CN106919641B - Interest point searching method and device and electronic equipment - Google Patents

Interest point searching method and device and electronic equipment Download PDF

Info

Publication number
CN106919641B
CN106919641B CN201710020025.1A CN201710020025A CN106919641B CN 106919641 B CN106919641 B CN 106919641B CN 201710020025 A CN201710020025 A CN 201710020025A CN 106919641 B CN106919641 B CN 106919641B
Authority
CN
China
Prior art keywords
city
user
interest
preset
circle
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN201710020025.1A
Other languages
Chinese (zh)
Other versions
CN106919641A (en
Inventor
陈骐
彭承龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Sankuai Online Technology Co Ltd
Original Assignee
Beijing Sankuai Online Technology Co Ltd
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 Beijing Sankuai Online Technology Co Ltd filed Critical Beijing Sankuai Online Technology Co Ltd
Priority to CN201710020025.1A priority Critical patent/CN106919641B/en
Publication of CN106919641A publication Critical patent/CN106919641A/en
Application granted granted Critical
Publication of CN106919641B publication Critical patent/CN106919641B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The application provides an interest point searching method, belongs to the technical field of computers, and is used for solving the problem that in the prior art, interest points recalled in a searching process are not abundant. The method comprises the following steps: determining a user scene of a user generating a search behavior, constructing a city circle interest point library of a target city corresponding to the search behavior aiming at the user scene, and then, mixing, sequencing and recalling the interest points in the city circle interest point library based on a preset model. The interest point searching method disclosed by the application effectively enriches the recalled interest points by constructing the city circle interest point library, and meanwhile improves the accuracy of the recalled interest points by distinguishing user scenes.

