所述多个用户对应的多个轨迹特征向量。
步骤S110,将所述多个初始推送数据输入到所述分类模型中。
示例性的,在所述召回阶段结束后,所述多个初始推送数据仅仅与所述目标用户轨迹相似,为了进一步筛选出与所述目标用户具有较强的关联关系的数据,本方案通过将所述多个初始推送数据输入到所述分类模型中,以得到与所述目标用户具有较强的关联关系的多个目标推送数据。在一些实施例中,所述目标推送数据可以是与所述目标用户具有较强的关联关系的用户,例如,两个用户在某个地铁站或者火车站共同出现,则相似度较高。但不以证明他们有朋友关系,有可能他们擦肩而过。因此还需要进一步分类。
步骤S112,通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据。
示例性的,所述分类模型可以基于预先设定的因素对所述多个初始推送数据进行关联关系排序,并将排序靠前的多个初始推送数据作为所述目标推送数据。其中,所述预先设定的因素包括位置熵、时间间隔序列和时间维度。
示例性的,所述目标用户与所述多个初始推送数据的关联关系的大小,可以根据所述位置熵来判断,而所述位置熵的大小可以根据区域热门程度来判断,例如,火车站这样的热门地点,很多人都共同出现过;而如果两个人经常出现在某个非热门区域,比如某小区, 则表示两人有较强的关系。因此在分类过程中需要判断两人共同访问过的地点热门程度,来确定所述位置熵。
Figure PCTCN2021097278-appb-000005
Figure PCTCN2021097278-appb-000006
位置熵H l越大表示该位置越热门,其中p l(u)是用户u去过地点l的概率。p l(u)的分子为用户u去过地点l的次数,分母为所有用户去过地点l的次数。以上海为例,位置熵较高的地点如下表。可以发现位置熵较高的都是人流量较大的火车站或者地铁站。如表1所示:
地点 H l
31.2335,121.4745(人民广场) 8.19999
31.2385,121.4155(曹杨路地铁站) 7.59356
31.2375,121.4845(南京东路地铁站) 7.59340
31.2295,121.5265(世纪大道) 7.54229
31.1945,121.3205(上海虹桥站) 7.3237
表1
考虑到实际情况:没有关系的用户倾向于共同访问热门的地点,有关系的用户也可能共同访问不热门的地点,比如住宅小区等。因此在位置熵的基础上引伸出“共现地点位置熵”和“共现频次位置熵”两个特征。
Figure PCTCN2021097278-appb-000007
Figure PCTCN2021097278-appb-000008
如表2所示:
  H(l 1) H(l 2) H(l 3) H(l 4) WL(u,u′) WO(u,u′)
Pair1 0.02 0.18 0.23 0.09 3.2481 8.7032
Pair2 3.24 0.54 0.38 0.66 1.8226 4.5353
Pair3 0.65 0.16 2.67 3.14 1.4867 3.5907
Pair4 1.25 2.18 3.34 2.29 0.5362 1.3286
表2
表2列出了4对用户,分别为pair1、pair2、pair3、pair4,每对用户均共同访问过4个不同的地址:l 1、l 2、l 3、l 4,可以发现pair1的4个地址位置熵较小,因此“共现地点位置熵”和“共现频次位置熵”较大,两人可能有关系。而pair4的两个用户访问过的地址位置上均较大,可能为某个地铁站或者商圈,因此两人的“共现地点位置熵”和“共现频次位置熵”较小,两人可能没有关系。
示例性的,所述目标用户与所述多个初始推送数据的关联关系的大小,还可以根据所述时间间隔序列来判断,所述时间间隔序列是用于判断所述两个用户访问同一地点的时间间隔,其中,时间间隔为0则表示两人同时访问某个地点,例如,用户u和u′均访问过地点l,u的访问行为为<u,t u,1,l>、<u,t u,2,l>、…<u,t u,m,l>,u′的访问行为为<u,t u′,1,l>、<u,t u′,2,l>、…<u,t u′,n,l>,表示u和u′都在多个时间下访问地点l。则可以构建两人的最大时间间序列和最小时间间隔序列,分别为
Figure PCTCN2021097278-appb-000009
Figure PCTCN2021097278-appb-000010
从上面两个公式可以看出,
Figure PCTCN2021097278-appb-000011
Figure PCTCN2021097278-appb-000012
均是长度为m+n的序列。
Figure PCTCN2021097278-appb-000013
为两人访问l的所有时间差的最大值的集合,
Figure PCTCN2021097278-appb-000014
为两人访问l的所有时间差的最小值的集合。