Description

Interest point searching method and device and electronic equipment
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for searching for a point of interest, and an electronic device.
Background
With the development of internet technology and the explosive increase of information amount of internet platforms, some search engines generally search relevant information for users according to search keywords input by the users and in combination with the geographic positions of the users in order to provide completely personalized decision support and information services for the users. The common practice in the prior art is: the method comprises the steps that a search server firstly searches according to search keywords and search requests of users to obtain a plurality of search results; then, the search server further acquires the current geographic position of the user and the user identification of the user, and acquires the historical position information of the user according to the user identification; and finally, the search server performs sequencing optimization on the plurality of search results according to the historical position information so as to provide rich search results for the user. By adopting the method in the prior art, when a certain user searches tourist attractions in the place A, if tourist resources in the place A are limited, the problem of insufficient tourist attractions can occur.
Therefore, in the interest point searching method in the prior art, in the application of searching interest points related to geographical positions such as tourist attractions and the like, the problem that the recalled interest points are not abundant exists.
Disclosure of Invention
The application provides an interest point searching method, which solves the problem that in the prior art, recalled interest points are not abundant in a searching process.
In order to solve the above problem, in a first aspect, an embodiment of the present application provides a method for searching for a point of interest, including:
determining a user scenario of a user generating a search behavior;
aiming at the user scene, constructing a city circle interest point library of a target city corresponding to the search behavior;
and performing mixed sequencing and recalling the interest points in the city circle interest point library based on a preset model.
In a second aspect, an embodiment of the present application provides an apparatus for searching for a point of interest, including:
the user scene determining module is used for determining the user scene of the user generating the searching behavior;
the city circle interest point library construction module is used for constructing a city circle interest point library of a target city corresponding to the search behavior aiming at the user scene determined by the user scene determination module;
and the mixed sorting module is used for performing mixed sorting and recalling the interest points in the city circle interest point library constructed by the city circle interest point library construction module based on a preset model.
In a third aspect, an embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the point of interest search method disclosed in the embodiment of the present application when executing the computer program.
In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, performs the steps of the point of interest search method disclosed in the present application.
According to the interest point searching method disclosed by the embodiment of the application, the user scene of the user generating the searching behavior is determined, the city circle interest point library of the target city corresponding to the searching behavior is constructed according to the user scene, then the interest points in the city circle interest point library are mixed, ordered and recalled based on the preset model, and the problem that the interest points recalled in the searching process are not abundant in the prior art is solved. The method has the advantages that recalled interest points are effectively enriched by constructing the city circle interest point library, and meanwhile, accuracy of the recalled interest points is improved by distinguishing user scenes.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a flowchart of a point of interest search method according to a first embodiment of the present application;
FIG. 2 is a flowchart of a point of interest searching method according to a second embodiment of the present application;
FIG. 3 is a schematic view of the city circle of city A in the second embodiment of the present application;
FIG. 4 is a schematic structural diagram of a third embodiment of an interest point searching apparatus;
fig. 5 is a second schematic structural diagram of a third interest point searching apparatus according to a third embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The interest points related to the interest point searching method disclosed by the embodiment of the application can be scenic spots, restaurants, hotels and the like with regional differences.
Example one
As shown in fig. 1, a method for searching a point of interest disclosed in the present application includes: step 100 to step 120.
Step 100, determining a user scenario of a user generating a search behavior.
The search behavior of the user includes a search behavior initiated after the user inputs a search keyword on a certain page, a search behavior initiated automatically by a page program after the user browses a certain page, a search behavior initiated automatically by the program after a user performs a search related operation such as a screening operation on a page, and the like. The specific mode of initiating the search behavior by the user is not limited by the application.
In specific implementation, each search behavior corresponds to a city, which is referred to as a target city of the search behavior in the application. If the user enters the page of city a and then enters the search behavior initiated after the search keyword is entered on the page, the target city of the search behavior is city a. If the user selects city a by screening in the city list, the target city of the search behavior is city a.
In specific implementation, the user scene comprises a local user and a remote user. After a user accesses a certain page of an application or a website, the page of the application or the website can acquire the id of the user through a remote server, and further acquire data such as registration information, behavior logs and the like of the user, so as to acquire a resident city of the user. The page of the application or the website can also acquire the current positioning city of the user by positioning the user equipment. Then, determining the resident city or the current positioning city of the user generating the search behavior, and determining the user scene according to the target city and the resident city or the current positioning city of the user, for example: and if the target city corresponding to the searching behavior is the resident city or the current positioning city, determining that the user scene of the user is a local user, otherwise, determining that the user scene of the user is a remote user.
And step 110, constructing a city circle interest point library of the target city corresponding to the search behavior aiming at the user scene.
After determining a user scene of a user generating a search behavior, if the user scene is a local user, constructing a city circle interest point library of the local user of a target city corresponding to the search behavior; and if the user scene is a remote user, constructing a city circle interest point library of the remote user of the target city corresponding to the search behavior.
When constructing a city circle interest point library of a local user of a target city corresponding to the search behavior, firstly determining city correlation first scores of other cities related to the target city according to city dimensions respectively according to historical behavior logs (such as click or purchase logs) of the local user of the target city; determining a second relevance score of the city and other cities related to the target city according to the historical behavior logs of all users of the target city and the user dimension, and determining a city relevance score of the other cities related to the target city according to the first score and the second score; and then determining the city circle of the target city according to the obtained city relevance score, or determining the city circle of the target city according to the obtained city relevance score and the distance between each associated city and the target city. In particular, the city circle of the target city usually includes a plurality of cities. And finally, determining all interest points of each city in the city circle, which meet preset conditions, and forming a city circle interest point library of the local user of the target city.
When constructing a city circle interest point library of a remote user of a target city corresponding to the search behavior, firstly determining city relevance first scores of other cities related to the target city according to city dimensions and according to historical behavior logs of the remote user of the target city; determining a second relevance score of the city and other cities related to the target city according to the historical behavior logs of all users of the target city and the user dimension, and determining a city relevance score of the other cities related to the target city according to the first score and the second score; and then determining the city circle of the target city according to the obtained city relevance score, or determining the city circle of the target city according to the obtained city relevance score and the distance between each associated city and the target city. In particular, the city circle of the target city usually includes a plurality of cities. And finally, determining all interest points of each city in the city circle meeting preset conditions to form a city circle interest point library of the allopatric users of the target city.
And 120, performing mixed sequencing and recalling the interest points in the city circle interest point library based on a preset model.
In specific implementation, the city circle interest point library of the local user or the city circle interest point library of the remote user generated in the previous step comprises a plurality of interest points. In specific implementation, the interest points in the city circle interest point library include: interest points where a preset user action (e.g., clicking and/or purchasing a preset interest point) has occurred, recent popular interest points in each city.
In specific implementation, the preset model comprises: the click rate estimation model, the short-term purchase rate estimation model and the long-term purchase rate estimation model. When the interest points in the city circle interest point library are mixed, sorted and recalled based on a preset model, for each interest point in the city circle interest point library, respectively obtaining a sorting score of the corresponding model through a preset click rate estimation model, a short-term visit rate estimation model and a long-term visit rate estimation model; then, for each interest point in the city circle interest point library, obtaining a mixed ranking score of the interest point by synthesizing the ranking scores obtained by the models; and sequencing the interest points in the city circle interest point library according to the mixed sequencing score, and recalling.
According to the interest point searching method disclosed by the embodiment of the application, the user scene of the user generating the searching behavior is determined, the city circle interest point library of the target city corresponding to the searching behavior is constructed according to the user scene, then the interest points in the city circle interest point library are mixed, ordered and recalled based on the preset model, and the problem that the interest points recalled in the searching process are not abundant in the prior art is solved. The method has the advantages that recalled interest points are effectively enriched by constructing the city circle interest point library, and meanwhile, accuracy of the recalled interest points is improved by distinguishing user scenes.
Example two
As shown in fig. 2, a method for searching a point of interest disclosed in the present application includes: step 200 to step 240.
Step 200, a plurality of city interest point libraries are created.
In specific implementation, before constructing an interest point library of a city circle, the interest point library of each city is preset, and the interest point library of the city comprises: a local user point of interest library and a remote user point of interest library. The interest points in the local user interest point library are interest points concerned by local users in the city, and the interest points in the remote user interest point library are interest points concerned by remote users in the city.
And step 210, training a click rate estimation model and purchase rate estimation models in different periods.
In the embodiment of the application, when the click rate estimation model is trained, the search logs of interest points for T-T + N days are selected as training samples, the search logs of interest points for T + N + 1-T + N + M days are selected as test samples, wherein the search logs corresponding to the recalled interest points clicked by the user are used as positive samples, the search logs corresponding to the recalled interest points which are displayed but not clicked are used as negative samples, and the click rate estimation model is obtained through non-linear model training such as GBRT. Where N is the period of the training sample, for example, if N is 30, a search log of the interest point is selected for training for 30 days. M is the period of the test sample to evaluate the generalization ability of the model, for example, M is 5, and then the search logs of the interest points 5 days later are used to evaluate the model effect. In particular implementations, M and N may be integers greater than 1.
Then, the preset features of each interest point need to be extracted according to the interest point search logs. The extracted features include, but are not limited to, the following four dimensions: a user dimension, a point of interest quality dimension, an interaction dimension, and a context dimension. Wherein the user dimensions further include: sub-dimensions such as user labels, prices/categories/geographical preferences, POI-CF (Point of interest collaborative filtering) -based recommendation features, etc.; the point of interest quality dimension further includes: sub-dimensions of sales, clicks, comments, prices, conversion rates and the like of the interest points; the interaction dimension further includes: whether the user clicks/purchases/collects the interest point sub-dimension in the past first cycle/second cycle/third cycle; the context dimension further includes: the time, city, weather, distance from the target city, etc. of the current search behavior. The user tags are used for distinguishing local user interest points or allopatric user interest points. Features in the point of interest quality dimension distinguish local user points of interest from off-site user points of interest. Wherein, the first period can be 7 days, the second period can be 30 days, and the third period can be 90 days.
In specific implementation, the features are extracted from the search log of each interest point, and an interest point feature vector consisting of a plurality of preset features is obtained, such as (F)10，F11，F12，F20，F21，F30，...，F40,F41,...). And then, training the obtained feature vectors by utilizing non-linear model training such as GBRT and the like to obtain a corresponding click rate estimation model.
The purchase rate estimation model is a model for predicting the probability of the user being visited to purchase. In the specific implementation of the application, at least two different periods of the purchase rate estimation model are usually trained. When the purchase rate estimation model is trained, the search logs of the interest points in T-T + N days are selected as training samples, the search logs of the interest points in T + N + 1-T + N + M days are selected as testing samples, the search logs corresponding to the recall interest points of the user who has paid behavior are used as positive samples, the search logs corresponding to the recall interest points which are displayed but not clicked are used as negative samples, and the click rate estimation model is obtained through nonlinear model training such as GBRT. Where N is the cycle length of the training sample, for example, if N is 30, a log of 30 days is selected for training. When the purchasing rate estimation models in different periods are trained, the periods of the training samples are different, taking the training short-period purchasing rate estimation model and the training long-period purchasing rate estimation model as an example, and when the short-period purchasing rate estimation model is trained, the period of the samples can be selected to be N-30; when the long-period purchase rate estimation model is trained, the period of the sample can be selected to be N-365. M is the period of the test sample to evaluate the generalization ability of the model, for example, M is 5, and then the search log of the next 5 days is used to evaluate the model effect.
Then, the preset features of each interest point need to be extracted according to the interest point search logs. The extracted feature dimension and the extraction method are the same as those in the training of the click rate estimation model, and are not described again here.
In specific implementation, the features are extracted from each interest point search log, and an interest point feature vector formed by a plurality of preset features is obtained, such as (F)10，F11，F12，F20，F21，F30，...，F40,F41,...). And then, training the obtained feature vectors by utilizing non-linear model training such as GBRT and the like to obtain a corresponding purchase rate estimation model.
When the model is trained, the user dimension characteristics are extracted, the user labels are set to identify the interest points of local users or the interest points of allopatric users, and the characteristics of the quality dimension of the interest points are extracted according to the local users and the allopatric users respectively, so that the click rate estimation model and the visit rate estimation model obtained by training can fully represent the behavior characteristics of the local users and the allopatric users.
At step 220, the user context of the user that generated the search behavior is determined.
In specific implementation, the specific implementation manner of obtaining the search behavior of the user is referred to as the first embodiment, and details are not described here.
When the search behavior of the user is obtained, the target city of the search behavior can be determined at the same time, and the resident city or the current positioning city of the user initiating the search behavior can be determined. For the specific implementation of determining the target city of the search behavior, and determining the resident city or the current location city of the user initiating the search behavior, reference is made to embodiment one, and details are not described here.
In the implementation of the application, the user scene comprises a local user and a remote user. In specific implementation, the determining the user scenario of the user generating the search behavior includes: determining a resident city or a current positioning city of a user generating the search behavior; and if the target city corresponding to the searching behavior is the resident city or the current positioning city, determining that the user scene of the user is a local user, otherwise, determining that the user scene of the user is a remote user. In specific implementation, preferably, if the resident city of the user generating the search behavior is obtained, whether the target city corresponding to the search behavior is the resident city of the user generating the search behavior is judged, if the target city corresponding to the search behavior is the resident city or the current positioning city, the user scene of the user is determined to be the local user, and otherwise, the user scene of the user is determined to be the remote user. If the resident city of the user generating the search behavior is not obtained, obtaining the current positioning city of the user generating the search behavior, then judging whether the target city corresponding to the search behavior is the current positioning city of the user generating the search behavior, if the target city corresponding to the search behavior is the current positioning city, determining that the user scene of the user is a local user, otherwise, determining that the user scene of the user is a remote user.
Step 230, constructing a city circle interest point library of the target city corresponding to the search behavior aiming at the user scene.
After determining a user scene of a user generating a search behavior, if the user scene is a local user, constructing a city circle interest point library of the local user of a target city corresponding to the search behavior; and if the user scene is a remote user, constructing a city circle interest point library of the remote user of the target city corresponding to the search behavior. The step of constructing a city circle interest point library of the target city corresponding to the search behavior aiming at the user scene comprises the following steps: according to the historical behavior log of the user, according to the city correlation of the user dimension and the city correlation of the city dimension, constructing a city circle, matched with the user scene, of the target city corresponding to the search behavior; and determining all interest points of all cities in the city circle meeting preset conditions to form a city circle interest point library of the target city.
In specific implementation, according to the historical behavior log of the user, the city circle of the target city corresponding to the search behavior and matched with the user scene is constructed according to the city relevance of the user dimension and the city relevance of the city dimension, and the method comprises the following substeps S1 to substep S5.
Substep S1: and determining a city relevance first score of a city associated with the target city according to city dimensions according to the historical behavior log matched with the user scene.
When the city correlation first score of the city associated with the target city is determined according to the historical behavior log matched with the user scene and the city dimension, if the user scene is a local user, determining the city correlation first score of the city associated with the target city according to the historical behavior log of the local user of the target city and the city dimension; and if the user scene is a remote user, determining a first city relevance score of a city associated with the target city according to city dimensions according to the historical behavior log of the remote user of the target city.