实际两人的共现地址会有多个,而且两人访问同一地点的次数也不尽相同,因此需要先对地点l的时间序列平均化得到u和u′关于l的平均时间间隔,再对u和u′的所有共现地址的时间间隔平均化。即
Figure PCTCN2021097278-appb-000015
Figure PCTCN2021097278-appb-000016
其中L u,u′,为用户u和u′都访问过的地点。
示例性的,所述目标用户与所述多个初始推送数据的关联关系的大小,还可以根据所述时间维度来判断,为了拓展维度,可以在不同时间纬度下建立特征,时间分为工作日、非工作日以及所有时间。因此全部特征共15个,如表3所示:
Figure PCTCN2021097278-appb-000017
表3
全部特征构建好后,需要放入二分类Lightgbm模型中,通过贝叶斯参数寻优找到合适的参数进行预测。
步骤S114,将所述多个目标推送数据推送到前端进行展示。
在本实施例中,通过前后端交互的系统将经过召回、分类后处理后得到的多个目标推送数据再传给前端进行展示。在一些实施例中,所述前端可以采用HTML和Javascript编写,所述HTML是一种用于创建网页的标准标记语言,所述JavaScript是一种属于网络的脚本语言。前后端交互采用Flask和Ajax编写,Flask是一个轻量级Web应用框架,Ajax是一种创建交互式网页应用的网页开发技术,可以实现在不重新加载整个网页的情况下,对网页的某部分进行更新。
在本实施例中,基于目标轨迹特征向量可以得到对应的多个目标推送数据,具体来说,通过将所述目标轨迹特征向量输入到训练好的召回模型中,然后将召回模型输出的结果输入到分类模型中可以得到所述多个目标推送数据。将目标轨迹特征向量上传至区块链可保证其安全性和对用户的公正透明性。用户设备可以从区块链中下载得该目标轨迹特征向量,以便查证多个目标推送数据是否被篡改。本示例所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
实施例二
图6为本申请基于行为轨迹的数据推送系统实施例二的程序模块示意图。基于行为轨迹的数据推送系统20可以包括或被分割成一个或多个程序模块,一个或者多个程序模块被 存储于存储介质中,并由一个或多个处理器所执行,以完成本申请,并可实现上述基于行为轨迹的数据推送方法。本申请实施例所称的程序模块是指能够完成特定功能的一系列计算机程序指令段,比程序本身更适合于描述基于行为轨迹的数据推送系统20在存储介质中的执行过程。以下描述将具体介绍本实施例各程序模块的功能:
接收模块200,用于接收模块,用于接收目标用户通过目标用户终端触发的数据推送指令。
获取模块202,用于根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据。
构建模块204,用于根据所述目标历史行为数据的构建所述目标用户的目标轨迹图。
示例性的,所述构建模块204还用于:从所述目标历史行为数据中获取多个目标地点数据:l 1、l 2、l 3...l n和所述多个目标地点数据对应的多个目标时间数据:t 1、t 2、t 3...t n;及根据所述多个目标时间数据的时间顺序为所述目标用户u构建目标轨迹图:C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)}。
提取模块206,用于根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量。
示例性的,所述提取模块206还用于:根据每个目标地点数据和每个目标地点数据所对应的目标时间数据提取每个目标地点的目标地点特征向量,以得到多个目标地点特征向量E 1、E 2、E 3...E n;及
根据所述多个目标地点特征向量,确定所述目标用户u的目标轨迹特征向量:
Figure PCTCN2021097278-appb-000018
并将目标轨迹特征向量上传至区块链中,其中,E i为所述目标用户u的第i个目标地点特征向量,E u为所述目标用户u的目标轨迹特征向量。
召回模块208,用于将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据。
分类模块210,用于将所述多个初始推送数据输入到所述分类模型中。
排序模块212,用于通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据。
推送模块214,用于将所述多个目标推送数据推送到前端进行展示。
示例性的,所述基于行为轨迹的数据推送系统20还包括训练模块,所述训练模块用于:获取多个用户对应的多个历史行为数据;根据所述多个历史行为数据提取每个用户的轨迹特征向量;及将多个轨迹特征向量作为预先获取的待训练召回模型的输入,与每个轨迹特征向量具有相似轨迹的轨迹特征向量作为所述待训练召回模型的输出,来对所述待训练召回模型进行训练,直至损失函数收敛;将每个轨迹特征向量作为训练后的待训练召回模型的输入,以获得与每个轨迹特征向量对应的多个相识轨迹特征向量;判断每个轨迹特征向量召回的相识轨迹特征向量的准确率及召回率是否达到检测准确阈值及检测召回阈值,若达到,则训练后的待训练召回模型则为初始召回模型。