Determining, according to the historical behavior log matched with the user scenario, a city relevance first score for a city associated with the target city according to a city dimension includes: acquiring a preset user behavior log which takes the target city as an access entry in historical behavior logs matched with the user scene, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point; determining the city dimension single score of each interest point according to the preset user behavior log; aggregating the single city dimension scores of all interest points according to the city to which the interest point belongs to obtain city correlation first scores of the cities to which the interest point belongs and the target city; wherein the user scene is: a local user or a remote user; if the user scene is a local user, the historical behavior log matched with the user scene is the historical behavior log of the local user of the target city; and if the user scene is a remote user, the historical behavior log matched with the user scene is the historical behavior log of the remote user in the target city.
In specific implementation, the historical behavior log of the user comprises the identification of a local user record and a remote user record, and all the historical behavior logs of the local user in the target city or all the historical behavior logs of the remote user can be acquired according to the identification. The specific implementation of obtaining the user historical behavior log of the target city refers to the prior art, and is not described herein again.
Taking a target city as a city a, wherein the user historical behavior log of the city a includes logs of users U1 and U2, wherein the user U1 is a resident user of the city a, and the user U2 is not a resident user of the city a, a specific process of obtaining a preset user behavior log which takes the target city as an access entry in the historical behavior log matched with the user scenario is described below. Suppose that the user historical behavior log includes point of interest 1 for city B purchased after user U1 entered from the page of city a; after entering from the page of city a, the user U1 clicks the interest point 2 of city B; after entering from the page of city a, user U2 clicked on point of interest 2 of city B. Taking an application scene as a local user as an example, acquiring a preset user behavior log which takes the target city as an access entry in historical behavior logs matched with the user scene comprises the following steps: after entering from the page of city a, user U1 purchased point of interest 1 of city B; after entering from the page of city a, user U1 clicked on point of interest 2 of city B. Taking an application scene as an example of a remote user, acquiring a preset user behavior log which takes the target city as an access entry in historical behavior logs matched with the user scene comprises the following steps: after entering from the page of city a, user U2 clicked on point of interest 2 of city B. The specific implementation of obtaining the preset user behavior log which takes the target city as an access entry in the historical behavior log matched with the user scene is referred to in the prior art, and details are not repeated here.
And when the single urban dimension score of each interest point is determined according to the preset user behavior log, summing the operated times (such as clicked, purchased, shared and the like) of the interest points through a summing formula with a time attenuation factor, and taking the sum as the single urban dimension score of the interest points. The summation formula provided with the time attenuation factor is exemplified as follows:
Figure BDA0001207902980000101
wherein, T1Is a calculation cycle; i is the distance from the current timeDays in between, orderid (i) is a log of pre-set actions (e.g., purchases and/or clicks) made to a point of interest in a city on day i prior to the current time; count () is a count function; and a is a time attenuation coefficient and takes a value between (0, 1). As can be seen from the above formula, the longer the preset user behavior log is from the current time, the smaller the weight is when calculating the single score of the city dimension of a certain point of interest. In specific practice, T1The value of (A) is set according to specific requirements, such as T1Day 28.
In practice, orderid (i) may also be a log of preset behaviors that are generated for a certain point of interest in a certain city within the ith time period before the current time (e.g., dividing a month before the current time into 4 weeks, and taking each week as a time period). The preset behavior can be one or more of user behaviors such as clicking, purchasing, recommending and the like. In specific implementation, a can beiOr other forms of expression that gradually decrease with increasing i.
After the preset user behavior logs in the historical behavior logs matched with the user scene are obtained, interest points generating the preset user behaviors and generation time are recorded in each log, the logs of each interest point, in which the preset user behaviors occur, can be added into a set, and then the city dimension single score of each interest point is calculated according to the formula. Assuming that 5 interest points are determined according to the preset user behavior log in the obtained historical behavior log matched with the user scene, the city dimension single score of the 5 interest points can be obtained, for example: POI _ Score1 for point of interest 1, POI _ Score2 for point of interest 2, POI _ Score3 for point of interest 3, POI _ Score4 for point of interest 4, and POI _ Score5 for point of interest 5.
In specific implementation, each city has a preset interest point, that is, a corresponding relationship between an interest point and a city is preset. For example, city B has point of interest 1, point of interest 2, and point of interest 3; city C has points of interest 4 and points of interest 5. Aggregating the city dimension single scores of all interest points according to the city to which the interest point belongs to obtain the city relevance of each city to which the interest point belongs and the target cityA score. For example, the city dimension single scores of interest point 1, interest point 2 and interest point 3 are aggregated to obtain a city relevance first score of city B and the target city a, such as W1AB(ii) a Aggregating the single city dimension scores of the interest points 4 and 5 to obtain a first city relevance score of the city C and the target city A, such as W1AC. In practice, the polymerization method may be accumulation or averaging, etc., which are not exemplified herein.
And a substep S2 of determining a city relevance second score for the city associated with the target city according to the user dimension based on the historical behavior log of the user.
Determining a city relevance second score of a city associated with the target city according to the user dimension according to the historical behavior log of the user, wherein the city relevance second score comprises the following steps: acquiring a preset user behavior log of the historical behavior log of the resident user of the target city, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point; determining a user dimension single score of each interest point according to the preset user behavior log; and aggregating the user dimension single scores of all the interest points according to the cities to which the interest points belong to obtain city relevance second scores of the cities to which the interest points belong and the target city.
In specific implementation, the historical behavior log of the user includes user identification, interest points, preset user behaviors for the interest points, behavior occurrence places and the like. And further acquiring the resident city of the user according to the user identification, and judging whether the user is the resident user of the target city. In the embodiment of the application, a preset user behavior log of the historical behavior log of the resident user in the target city is obtained and used as a basis for calculating the single score of the user dimension. In specific implementation, logs of all user resident users of the target city can be obtained first, for a certain city, the behavior of the resident users is more representative, and the data calculation amount can be effectively reduced. Then, the logs of the preset user behaviors (such as clicking and purchasing interest points) are extracted and used as a basis for calculating the user dimension scores.
The specific implementation of obtaining the user historical behavior log of the target city refers to the prior art, and is not described herein again.
And when the user dimension single score of each interest point is determined according to the preset user behavior log, summing the operated times (such as clicked, purchased, shared and the like) of the interest points through a summing formula with a behavior acceleration factor, and taking the sum as the user dimension single score of the interest points. The summation formula provided with the behavior acceleration factor is exemplified as follows:
Figure BDA0001207902980000121
wherein i1And i2For the number of days of each time period, i1>i2(ii) a orderid (i) is a log of preset actions (e.g., purchases and/or clicks) that occur at a point of interest in a city during i days prior to the current time; count () is a count function; n and m are the number of periods; a is1And a2Is a behavior acceleration factor, a1<a2. As can be seen from the above formula, the weight of the long-period preset user behavior log is smaller than that of the short-period preset user behavior log when calculating the user dimension single score of a certain interest point. In specific practice, i1Can take 30, n can take 3, i2Can take the value of 7, and m can take the value of 4.
After the preset user behavior logs in the historical behavior logs of the user are obtained, the interest points generating the preset user behaviors and the generation time are recorded in each log, the logs of each interest point, which have generated the preset user behaviors, can be added into a set, and then the user dimension score of each interest point is calculated according to the formula. Assuming that 5 interest points are determined according to a preset user behavior log in the obtained historical behavior log of the user, user dimension scores of the 5 interest points can be obtained, such as: POI _ Score6 for point of interest 1, POI _ Score7 for point of interest 2, POI _ Score8 for point of interest 3, POI _ Score9 for point of interest 4, and POI _ Score10 for point of interest 5.
In specific implementation, each city has a preset interest point, that is, a corresponding relationship between an interest point and a city is preset. For example, city B has point of interest 1, point of interest 2, and point of interest 3; city C has points of interest 4 and points of interest 5. And aggregating the user dimension single scores of all the interest points according to the cities to which the interest points belong to obtain city relevance second scores of the cities to which the interest points belong and the target city. For example, the user dimension scores of interest point 1, interest point 2, and interest point 3 are aggregated to obtain a second city relevance score, such as W, between city B and the target city a2AB(ii) a Aggregating the user dimension scores of interest points 4 and 5 to obtain a city relevance second score of city C and the target city A, such as W2AC. In practice, the polymerization method may be accumulation or averaging, etc., which are not exemplified herein.
And a substep S3, for each city associated with the target city, fusing the city relevance first score and the second score to obtain a city relevance score of the city associated with the target city and matched with the user scenario.