实施例三
参阅图7,是本申请实施例三之计算机设备的硬件架构示意图。本实施例中,所述计算机设备2是一种能够按照事先设定或者存储的指令,自动进行数值计算和/或信息处理的设备。该计算机设备2可以是机架式服务器、刀片式服务器、塔式服务器或机柜式服务器(包括独立的服务器,或者多个服务器所组成的服务器集群)等。如图所示,所述计算机设备2至少包括,但不限于,可通过系统总线相互通信连接存储器21、处理器22、网络接口23、以及基于行为轨迹的数据推送系统20。
本实施例中,存储器21至少包括一种类型的计算机可读存储介质,所述可读存储介质包括闪存、硬盘、多媒体卡、卡型存储器(例如,SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘等。在一些实施例中,存储器21可以是计算机设备2的内部存储单元,例如该计算机设备2的硬盘或内存。在另一些实施例中,存储器21也可以是计算机设备2的外部存储设备,例如该计算机设备2上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。当然,存储器21还可以既包括计算机设备2的内部存储单元也包括其外部存储设备。本实施例中,存储器21通常用于存储安装于计算机设备2的操作系统和各类应用软件,例如实施例二的基于行为轨迹的数据推送系统20的程序代码等。此外,存储器21还可以用于暂时地存储已经输出或者将要输出的各类数据。
处理器22在一些实施例中可以是中央处理器(Central Processing Unit,CPU)、控制器、微控制器、微处理器、或其他数据处理芯片。该处理器22通常用于控制计算机设备2的总体操作。本实施例中,处理器22用于运行存储器21中存储的程序代码或者处理数据,例如运行基于行为轨迹的数据推送系统20,以实现实施例一的基于行为轨迹的数据推送方法。
所述网络接口23可包括无线网络接口或有线网络接口,该网络接口23通常用于在所述计算机设备2与其他电子装置之间建立通信连接。例如,所述网络接口23用于通过网络将所述计算机设备2与外部终端相连,在所述计算机设备2与外部终端之间的建立数据传输通道和通信连接等。所述网络可以是企业内部网(Intranet)、互联网(Internet)、全球移动通讯系统(Global System of Mobile communication,GSM)、宽带码分多址(Wideband Code Division Multiple Access,WCDMA)、4G网络、5G网络、蓝牙(Bluetooth)、Wi-Fi等无线或有线网络。需要指出的是,图7仅示出了具有部件20-23的计算机设备2,但是应理解的是,并不要求实施所有示出的部件,可以替代的实施更多或者更少的部件。
在本实施例中,存储于存储器21中的基于行为轨迹的数据推送系统20还可以被分割为一个或者多个程序模块,所述一个或者多个程序模块被存储于存储器21中,并由一个或多个处理器(本实施例为处理器22)所执行,以完成本申请。
例如,图6示出了本申请实施例二之所述实现基于行为轨迹的数据推送系统20的程序模块示意图,该实施例中,所述基于行为轨迹的数据推送系统20可以被划分为接收模块200、获取模块202、构建模块204、提取模块206、召回模块208、分类模块210、排序模块212和推送模块214。其中,本申请所称的程序模块是指能够完成特定功能的一系列计算机程序指令段,比程序更适合于描述所述基于行为轨迹的数据推送系统20在所述计算机设备2中的执行过程。所述程序模块200-214的具体功能在实施例二中已有详细描述,在此不再赘述。
实施例四
本实施例还提供一种计算机可读存储介质,所述计算机可读存储介质可以是非易失性,也可以是易失性,如闪存、硬盘、多媒体卡、卡型存储器(例如,SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘、服务器、App应用商城等等,其上存储有计算机可读指令,计算机可读指令执行时实现相应功能。本实施例的计算机可读存储介质用于基于行为轨迹的数据推送系统20,被处理器执行如下步骤:
接收目标用户通过目标用户终端触发的数据推送指令;
根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据;
将所述多个初始推送数据输入到所述分类模型中;
通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
将所述多个目标推送数据推送到前端进行展示。
上述本申请实施例序号仅仅为了描述,不代表实施例的优劣。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。
以上仅为本申请的优选实施例,并非因此限制本申请的专利范围,凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种基于行为轨迹的数据推送方法,其中,所述方法包括:
    接收目标用户通过目标用户终端触发的数据推送指令;
    根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
    根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
    根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
    将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据;
    将所述多个初始推送数据输入到所述分类模型中;
    通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
    将所述多个目标推送数据推送到前端进行展示。
  2. 如权利要求1所述的基于行为轨迹的数据推送方法,其中,所述根据所述目标历史行为数据的构建所述目标用户的目标轨迹图,包括:
    从所述目标历史行为数据中获取多个目标地点数据:l 1、l 2、l 3...l n和所述多个目标地点数据对应的多个目标时间数据:t 1、t 2、t 3...t n;及
    根据所述多个目标时间数据的时间顺序为所述目标用户u构建目标轨迹图:C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)},其中,l 1、l 2和l n分别为第一个目标地点数据、第二个目标地点数据和第n个目标地点数据,t 1、t 2和t n分别为第一个目标地点数据对应的目标时间数据、第二个目标地点数据对应的目标时间数据和第n个目标地点数据对应的目标时间数据。
  3. 如权利要求1所述的基于行为轨迹的数据推送方法,其中,所述根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量,包括:
    根据每个目标地点数据和每个目标地点数据所对应的目标时间数据提取每个目标地点的目标地点特征向量,以得到多个目标地点特征向量E 1、E 2、E 3...E n;及
    根据所述多个目标地点特征向量,确定所述目标用户u的目标轨迹特征向量:
    Figure PCTCN2021097278-appb-100001
    并将目标轨迹特征向量上传至区块链中,其中,E i为所述目标用户u的第i个目标地点特征向量,E u为所述目标用户u的目标轨迹特征向量。
  4. 如权利要求3所述的基于行为轨迹的数据推送方法,其中,还包括:
    判断所述目标用户是否为特殊用户;及
    如果所述目标用户为特殊用户时,则通过预定的规则计算未出现的地点的向量。
  5. 如权利要求1所述的基于行为轨迹的数据推送方法,其中,还包括:训练召回模型的步骤:
    获取多个用户对应的多个历史行为数据;
    根据所述多个历史行为数据提取每个用户的轨迹特征向量;
    将多个轨迹特征向量作为预先获取的待训练召回模型的输入,与每个轨迹特征向量具有相似轨迹的轨迹特征向量作为所述待训练召回模型的输出,来对所述待训练召回模型进行训练,直至损失函数收敛;
    将每个轨迹特征向量作为训练后的待训练召回模型的输入,以获得与每个轨迹特征向量对应的多个相识轨迹特征向量;及
    判断每个轨迹特征向量召回的相识轨迹特征向量的准确率及召回率是否达到检测准确阈值及检测召回阈值,若达到,则训练后的待训练召回模型则为初始召回模型。
  6. 一种基于行为轨迹的数据推送系统,其中,包括:
    接收模块,用于接收目标用户通过目标用户终端触发的数据推送指令;
    获取模块,用于根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
    构建模块,用于根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
    提取模块,用于根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
    召回模块,用于将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据;
    分类模块,用于将所述多个初始推送数据输入到所述分类模型中;
    排序模块,用于通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
    推送模块,用于将所述多个目标推送数据推送到前端进行展示。
  7. 如权利要求6所述的基于行为轨迹的数据推送系统,其中,所述构建模块,还用于:
    从所述目标历史行为数据中获取多个目标地点数据:l 1、l 2、l 3...l n和所述多个目标地点数据对应的多个目标时间数据:t 1、t 2、t 3...t n;及