In specific implementation, for each city associated with the target city, fusing the city relevance first score and the second score to obtain a city relevance score of a city associated with the target city and matched with the user scenario, including: and for each city associated with the target city, performing weighted fusion on the city relevance first score and the second score to obtain a city relevance score of the city associated with the target city and matched with the user scene. With weighted additive fusion, the weights can be adjusted according to the actual situation, such as WAB＝0.7×W1AB+0.3×W2AB。
In specific implementation, when the first city relevance score and the second city relevance score are fused, a relevance factor can be artificially increased by combining with domain knowledge. For example, for cities belonging to a geographical concept (e.g. Hunan West, Long triangular, etc.) in geography and customs, people can be involvedIncrease and assign W3ABThen according to the formula WAB＝0.7×W1AB+0.3×W2AB+W3ABCalculating a city relevance score for a city associated with the target city and matching the user scenario. By increasing the relevance factor according to the relevance condition set by the user, the relevance between cities can be enhanced, and the interest points can be recommended conveniently based on the relevance set by the user.
And a substep S4, taking the city corresponding to the city relevance score larger than a preset relevance score threshold value as a candidate city of the city circle matched with the user scene of the target city.
In specific implementation, a relevance score threshold may be preset, and when the calculated city relevance score of the city associated with the target city is greater than the preset relevance score threshold, the city is determined to be a city in the city circle of the target city. The correlation score threshold is determined according to specific requirements and can be set to a value greater than or equal to 0.
The determined candidate cities of the city circle of the target city typically include a plurality of associated cities. As shown in fig. 3, for city a, the cities in its city circle 300 include cities: B. c, D and E.
And a substep S5, forming a city circle of the target city, which is matched with the user scene, by the alternative cities, the distances of which from the target city to the alternative cities meet a preset distance threshold value.
In specific implementation, the preset distance threshold is a required radius defined according to user requirements, such as 1 km. By filtering a part of cities far away from the target city according to the distance between the city in the city circle and the target city, the calculation amount can be reduced and the accuracy of recalling the interest points can be improved on the premise of meeting the radius required by the user. Suppose that the city relevance scores for city A and cities B, C, D and E, respectively, are calculated as WAB、WAC、WADAnd WAEAre greater than a preset relevance score threshold, and the distance between city A and cities B, C, D and E is DAB、DAC、DADAnd DAE. If D isAB、DAC、DADAnd DAEIf the distance is less than or equal to the preset distance threshold, the city circle of the city a comprises the following cities: cities A, B, C, D and E. If D isAEIf the distance is greater than the preset distance threshold, the city circle of the city a includes the following cities: cities A, B, C and D.
In a specific implementation, the preset distance threshold may also be a distance range, such as within 10 km to 100 km.
The determined city circle for the target city typically includes a plurality of cities associated with the target city. After the city circle of the target city is determined, further determining all interest points of all cities in the city circle, which meet preset conditions, to form a city circle interest point library of the target city.
In specific implementation, the user scenario includes: a local user or a displaced user. The determining all interest points of each city in the city circle meeting preset conditions to form a city circle interest point library of the target city includes: if the user scene is a local user, determining interest points of the local user of the target city, which have preset user behaviors, interest points of other cities except the target city in a city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period; if the user scene is a remote user, determining interest points of remote users of the target city, which have preset user behaviors, interest points of other cities except the target city in the city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period; wherein the preset user behavior comprises: click and/or purchase a predetermined point of interest.
Taking the target city as a city A, wherein the cities in the city circle comprise cities: A. b, C, D, a specific method for determining all points of interest of each city in the city circle that satisfy the preset conditions to form the city circle point of interest library of the target city is described.
If the user scene is a local user, adding an interest point in which a preset user action (such as clicking and/or purchasing) occurs in the local user interest points in the city A into a city circle interest point library of the target city A; then, the interest points in the cities B, C, D and E, which have occurred with the city A as the pre-set user behavior (e.g., clicking and/or purchasing) of the access portal, are added to the city circle interest point library of the target city A.
In order to enrich recalled interest points, after adding the interest points which are determined according to the historical behavior log of the user and have preset user behaviors into the city circle interest point library, further adding popular interest points (such as the interest points with the highest sales volume) of all cities in the city circle in a preset time period into the city circle interest point library.
If the user scene is a remote user, adding an interest point in which a preset user action (such as clicking and/or purchasing) occurs in the remote user interest points in the city A into the city circle interest point library of the target city A; then, the interest points in the cities B, C, D and E, which have occurred the preset user behavior with the city A as the access entrance (e.g., with the city A as the access entrance and performing the clicking and/or purchasing operation), are added to the city circle interest point library of the target city A.
Similarly, in order to enrich recalled points of interest, popular points of interest (e.g., the points of interest with the highest sales volume) of each city in the city circle within a preset time period are further added to the city circle point of interest library of the target city a.
In specific implementation, the second city relevance score of the city associated with the target city may be determined according to the user dimension according to the historical behavior log of the user, and then the first city relevance score of the city associated with the target city may be determined according to the city dimension according to the historical behavior log matched with the user scenario, where the specific execution order of the substep S1 and the substep S2 is not limited.
And 240, performing mixed sequencing and recalling the interest points in the city circle interest point library based on a preset model.
In specific implementation, the city circle interest point library of the local user or the city circle interest point library of the remote user generated in the previous step comprises a plurality of interest points. In specific implementation, based on a preset model, performing mixed sorting and recalling on the interest points in the city circle interest point library, including: for each interest point in the city circle interest point library, respectively obtaining the ranking score of the corresponding model through a preset model; for each interest point in the city circle interest point library, integrating the ranking scores obtained through the models to obtain a mixed ranking score of the interest point; sorting the interest points in the city circle interest point library according to the mixed sorting score, and recalling; wherein the preset model comprises: the click rate estimation model and the purchase rate estimation model of at least two different periods; the purchase rate estimation model is obtained by training according to the interest point purchase log in a set length period and is used for predicting the purchase rate of the interest points. In the embodiment, the preset models include a click rate estimation model, a short-term purchase rate estimation model and a long-term purchase rate estimation model.
For example, after determining the city circle of interest point library that needs to be returned to the city a for the current search behavior according to the user scenario and the target city, all the interest points in the city circle of interest point library of the city a need to be sorted. Suppose that the local user city circle interest point library of city a includes: POI 1 for city A, POI 2 for city A, POI 3 for city B, POI 4 for city B, POI 5 for city C, POI 6 for city D, and POI 7 for city E. For the interest point 1, firstly, the corresponding models are obtained and ranked through a preset click rate estimation model, a short-term purchase rate estimation model and a long-term purchase rate estimation model respectively. Namely, the click rate estimated ranking score1 is obtained through a preset click rate estimated model, the short-term purchase rate estimated ranking score2 is obtained through a short-term purchase rate estimated model, and the long-term purchase rate estimated ranking score3 is obtained through a long-term purchase rate estimated model. Then, the ranking scores obtained by the models are integrated to obtain a mixed ranking score of the interest point. For example, for each moldWeighted summation of ranking scores obtained from the types, or according to a formula
Figure BDA0001207902980000161
A hybrid ranking score for the points of interest is calculated.
According to the foregoing method, scores of mixed ranking of the points of interest 1 to 7, respectively, can be obtained. And finally, sequencing the interest points in the city circle interest point library according to the mixed sequencing score, and recalling.
When the corresponding model score is obtained through each preset model, the preset features of each interest point need to be extracted. The extracted features include, but are not limited to, the following four dimensions: a user dimension, a point of interest quality dimension, an interaction dimension, and a context dimension. Wherein the user dimensions further include: sub-dimensions such as user labels, prices/categories/geographical preferences, POI-CF (Point of interest collaborative filtering) -based recommendation features, etc.; the point of interest quality dimension further includes: sub-dimensions of sales, clicks, comments, prices, conversion rates and the like of the interest points; the interaction dimension further includes: whether the user clicks/purchases/collects the interest point sub-dimension in the past first cycle/second cycle/third cycle; the context dimension further includes: the time, city, weather, distance from the target city, etc. of the current search behavior. The user tags are used for distinguishing local user interest points or allopatric user interest points. Features in the point of interest quality dimension distinguish local user points of interest from off-site user points of interest. Wherein, the first period can be 7 days, the second period can be 30 days, and the third period can be 90 days.
In specific implementation, the click rate estimation model, the short-term purchase rate estimation model and the long-term purchase rate estimation model respectively extract preset features from the current interest points to form feature vectors. Taking click rate estimation model H as an example, when model H obtains the click rate estimation score of a certain interest point, the model H firstly extracts the preset characteristics of the interest point to form a characteristic vector such as (F)10，F11，F12，F20，F21，F30，...，F40,F41,..) the predetermined characteristics include the four dimensions described aboveA feature vector of degrees; and then, calculating the ranking score corresponding to the click rate estimation model of the current interest point by the model H according to the extracted features.