    根据所述多个目标时间数据的时间顺序为所述目标用户u构建目标轨迹图:C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)},其中,l 1、l 2和l n分别为第一个目标地点数据、第二个目标地点数据和第n个目标地点数据,t 1、t 2和t n分别为第一个目标地点数据对应的目标时间数据、第二个目标地点数据对应的目标时间数据和第n个目标地点数据对应的目标时间数据。
  8. 如权利要求7所述的基于行为轨迹的数据推送系统,其中,所述提取模块,还用于:
    根据每个目标地点数据和每个目标地点数据所对应的目标时间数据提取每个目标地点的目标地点特征向量,以得到多个目标地点特征向量E 1、E 2、E 3...E n;及
    根据所述多个目标地点特征向量,确定所述目标用户u的目标轨迹特征向量:
    Figure PCTCN2021097278-appb-100002
    并将目标轨迹特征向量上传至区块链中,其中,E i为所述目标用户u的第 i个目标地点特征向量,E u为所述目标用户u的目标轨迹特征向量。
  9. 如权利要求8所述的基于行为轨迹的数据推送系统,其中,还包括判断模块,所述判断模块,用于:
    判断所述目标用户是否为特殊用户;及
    如果所述目标用户为特殊用户时,则通过预定的规则计算未出现的地点的向量。
  10. 如权利要求6所述的基于行为轨迹的数据推送系统,其中,还包括训练模块,所述训练模块,用于:
    获取多个用户对应的多个历史行为数据;
    根据所述多个历史行为数据提取每个用户的轨迹特征向量;
    将多个轨迹特征向量作为预先获取的待训练召回模型的输入,与每个轨迹特征向量具有相似轨迹的轨迹特征向量作为所述待训练召回模型的输出,来对所述待训练召回模型进行训练,直至损失函数收敛;
    将每个轨迹特征向量作为训练后的待训练召回模型的输入,以获得与每个轨迹特征向量对应的多个相识轨迹特征向量;及
    判断每个轨迹特征向量召回的相识轨迹特征向量的准确率及召回率是否达到检测准确阈值及检测召回阈值,若达到,则训练后的待训练召回模型则为初始召回模型。
  11. 一种计算机设备,所述计算机设备包括存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机可读指令,其中,所述计算机可读指令被处理器执行时实现以下步骤:
    接收目标用户通过目标用户终端触发的数据推送指令;
    根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
    根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
    根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
    将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回多个初始推送数据;
    将所述多个初始推送数据输入到所述分类模型中;
    通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
    将所述多个目标推送数据推送到前端进行展示。
  12. 如权利要求11所述的计算机设备,其中,所述计算机可读指令被处理器执行时还实现以下步骤:
    从所述目标历史行为数据中获取多个目标地点数据:l 1、l 2、l 3...l n和所述多个目标地点数据对应的多个目标时间数据:t 1、t 2、t 3...t n;及
    根据所述多个目标时间数据的时间顺序为所述目标用户u构建目标轨迹图: C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)},其中,l 1、l 2和l n分别为第一个目标地点数据、第二个目标地点数据和第n个目标地点数据,t 1、t 2和t n分别为第一个目标地点数据对应的目标时间数据、第二个目标地点数据对应的目标时间数据和第n个目标地点数据对应的目标时间数据。
  13. 如权利要求11所述的计算机设备,其中,所述计算机可读指令被处理器执行时还实现以下步骤:
    根据每个目标地点数据和每个目标地点数据所对应的目标时间数据提取每个目标地点的目标地点特征向量,以得到多个目标地点特征向量E 1、E 2、E 3...E n;及
    根据所述多个目标地点特征向量,确定所述目标用户u的目标轨迹特征向量:
    Figure PCTCN2021097278-appb-100003
    并将目标轨迹特征向量上传至区块链中,其中,E i为所述目标用户u的第i个目标地点特征向量,E u为所述目标用户u的目标轨迹特征向量。
  14. 如权利要求13所述的计算机设备,其中,所述计算机可读指令被处理器执行时还实现以下步骤:
    判断所述目标用户是否为特殊用户;及
    如果所述目标用户为特殊用户时,则通过预定的规则计算未出现的地点的向量。
  15. 如权利要求11所述的计算机设备,其中,所述计算机可读指令被处理器执行时还实现以下步骤:
    获取多个用户对应的多个历史行为数据;
    根据所述多个历史行为数据提取每个用户的轨迹特征向量;
    将多个轨迹特征向量作为预先获取的待训练召回模型的输入,与每个轨迹特征向量具有相似轨迹的轨迹特征向量作为所述待训练召回模型的输出,来对所述待训练召回模型进行训练,直至损失函数收敛;
    将每个轨迹特征向量作为训练后的待训练召回模型的输入,以获得与每个轨迹特征向量对应的多个相识轨迹特征向量;及
    判断每个轨迹特征向量召回的相识轨迹特征向量的准确率及召回率是否达到检测准确阈值及检测召回阈值,若达到,则训练后的待训练召回模型则为初始召回模型。
  16. 一种计算机可读存储介质,其中,所述计算机可读存储介质内存储有计算机可读指令,所述计算机可读指令可被至少一个处理器所执行,以使所述至少一个处理器执行如下步骤:
    接收目标用户通过目标用户终端触发的数据推送指令;
    根据所述数据推送指令从行为日志系统中获取目标用户的目标历史行为数据;
    根据所述目标历史行为数据的构建所述目标用户的目标轨迹图;
    根据所述目标轨迹图提取所述目标用户的目标轨迹特征向量;
    将所述目标轨迹特征向量输入到召回模型中,以通过所述召回模型从多个用户中召回 多个初始推送数据;
    将所述多个初始推送数据输入到所述分类模型中;
    通过所述分类模型对所述多个初始推送数据进行关联关系排序,以得到多个目标推送数据;及
    将所述多个目标推送数据推送到前端进行展示。
  17. 如权利要求16所述的计算机可读存储介质,其中,所述计算机可读指令还可被至少一个处理器所执行,以使所述至少一个处理器执行如下步骤:
    从所述目标历史行为数据中获取多个目标地点数据:l 1、l 2、l 3...l n和所述多个目标地点数据对应的多个目标时间数据:t 1、t 2、t 3...t n;及
    根据所述多个目标时间数据的时间顺序为所述目标用户u构建目标轨迹图:C u={(u,t 1,l 1),(u,t 2,l 2)...(u,t n,l n)},其中,l 1、l 2和l n分别为第一个目标地点数据、第二个目标地点数据和第n个目标地点数据,t 1、t 2和t n分别为第一个目标地点数据对应的目标时间数据、第二个目标地点数据对应的目标时间数据和第n个目标地点数据对应的目标时间数据。
  18. 如权利要求16所述的计算机可读存储介质,其中,所述计算机可读指令还可被至少一个处理器所执行,以使所述至少一个处理器执行如下步骤:
    根据每个目标地点数据和每个目标地点数据所对应的目标时间数据提取每个目标地点的目标地点特征向量,以得到多个目标地点特征向量E 1、E 2、E 3...E n;及
    根据所述多个目标地点特征向量,确定所述目标用户u的目标轨迹特征向量:
    Figure PCTCN2021097278-appb-100004
    并将目标轨迹特征向量上传至区块链中,其中,E i为所述目标用户u的第i个目标地点特征向量,E u为所述目标用户u的目标轨迹特征向量。
  19. 如权利要求18所述的计算机可读存储介质,其中,所述计算机可读指令还可被至少一个处理器所执行,以使所述至少一个处理器执行如下步骤:
    判断所述目标用户是否为特殊用户;及
    如果所述目标用户为特殊用户时,则通过预定的规则计算未出现的地点的向量。
  20. 如权利要求16所述的计算机可读存储介质,其中,所述计算机可读指令还可被至少一个处理器所执行,以使所述至少一个处理器执行如下步骤:
    获取多个用户对应的多个历史行为数据;
    根据所述多个历史行为数据提取每个用户的轨迹特征向量;
    将多个轨迹特征向量作为预先获取的待训练召回模型的输入,与每个轨迹特征向量具有相似轨迹的轨迹特征向量作为所述待训练召回模型的输出,来对所述待训练召回模型进行训练,直至损失函数收敛;
    将每个轨迹特征向量作为训练后的待训练召回模型的输入,以获得与每个轨迹特征向量对应的多个相识轨迹特征向量;及
    判断每个轨迹特征向量召回的相识轨迹特征向量的准确率及召回率是否达到检测准确阈值及检测召回阈值,若达到,则训练后的待训练召回模型则为初始召回模型。
PCT/CN2021/097278 2020-06-24 2021-05-31 基于行为轨迹的数据推送方法、系统和计算机设备 WO2021259007A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010588495.XA CN111753214A (zh) 2020-06-24 2020-06-24 基于行为轨迹的数据推送方法、系统和计算机设备
CN202010588495.X 2020-06-24