The method for extracting corresponding features from the interest points and the method for combining the extracted dimensional features into the overall features of the interest points are the same as the method for extracting features from training samples and combining features when training the models, and are not repeated here.
The ranking scores of the interest points are calculated by using the visit rate estimation models in different periods so as to be used for mixed ranking, the interest points are recalled by combining the periodic and seasonal characteristics of the demand of the interest points, the recall result is more in line with the demand of the user, and the user experience is further improved. The sensing ability of seasonal variation can be improved through the short-period model, and the stability of the interest point recall effect is improved through the long-period model. In the O2O consumption field, the click rate estimation model and the visit rate prediction model are combined to perform sorting sub-calculation so as to achieve the balance between user interest mining and final demand meeting.
According to the interest point searching method disclosed by the embodiment of the application, a click rate estimation model and a plurality of visiting and purchasing rate estimation models in different periods are trained in advance, after a searching behavior of a user is obtained, a user scene of the user generating the searching behavior is determined, a city circle interest point library of a target city corresponding to the searching behavior is established according to the user scene, then interest points in the city circle interest point library are mixed, ordered and recalled based on the preset model, and the problem that the interest points recalled in the searching process are not abundant in the prior art is solved. The method has the advantages that recalled interest points are effectively enriched by constructing the city circle interest point library, and meanwhile, accuracy of the recalled interest points is improved by distinguishing user scenes.
The interest points are mixed and sorted by combining the sorting scores of the interest points calculated by the click rate pre-estimation model and the purchase rate pre-estimation models in different periods, so that the interest points are recalled by combining the periodic and seasonal characteristics of the demand of the interest points, the recall result is more in line with the demand of a user, and the user experience is further improved.
EXAMPLE III
As shown in fig. 4, an interest point searching apparatus disclosed in this embodiment includes:
a user scenario determination module 400 for determining a user scenario of a user that generated a search action;
a city circle interest point library constructing module 410, configured to construct, for the user scenario determined by the user scenario determining module 400, a city circle interest point library of the target city corresponding to the search behavior;
and a mixed ranking module 420, configured to perform mixed ranking and recall on the interest points in the city circle interest point library constructed by the city circle interest point library construction module 410 based on a preset model.
In the implementation of the application, the user scene comprises a local user and a remote user. In specific implementation, the user context determining module 400 is specifically configured to: determining a resident city or a current positioning city of a user generating the search behavior; and if the target city corresponding to the searching behavior is the resident city or the current positioning city, determining that the user scene of the user is a local user, otherwise, determining that the user scene of the user is a remote user.
Optionally, as shown in fig. 5, the city circle interest point library building module 410 includes:
the city circle constructing unit 4101 is configured to construct, according to a historical behavior log of a user, a city circle of a target city corresponding to the search behavior and matching with the user scene according to the city relevance of the user dimension and the city relevance of the city dimension;
an interest point database construction unit 4102 for determining all interest points of each city in the city circle, which satisfy a preset condition, to form an interest point database for the city circle of the target city.
Optionally, as shown in fig. 5, the city circle constructing unit 4101 includes:
a city dimension score determining subunit 41011, configured to determine, according to a history behavior log matched with the user scenario, a city relevance first score of a city associated with the target city according to a city dimension; and the number of the first and second groups,
a user dimension score determining subunit 41012, configured to determine, according to the historical behavior log of the user, a second city relevance score of the city associated with the target city according to the user dimension;
a score fusion subunit 41013, configured to fuse, for each city associated with the target city, the city relevance first score and the second score to obtain a city relevance score of a city associated with the target city and matching the user scenario;
a relevant city determining subunit 41014, configured to use a city corresponding to the city relevance score greater than a preset relevance score threshold as an alternative city of the city circle of the target city that matches the user scenario;
a city circle constructing subunit 41015, configured to construct a city circle of the target city, which is matched with the user scenario, from the candidate cities whose distances from the target city satisfy a preset distance threshold.
Optionally, the city dimension score determining subunit 41011 is specifically configured to:
acquiring a preset user behavior log which takes the target city as an access entry in historical behavior logs matched with the user scene, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point;
determining the city dimension single score of each interest point according to the preset user behavior log;
aggregating the single city dimension scores of all interest points according to the city to which the interest point belongs to obtain city correlation first scores of the cities to which the interest point belongs and the target city;
wherein the user scene is: a local user or a remote user; if the user scene is a local user, the historical behavior log matched with the user scene is the historical behavior log of the local user of the target city; and if the user scene is a remote user, the historical behavior log matched with the user scene is the historical behavior log of the remote user in the target city.
Optionally, the user dimension score determining subunit 41012 is specifically configured to:
acquiring a preset user behavior log of the historical behavior log of the resident user of the target city, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point;
determining a user dimension single score of each interest point according to the preset user behavior log;
and aggregating the user dimension single scores of all the interest points according to the cities to which the interest points belong to obtain city relevance second scores of the cities to which the interest points belong and the target city.
Optionally, the user scenario includes: the city circle interest point library constructing unit 4102 is specifically configured to:
if the user scene is a local user, determining interest points of the local user of the target city, which have preset user behaviors, interest points of other cities except the target city in a city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period;
if the user scene is a remote user, determining interest points of remote users of the target city, which have preset user behaviors, interest points of other cities except the target city in the city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period;
wherein the preset user behavior comprises: click and/or purchase a predetermined point of interest.
Optionally, the mixing and sorting module 420 is specifically configured to:
for each interest point in the city circle interest point library, respectively obtaining the ranking score of the corresponding model through a preset model;
for each interest point in the city circle interest point library, integrating the ranking scores obtained through the models to obtain a mixed ranking score of the interest point;
sorting the interest points in the city circle interest point library according to the mixed sorting score, and recalling;
wherein the preset model at least comprises: the click rate estimation model and the purchase rate estimation model of at least two different periods; the purchase rate estimation model is obtained by training according to the interest point purchase log in a set length period and is used for predicting the purchase rate of the interest points.
For specific implementation of each module, unit and subunit of the interest point searching apparatus disclosed in this embodiment, reference is made to relevant steps in the method embodiment, and details are not described here again.
According to the interest point searching device disclosed by the embodiment of the application, after the searching behavior of the user is obtained, the user scene of the user generating the searching behavior is determined, the city circle interest point library of the target city corresponding to the searching behavior is constructed according to the user scene, then the interest points in the city circle interest point library are mixed, ordered and recalled based on the preset model, and the problem that the interest points recalled in the searching process are not abundant in the prior art is solved. The method has the advantages that recalled interest points are effectively enriched by constructing the city circle interest point library, and meanwhile, accuracy of the recalled interest points is improved by distinguishing user scenes.
The interest points are mixed and sorted by combining the sorting scores of the interest points calculated by the click rate pre-estimation model and the purchase rate pre-estimation models in different periods, so that the interest points are recalled by combining the periodic and seasonal characteristics of the demand of the interest points, the recall result is more in line with the demand of a user, and the user experience is further improved.
The application also discloses an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and is characterized in that the processor implements the interest point searching method in the first embodiment and the second embodiment when executing the computer program. The electronic device can be a PC, a mobile terminal, a personal digital assistant, a tablet computer and the like.
The present application also discloses a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps of the point of interest search method as described in the first and second embodiments of the present application.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
The method and the device for searching for the point of interest provided by the application are introduced in detail, a specific example is applied in the text to explain the principle and the implementation of the application, and the description of the embodiment is only used for helping to understand the method and the core idea of the application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.