Publications (1)

Publication Number Publication Date
WO2021259007A1 true WO2021259007A1 (zh) 2021-12-30

Family

ID=72677088

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/097278 WO2021259007A1 (zh) 2020-06-24 2021-05-31 基于行为轨迹的数据推送方法、系统和计算机设备

Country Status (2)

Country Link
CN (1) CN111753214A (zh)
WO (1) WO2021259007A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114418093A (zh) * 2022-01-19 2022-04-29 北京百度网讯科技有限公司 训练路径表征模型、输出信息的方法和装置
CN114785852A (zh) * 2022-04-07 2022-07-22 北京达佳互联信息技术有限公司 一种推送内容确定方法、装置、电子设备及存储介质
CN114925920A (zh) * 2022-05-25 2022-08-19 中国平安财产保险股份有限公司 离线位置预测方法、装置、电子设备及存储介质
CN115408617A (zh) * 2022-09-15 2022-11-29 亢颖灏 基于人工智能的互联网用户整合方法及大数据服务系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111753214A (zh) * 2020-06-24 2020-10-09 平安科技(深圳)有限公司 基于行为轨迹的数据推送方法、系统和计算机设备
CN113868532B (zh) * 2021-09-30 2022-11-01 北京百度网讯科技有限公司 一种地点推荐方法、装置、电子设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015188324A1 (zh) * 2014-06-11 2015-12-17 华为技术有限公司 移动终端位置预测方法及装置
CN105279207A (zh) * 2014-07-25 2016-01-27 腾讯科技(深圳)有限公司 一种用户推荐方法、装置和系统
CN106022934A (zh) * 2016-05-05 2016-10-12 北京邮电大学 一种基于移动轨迹模式的潜在好友发现方法及系统
CN108733692A (zh) * 2017-04-18 2018-11-02 腾讯科技(深圳)有限公司 一种社交信息推荐方法和装置
CN111753214A (zh) * 2020-06-24 2020-10-09 平安科技(深圳)有限公司 基于行为轨迹的数据推送方法、系统和计算机设备

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109242553A (zh) * 2018-08-22 2019-01-18 平安科技(深圳)有限公司 一种用户行为数据推荐方法、服务器及计算机可读介质
CN109934704A (zh) * 2019-03-22 2019-06-25 深圳乐信软件技术有限公司 信息推荐方法、装置、设备和存储介质
CN110781342B (zh) * 2019-10-09 2024-02-02 上海麦克风文化传媒有限公司 一种基于用户行为序列和数据融合的推荐系统召回方法及系统
CN111061946B (zh) * 2019-11-15 2023-06-30 汉海信息技术(上海)有限公司 场景化内容推荐方法、装置、电子设备及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015188324A1 (zh) * 2014-06-11 2015-12-17 华为技术有限公司 移动终端位置预测方法及装置
CN105279207A (zh) * 2014-07-25 2016-01-27 腾讯科技(深圳)有限公司 一种用户推荐方法、装置和系统
CN106022934A (zh) * 2016-05-05 2016-10-12 北京邮电大学 一种基于移动轨迹模式的潜在好友发现方法及系统
CN108733692A (zh) * 2017-04-18 2018-11-02 腾讯科技(深圳)有限公司 一种社交信息推荐方法和装置
CN111753214A (zh) * 2020-06-24 2020-10-09 平安科技(深圳)有限公司 基于行为轨迹的数据推送方法、系统和计算机设备

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114418093A (zh) * 2022-01-19 2022-04-29 北京百度网讯科技有限公司 训练路径表征模型、输出信息的方法和装置
CN114418093B (zh) * 2022-01-19 2023-08-25 北京百度网讯科技有限公司 训练路径表征模型、输出信息的方法和装置
CN114785852A (zh) * 2022-04-07 2022-07-22 北京达佳互联信息技术有限公司 一种推送内容确定方法、装置、电子设备及存储介质
CN114785852B (zh) * 2022-04-07 2024-04-09 北京达佳互联信息技术有限公司 一种推送内容确定方法、装置、电子设备及存储介质
CN114925920A (zh) * 2022-05-25 2022-08-19 中国平安财产保险股份有限公司 离线位置预测方法、装置、电子设备及存储介质
CN114925920B (zh) * 2022-05-25 2024-05-03 中国平安财产保险股份有限公司 离线位置预测方法、装置、电子设备及存储介质
CN115408617A (zh) * 2022-09-15 2022-11-29 亢颖灏 基于人工智能的互联网用户整合方法及大数据服务系统

Also Published As

Publication number Publication date
CN111753214A (zh) 2020-10-09

Similar Documents

Publication Publication Date Title
WO2021259007A1 (zh) 基于行为轨迹的数据推送方法、系统和计算机设备
Li et al. Multiple location profiling for users and relationships from social network and content
Hang et al. Exploring student check-in behavior for improved point-of-interest prediction
Balaanand et al. An enhanced graph-based semi-supervised learning algorithm to detect fake users on Twitter
US9820094B2 (en) Travel recommendations on online social networks
Yuan et al. Measuring similarity of mobile phone user trajectories–a Spatio-temporal Edit Distance method
EP2820616B1 (en) Empirical expert determination and question routing system and method
US9311683B1 (en) Measuring engagement with a social networking platform
US8600797B1 (en) Inferring household income for users of a social networking system
WO2011094341A2 (en) System and method for social networking
US11490220B2 (en) System and method for accurately and efficiently generating ambient point-of-interest recommendations
WO2021159738A1 (zh) 基于医学领域的数据推荐方法、设备、服务器及存储介质
US20180060340A1 (en) Customized Keyword Query Suggestions on Online Social Networks
Cooper et al. Some typical properties of the spatial preferred attachment model
US11341109B2 (en) Method and system for detecting and using locations of electronic devices of users in a specific space to analyze social relationships between the users
Nishida et al. Probabilistic identification of visited point-of-interest for personalized automatic check-in
US20100250366A1 (en) Merge real-world and virtual markers
US10949434B1 (en) User identification system
CN109891190B (zh) 基于衍生社交网络对个人进行地理定位
Lu et al. Travel attractions recommendation with knowledge graphs
CN107809370B (zh) 用户推荐方法及装置
Ju et al. Relationship strength estimation based on Wechat Friends Circle
Zhang et al. Where are we in embedding spaces?
Garg et al. Smart geo-fencing with location sensitive product affinity
CN106575418B (zh) 建议的关键词

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21829694

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21829694

Country of ref document: EP

Kind code of ref document: A1