Claims (12)

1. A method for searching a point of interest, comprising:
determining a user scene of a user generating a search behavior, wherein the user scene comprises a local user or a remote user;
aiming at the user scene, constructing a city circle interest point library of a target city corresponding to the search behavior;
based on a preset model, carrying out mixed sequencing and recalling on the interest points in the city circle interest point library;
wherein, the step of constructing the city circle interest point library of the target city corresponding to the search behavior aiming at the user scene comprises the following steps:
according to the historical behavior log of the user, according to the city correlation of the user dimension and the city correlation of the city dimension, constructing a city circle, matched with the user scene, of the target city corresponding to the search behavior;
determining all interest points of each city in the city circle, which meet preset conditions, and forming a city circle interest point library of the target city;
wherein, the determining all interest points of each city in the city circle which meet the preset conditions to form the city circle interest point library of the target city comprises:
if the user scene is a local user, determining interest points of the local user of the target city, which have preset user behaviors, interest points of other cities except the target city in a city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period;
if the user scene is a remote user, determining interest points of remote users of the target city, which have preset user behaviors, interest points of other cities except the target city in the city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period;
wherein the preset user behavior comprises: click and/or purchase a predetermined point of interest.
2. The method according to claim 1, wherein the step of constructing a city circle matching the user scene of the target city corresponding to the search behavior according to the historical behavior log of the user, the city correlation of the user dimension, and the city correlation of the city dimension comprises:
determining a city relevance first score of a city associated with the target city according to city dimensions according to a historical behavior log matched with the user scene; and the number of the first and second groups,
determining a city relevance second score of a city associated with the target city according to the user dimension according to the historical behavior log of the user;
for each city associated with the target city, fusing the city relevance first score and the second score to obtain a city relevance score of the city associated with the target city and matched with the user scene;
taking the city corresponding to the city relevance score larger than a preset relevance score threshold value as an alternative city of the city circle matched with the user scene of the target city;
and forming a city circle, matched with the user scene, of the target city by using the alternative cities, the distances of which to the target city meet a preset distance threshold value.
3. The method of claim 2, wherein the step of determining a city relevance first score for a city associated with the target city by city dimension based on historical behavior logs matching the user scenario comprises:
acquiring a preset user behavior log which takes the target city as an access entry in historical behavior logs matched with the user scene, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point;
determining the city dimension single score of each interest point according to the preset user behavior log;
aggregating the single city dimension scores of all interest points according to the city to which the interest point belongs to obtain city correlation first scores of the cities to which the interest point belongs and the target city;
wherein the user scene is: a local user or a remote user; if the user scene is a local user, the historical behavior log matched with the user scene is the historical behavior log of the local user of the target city; and if the user scene is a remote user, the historical behavior log matched with the user scene is the historical behavior log of the remote user in the target city.
4. The method of claim 2, wherein the step of determining a city relevance second score for the city associated with the target city by user dimension based on the user's historical behavior log comprises:
acquiring a preset user behavior log of the historical behavior log of the resident user of the target city, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point;
determining a user dimension single score of each interest point according to the preset user behavior log;
and aggregating the user dimension single scores of all the interest points according to the cities to which the interest points belong to obtain city relevance second scores of the cities to which the interest points belong and the target city.
5. The method of claim 1, wherein the step of performing mixed ranking and recalling of the points of interest in the city circle point of interest library based on the preset model comprises:
for each interest point in the city circle interest point library, respectively obtaining the ranking score of the corresponding model through a preset model;
for each interest point in the city circle interest point library, integrating the ranking scores obtained through the models to obtain a mixed ranking score of the interest point;
sorting the interest points in the city circle interest point library according to the mixed sorting score, and recalling;
wherein the preset model at least comprises: the click rate estimation model and the purchase rate estimation model of at least two different periods; the purchase rate estimation model is obtained by training according to the interest point purchase log in a set length period and is used for predicting the purchase rate of the interest points.
6. An apparatus for searching for a point of interest, comprising:
a user context determination module for determining a user context of a user generating a search action, the user context comprising: a local user or a remote user;
the city circle interest point library construction module is used for constructing a city circle interest point library of a target city corresponding to the search behavior aiming at the user scene determined by the user scene determination module;
the mixed ordering module is used for carrying out mixed ordering and recalling the interest points in the city circle interest point library constructed by the city circle interest point library construction module based on a preset model;
the city circle interest point library construction module comprises:
the city circle construction unit is used for constructing a city circle, matched with the user scene, of a target city corresponding to the search behavior according to the historical behavior log of the user and the city correlation of the user dimension and the city correlation of the city dimension;
the city circle interest point library construction unit is used for determining all interest points of each city in the city circle, which meet preset conditions, and forming a city circle interest point library of the target city;
the city circle interest point library construction unit is specifically used for:
if the user scene is a local user, determining interest points of the local user of the target city, which have preset user behaviors, interest points of other cities except the target city in a city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period;
if the user scene is a remote user, determining interest points of remote users of the target city, which have preset user behaviors, interest points of other cities except the target city in the city circle which takes the target city as an access entrance and has preset user behaviors, and popular interest points of all cities in the city circle within a preset time period;
wherein the preset user behavior comprises: click and/or purchase a predetermined point of interest.
7. The apparatus of claim 6, wherein the city circle building unit comprises:
a city dimension score determining subunit, configured to determine, according to a city dimension, a city relevance first score of a city associated with the target city, according to a historical behavior log matched with the user scenario; and the number of the first and second groups,
the user dimension score determining subunit is used for determining a second city relevance score of the city associated with the target city according to the user dimension according to the historical behavior log of the user;
a score fusion subunit, configured to fuse, for each city associated with the target city, the city relevance first score and the second score to obtain a city relevance score of a city associated with the target city and matched with the user scenario;
a relevant city determining subunit, configured to use a city corresponding to the city relevance score greater than a preset relevance score threshold as an alternative city of the city circle of the target city matching the user scenario;
and the city circle constructing subunit is used for constructing the city circle, matched with the user scene, of the target city by using the alternative cities, the distance between which and the target city meets a preset distance threshold value.
8. The apparatus of claim 7, wherein the city dimension score determination sub-is specifically configured to:
acquiring a preset user behavior log which takes the target city as an access entry in historical behavior logs matched with the user scene, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point;
determining the city dimension single score of each interest point according to the preset user behavior log;
aggregating the single city dimension scores of all interest points according to the city to which the interest point belongs to obtain city correlation first scores of the cities to which the interest point belongs and the target city;
wherein the user scene is: a local user or a remote user; if the user scene is a local user, the historical behavior log matched with the user scene is the historical behavior log of the local user of the target city; and if the user scene is a remote user, the historical behavior log matched with the user scene is the historical behavior log of the remote user in the target city.
9. The apparatus of claim 7, wherein the user dimension score determining subunit is specifically configured to:
acquiring a preset user behavior log of the historical behavior log of the resident user of the target city, wherein the preset user behavior comprises the following steps: clicking and/or purchasing a preset interest point;
determining a user dimension single score of each interest point according to the preset user behavior log;
and aggregating the user dimension single scores of all the interest points according to the cities to which the interest points belong to obtain city relevance second scores of the cities to which the interest points belong and the target city.
10. The apparatus of claim 6, wherein the mix ordering module is specifically configured to:
for each interest point in the city circle interest point library, respectively obtaining the ranking score of the corresponding model through a preset model;
for each interest point in the city circle interest point library, integrating the ranking scores obtained through the models to obtain a mixed ranking score of the interest point;
sorting the interest points in the city circle interest point library according to the mixed sorting score, and recalling;
wherein the preset model at least comprises: the click rate estimation model and the purchase rate estimation model of at least two different periods; the purchase rate estimation model is obtained by training according to the interest point purchase log in a set length period and is used for predicting the purchase rate of the interest points.
11. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the point of interest search method of any one of claims 1 to 5 when executing the computer program.
12. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the point of interest search method of any one of claims 1 to 5.
CN201710020025.1A 2017-01-12 2017-01-12 Interest point searching method and device and electronic equipment Active CN106919641B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710020025.1A CN106919641B (en) 2017-01-12 2017-01-12 Interest point searching method and device and electronic equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710020025.1A CN106919641B (en) 2017-01-12 2017-01-12 Interest point searching method and device and electronic equipment

Publications (2)

Publication Number Publication Date
CN106919641A CN106919641A (en) 2017-07-04
CN106919641B true CN106919641B (en) 2020-04-17

Family

ID=59453985

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710020025.1A Active CN106919641B (en) 2017-01-12 2017-01-12 Interest point searching method and device and electronic equipment

Country Status (1)

Country Link
CN (1) CN106919641B (en)

Families Citing this family (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107396432A (en) 2017-07-20 2017-11-24 深圳市沃特沃德股份有限公司 Position control method and device
CN107315841A (en) * 2017-07-20 2017-11-03 北京三快在线科技有限公司 A kind of information search method, apparatus and system
CN108009877B (en) * 2017-11-24 2021-10-15 创新先进技术有限公司 Information mining method and device
CN107895213A (en) * 2017-12-05 2018-04-10 北京三快在线科技有限公司 Forecasting Methodology, device and the electronic equipment of spending limit
CN108288179B (en) * 2018-01-25 2021-02-02 贝壳找房(北京)科技有限公司 User preference house source calculation method and system
CN110599280A (en) * 2018-06-12 2019-12-20 阿里巴巴集团控股有限公司 Commodity information preference model training and predicting method and device and electronic equipment
CN109582751B (en) * 2018-11-29 2021-01-01 百度在线网络技术（北京）有限公司 Retrieval effect measuring method and server
CN110659409B (en) * 2018-12-29 2022-06-10 北京奇虎科技有限公司 Point of interest (POI) recommendation method and device
CN111382218B (en) * 2018-12-29 2023-09-26 北京嘀嘀无限科技发展有限公司 System and method for searching point of interest (POI)
CN110096525A (en) * 2019-04-29 2019-08-06 北京三快在线科技有限公司 Calibrate method, apparatus, equipment and the storage medium of interest point information
CN111859168A (en) * 2019-05-24 2020-10-30 北京嘀嘀无限科技发展有限公司 Method and system for determining interest points
CN110472163B (en) * 2019-08-22 2022-09-23 百度在线网络技术（北京）有限公司 Method, device, electronic equipment and medium for determining display of map search result
CN111061946B (en) * 2019-11-15 2023-06-30 汉海信息技术(上海)有限公司 Method, device, electronic equipment and storage medium for recommending scenerized content
CN113221025B (en) * 2020-01-21 2024-04-02 百度在线网络技术（北京）有限公司 Point-of-interest recall method, device, equipment and medium
CN113282690B (en) * 2020-02-19 2024-04-02 百度在线网络技术（北京）有限公司 Method, device, equipment and storage medium for ordering recall of interest points
CN111737382A (en) * 2020-05-15 2020-10-02 百度在线网络技术（北京）有限公司 Ranking method of geographic position points, method for training ranking model and corresponding device
CN112100523B (en) * 2020-09-15 2024-03-08 北京百度网讯科技有限公司 Method and device for ordering interest points based on environment information, electronic equipment and medium
CN112417318B (en) * 2020-10-29 2024-08-30 汉海信息技术(上海)有限公司 Method and device for determining states of interest points, electronic equipment and medium
CN112559879B (en) * 2020-12-24 2023-10-03 北京百度网讯科技有限公司 Interest model training method, interest point recommending method, device and equipment
CN114861032A (en) * 2021-02-04 2022-08-05 华为技术有限公司 Searching method and electronic equipment
CN113343046B (en) * 2021-05-20 2023-08-25 成都美尔贝科技股份有限公司 Intelligent search ordering system
CN113297456B (en) * 2021-05-20 2023-04-07 北京三快在线科技有限公司 Searching method, searching device, electronic equipment and storage medium
CN115515143B (en) * 2021-06-22 2024-07-02 中国移动通信集团重庆有限公司 Base station coverage scene determination method, device, equipment and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103914536A (en) * 2014-03-31 2014-07-09 北京百度网讯科技有限公司 Interest point recommending method and system for electronic maps
CN104102638A (en) * 2013-04-02 2014-10-15 北京百度网讯科技有限公司 Position information based push method, push system and push device
CN104504064A (en) * 2014-12-19 2015-04-08 百度在线网络技术（北京）有限公司 Information recommendation method and device
CN105045901A (en) * 2015-08-05 2015-11-11 百度在线网络技术（北京）有限公司 Search keyword push method and device
CN106202260A (en) * 2016-06-29 2016-12-07 百度在线网络技术（北京）有限公司 searching method, device and search engine

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104102638A (en) * 2013-04-02 2014-10-15 北京百度网讯科技有限公司 Position information based push method, push system and push device
CN103914536A (en) * 2014-03-31 2014-07-09 北京百度网讯科技有限公司 Interest point recommending method and system for electronic maps
CN104504064A (en) * 2014-12-19 2015-04-08 百度在线网络技术（北京）有限公司 Information recommendation method and device
CN105045901A (en) * 2015-08-05 2015-11-11 百度在线网络技术（北京）有限公司 Search keyword push method and device
CN106202260A (en) * 2016-06-29 2016-12-07 百度在线网络技术（北京）有限公司 searching method, device and search engine

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"[Android]百度地图开发（二）.定位城市位置和城市POI搜索";eastmount;《https://yq.aliyun.com/articles/26128》;20141224;第9-10页 *

Also Published As

Publication number Publication date
CN106919641A (en) 2017-07-04

Similar Documents

Publication Publication Date Title
CN106919641B (en) Interest point searching method and device and electronic equipment
Li et al. The where in the tweet
Bennett et al. Inferring and using location metadata to personalize web search
US11449785B2 (en) Method and system for artificial intelligence learning using messaging service and method and system for relaying answer using artificial intelligence
CN106372249B (en) A kind of clicking rate predictor method, device and electronic equipment
CN105532030B (en) For analyzing the devices, systems, and methods of the movement of target entity
KR101871747B1 (en) Similarity tendency based user-sightseeing recommendation system and method thereof
Zheng Location-based social networks: Users
US8458177B2 (en) Automated search
CN107424043A (en) A kind of Products Show method and device, electronic equipment
CN105493057B (en) The content selection controlled using accuracy
WO2017166647A1 (en) Navigation information pushing method and device, and apparatus
CN111061946A (en) Scenario content recommendation method and device, electronic equipment and storage medium
CN107222526B (en) Method, device and equipment for pushing promotion information and computer storage medium
CN108550068A (en) A kind of individual commodity recommendation method and system based on user behavior analysis
CN110597962B (en) Search result display method and device, medium and electronic equipment
CN107436893A (en) A kind of webpage recommending method and device
CN108596695B (en) Entity pushing method and system
JP2009537891A (en) Advertisement / information exposure method for improving hit ratio of context advertisement / information mapping and context advertisement / information recommendation service system using the same
CN111415195B (en) Advertisement recommendation method and device, terminal equipment and storage medium
WO2019153889A1 (en) Implementation of geographical location service
JP2013093015A (en) Information recommendation method, device, and program
KR20200003109A (en) Method and apparatus for setting sample weight, electronic device
CN110532469B (en) Information recommendation method, device, equipment and storage medium
JP2010009315A (en) Recommended store presentation system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant