CN109614556B - Access path prediction and information push method and device - Google Patents
Access path prediction and information push method and device Download PDFInfo
- Publication number
- CN109614556B CN109614556B CN201811284144.9A CN201811284144A CN109614556B CN 109614556 B CN109614556 B CN 109614556B CN 201811284144 A CN201811284144 A CN 201811284144A CN 109614556 B CN109614556 B CN 109614556B
- Authority
- CN
- China
- Prior art keywords
- interest
- user
- interest point
- point
- users
- 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
Links
Images
Abstract
The embodiment of the invention discloses an access path prediction method, an information push method and an information push device, wherein the access path prediction method comprises the following steps: acquiring the current position of a user in a target area and user characteristic parameters; predicting the interest point and/or the interest point type with the highest user access probability according to the current position, the user characteristic parameters and historical access data; and taking the interest point and/or the interest point type with the highest access probability as a starting interest point and/or an interest point type of a predicted access path.
Description
Technical Field
The invention relates to the technical field of mobile internet, in particular to an access path prediction method, an information push method and an information push device.
Background
Currently, mobile internet technology is rapidly developing, and the amount of smart devices such as smart phones is increasing, so that Location-based Services (LBS) is closely related to life, and more applications are mainly used for "around-the-body" Services. For example, a merchant can send personalized advertisement information to a mobile phone of a user by analyzing the position information of the user, so that the accuracy of advertisement putting is improved, the user can quickly receive discount information, coupons and the like of nearby merchants, and the potential consumption desire of the user is greatly stimulated.
However, while the value of LBS is recognized, the privacy problem brought by LBS is getting more and more concerned, and the development of LBS is seriously hindered.
Disclosure of Invention
In view of this, embodiments of the present invention provide an access path prediction method, an information push method, and an apparatus, so as to solve a problem that an LBS service in the prior art cannot effectively protect user privacy.
According to a first aspect, an embodiment of the present invention provides an access path prediction method, including: acquiring the current position of a user in a target area and user characteristic parameters; predicting the interest point and/or the interest point type with the highest user access probability according to the current position, the user characteristic parameters and historical access data; and taking the interest point and/or the interest point type with the highest access probability as a starting interest point and/or an interest point type of a predicted access path.
Optionally, the historical visiting data includes collected user characteristic parameters of a plurality of users and a historical visiting path in the target area, the historical visiting path includes at least one interest point and/or interest point type, and the obtaining of the interest point and/or interest point type with the highest user visiting probability according to the current location, the user characteristic parameters, and the historical visiting data includes: matching corresponding user groups for the users according to the user characteristic parameters, wherein the user groups are obtained according to the user characteristic parameters of the users; obtaining interest points and/or interest point types within a preset distance from the current position; and selecting the interest point and/or the interest point type with the highest user access probability in the matched user group from the obtained interest points and/or the interest point types, wherein the access probability of the interest point and/or the interest point type in each user group is obtained according to the historical access paths of the plurality of users.
Optionally, the step of obtaining the user group according to the user characteristic parameters of the plurality of users includes: calculating the similarity between the user characteristic parameters of each user; and dividing the plurality of users into a plurality of user groups according to the similarity.
Optionally, the step of obtaining the user group according to the user characteristic parameters of the plurality of users includes: and clustering the user characteristic parameters of all users by using an unsupervised learning mode to form a plurality of user groups.
Optionally, the matching, according to the user characteristic parameter, a corresponding user group for the user includes: respectively calculating the similarity between the user characteristic parameters of the users and the user characteristic parameters of the users in the user groups; and selecting the user group with the highest similarity to be matched as the user group corresponding to the user.
Optionally, the method further comprises: taking the interest point and/or the interest point type with the highest relevance degree with the starting interest point and/or the interest point type in the matched user group as a next interest point and/or an interest point type of the predicted access path; and/or using the interest point and/or the interest point type with the highest relevance degree with the currently predicted interest point and/or interest point type in the matched user group as the next interest point and/or interest point type of the predicted access path.
Optionally, when there are interest points and/or interest point types with the same access probability or when there are interest points and/or interest point types with the same association degree, selecting the interest points and/or interest point types which are more likely to be accessed by the user according to the current time.
Optionally, the historical visiting data includes collected user characteristic parameters of a plurality of users and a historical visiting path in the target area, the historical visiting path includes a starting point and at least one interest point and/or interest point type, and the obtaining of the interest point and/or interest point type with the highest user visiting probability according to the current position and the user characteristic parameters includes: and inputting the current position and the user characteristic parameters into a first neural network model to obtain the interest point and/or the interest point type with the highest user visit probability, wherein the first neural network model is obtained by training the historical visit data.
Optionally, the method further comprises: inputting the starting interest point and/or interest point type and the user characteristic parameter into a second neural network model to obtain a subsequent interest point and/or interest point type of the predicted visit path, wherein the second neural network model is obtained by training the historical visit data.
Optionally, the predicted access path includes n points of interest and/or types of points of interest, where n satisfies the following condition: t1+ t2+ … + tn is less than or equal to min { Tb-Ta, Tc }; wherein t1, t2 and … … tn are respectively the average stay time of the users in the matched user group at the n interest points and/or interest point types, Ta represents the current time, Tb represents the closing time of the stores in the target area, Tc represents the average total time consumption of the users in the matched user group on a historical access path, min { Tb-Ta, Tc } represents the smaller value of Tb-Ta and Tc, and n is a natural number.
Optionally, the method further comprises: and screening the historical access data.
Optionally, the filtering the historical access data includes: respectively acquiring the staying time of each user accessing each interest point and/or each interest point type in the historical access data; and removing the interest points and/or the interest point types with the stay time less than the preset time.
Optionally, the filtering the historical access data includes: respectively counting the access probability of each user for accessing each interest point and/or each interest point type; points of interest and/or types of points of interest having a visit probability below a predetermined threshold are removed.
According to a second aspect, an embodiment of the present invention provides an information pushing method, including: when receiving a current location of a user, obtaining a predicted access path of the user according to the method of any one of the above first aspects; and pushing corresponding information to the user according to the predicted access path.
According to a third aspect, an embodiment of the present invention provides an access path prediction apparatus, including: the acquisition unit is used for acquiring the current position of a user in the target area and the characteristic parameters of the user; the prediction unit is used for acquiring the interest point and/or the type of the interest point with the highest user access probability according to the current position, the user characteristic parameters and historical access data; and the path unit is used for taking the interest point and/or the interest point type with the highest access probability as the initial interest point and/or the interest point type of the predicted access path.
According to a fourth aspect, an embodiment of the present invention provides an information pushing apparatus, including: the access path prediction device according to the third aspect is configured to, when receiving a current location of a user, obtain a predicted access path of the user; and the pushing unit is used for pushing corresponding information to the user according to the predicted access path.
According to a fifth aspect, an embodiment of the present invention provides a server, including: a memory and a processor, the memory and the processor being communicatively coupled to each other, the memory having stored therein computer instructions, the processor performing the method of any of the first or second aspects by executing the computer instructions.
According to a fifth aspect, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing a computer to perform the method of any one of the first or second aspects.
Drawings
The features and advantages of the present invention will be more clearly understood by reference to the accompanying drawings, which are illustrative and not to be construed as limiting the invention in any way, and in which:
FIG. 1 shows a flow diagram illustrating an access path prediction method according to an embodiment of the invention;
FIG. 2 shows a flow diagram illustrating an access path prediction method according to another embodiment of the invention;
FIG. 3 is a diagram illustrating grouping of users according to similarity between user characteristic parameters according to an embodiment of the present invention;
FIG. 4 shows a flow diagram illustrating an access path prediction method according to another embodiment of the invention;
FIG. 5 shows a flow diagram of an information push method according to an embodiment of the invention;
FIG. 6 shows a schematic diagram of an access path prediction apparatus according to an embodiment of the invention;
FIG. 7 shows a schematic diagram of an information pushing apparatus according to an embodiment of the invention;
fig. 8 shows a schematic diagram of a server according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. 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 invention.
Fig. 1 illustrates an access path prediction method according to an embodiment of the present invention, which may predict a possible access path of a user according to historical access data of a plurality of users when the user enters a target area (e.g., a certain business district), without constantly obtaining real-time location information of the user, where the access path may include at least one Point of Interest (POI) and/or a type of the Point of Interest (POI), each of which may represent, for example, a specific store, a specific movie theater, a specific restaurant, etc., each of which has a type, one type of the POI is a set of stores having the type, for example, the type of the POI is a cosmetic, and the type of the POI includes a set of cosmetic stores within the business district. For some scenes, specific interest points do not need to be concerned, only interest point types interested by the user need to be given, the interest points in the business circles in the types are displayed to the user, the user is given certain free option, and the consumption desire of the user can be promoted. As shown in fig. 1, the method may include the steps of:
s11, obtaining the current position of the user in the target area and the user characteristic parameters.
After a user enters a certain target area (such as a certain business district), if the user does not have an explicit idea to access which point of interest next step, the user can open a corresponding APP in the intelligent mobile terminal, so that the APP can acquire the current position of the user through a positioning module of the intelligent mobile terminal and transmit the current position information to the server through the intelligent mobile terminal. The user characteristic parameter is a parameter related to the user itself, such as the user's age, gender, income, hobbies, and the like.
However, the present invention is not limited to this, if the user holds an intelligent mobile terminal that is not equipped with a positioning function, the user may also authenticate on a certain service terminal by means of card swiping, fingerprint identification, password confirmation, etc., and the server may obtain the current location of the user through the location information of the service terminal, and may also obtain the user characteristic parameters that the user fills in when registering.
S12, predicting the interest points and/or interest point types with the highest user access probability according to the current position, the user characteristic parameters and the historical access data.
The server stores a large amount of historical access data, and the historical access data records user characteristic parameters of a plurality of users and historical access paths in the target area, wherein the historical access paths are at least one point of interest and/or type of point of interest accessed by a plurality of users in the target area such as a certain business district and the like which are acquired previously. Taking the interest points as an example, if a user visits a certain cinema a, then visits a certain bookstore B, and finally visits a certain restaurant C in a business circle, the cinema a, the bookstore B, and the restaurant C constitute an access path of the user, the access path may be ordered or unordered, if the user tends to visit the three interest points in the order of cinema a, bookstore B, and restaurant C, but not in another order, the access path is ordered, and if the user does not have an obvious tendency to visit the sequence of cinema a, bookstore B, and restaurant C, the access path is unordered. Similar users generally have similar consumption tendencies, the more dimensionality of the characteristic parameters of the users, the more accurate the similar users are matched, and the most probable interest points and/or interest point types accessed by the users can be predicted by analyzing a large amount of historical access data.
And S13, taking the interest point and/or the interest point type with the highest access probability as the starting interest point and/or the interest point type of the predicted access path.
Generally, after entering a certain business district, a user does not access only one point of interest or one type of point of interest, and the user may access other points of interest or types of points of interest, so the point of interest and/or the type of point of interest with the highest access probability may be used as the starting point of interest and/or the type of point of interest of the predicted access path.
Through the steps of the method, the server can predict the interest points and/or interest point types which are possibly visited by the user only by knowing the initial position of the user, and does not need to acquire the real position of the user all the time, so that the method has certain commercial value while ensuring the privacy of the user.
Fig. 2 illustrates an access path prediction method according to another embodiment of the present invention, which may include the steps of:
and S21, acquiring the current position of the user in the target area and the user characteristic parameters, wherein the specific content can be referred to the description of the step S11.
And S22, matching the corresponding user group for the user according to the user characteristic parameters, wherein the user group is obtained according to the user characteristic parameters of a plurality of users in the historical access data.
In this embodiment, a plurality of methods may be employed to obtain the user group according to the user characteristic parameters of the plurality of users in the historical access data. In some optional implementations of this embodiment, the step of obtaining the user group according to the user characteristic parameters of the plurality of users in the historical access data may include
S22a) calculating the similarity among the user characteristic parameters of each user;
s22b) dividing the plurality of users into a plurality of user groups according to the similarity.
One index for measuring the similarity between the user characteristic parameters may be the spatial distance between the user characteristic parameters, taking a two-dimensional user characteristic parameter including age and income as an example, the ages and incomes of a plurality of users in the historical visit data may be represented in a coordinate system with income as abscissa and age as ordinate, as shown in fig. 3, each user sample may be represented by a circle, the computer may calculate the spatial distance between two points, so that points with the spatial distance within a predetermined range may be divided into a group, as shown by a dashed box in fig. 3, for example, a young person with low income, a young person with high income, an old person with high income, and the like, and for samples with the spatial distance between some and any point exceeding the predetermined range, as shown by a dot not within a dashed box in fig. 3, may be discarded. In fig. 3, the present embodiment is illustrated by taking only two-dimensional user characteristic parameters as an example, and in practical applications, in order to implement more precise user group division, there may be tens or even hundreds of dimensional user characteristic parameters. Thus, users can be grouped according to the similarity between the characteristic parameters of the users.
In other alternative embodiments of this embodiment, the step S22 is implemented by, for example, unsupervised learning, and the computer can automatically cluster the user samples to form a plurality of user groups. Unsupervised learning can classify untagged samples by data analysis of a large number of samples. Unsupervised learning can be implemented based on probability density function estimation, such as maximum likelihood or Bayes estimation, or based on inter-sample similarity, such as K-Means algorithm.
After obtaining a plurality of user groups, for example, the similarity between the user characteristic parameter of the user and the user characteristic parameters of the users in each user group may be calculated, and the user group with the highest similarity is selected to match with the user group corresponding to the user.
S23, obtaining the interest points and/or the interest point types within the preset distance from the current position.
When the server acquires the current position of the user, the server may diverge outward along each road of the current position, for example, measure and calculate a range that the user can reach by walking for 5 minutes according to the normal walking speed, acquire the interest points within the range, and further obtain corresponding interest point types according to the interest points. As another alternative, the server may also make a circle with the current position of the user as a center and a length of, for example, 500 meters as a radius, and obtain the interest point within the range.
And S24, selecting the interest point and/or the interest point type with the highest user access probability in the matched user group from the obtained interest points and/or the interest point types, wherein the access probability of the interest point and/or the interest point type in each user group is obtained according to historical access paths of a plurality of users.
Also taking the two-dimensional user characteristic parameters including age and income as an example, the server matches the user to a user group of young people with low income, it can be considered that the user group has similar consumption tendency to the young people with low income, and in order to take distance into account, the interest point with the highest user visit probability in the same user group can be selected from the interest points within the range that the user can quickly reach from the current position. Since the selected interest point is the interest point with the highest user access probability in the same user group and is within the range which can be quickly reached by the user from the current position, the user can be predicted to possibly access the interest point.
The following describes, as an example, the probability of accessing a point of interest in each user group according to historical access paths of a plurality of users. As shown in table 1 below, a user group includes three users, user a, user B and user C, and the table lists the visit rates of each user for visiting the points of interest 1-4, so that it can be counted that the visit rate of the point of interest 1 is 24 times, the visit rate of the point of interest 2 is 39 times, the visit rate of the point of interest 3 is 22 times, and the visit rate of the point of interest 4 is 15 times, and with respect to the total 100 visit rates, the visit probability of the point of interest 1 is 24%, the visit probability of the point of interest 2 is 39%, the visit probability of the point of interest 3 is 22%, and the visit probability of the point of interest 4 is 15%. By such statistics it is possible to know which point of interest the users within the user group are more interested in.
| Point of interest 1 | Point of interest 2 | Point of interest 3 | Point of interest 4 | |
| User A | 15 | 12 | 3 | |
| User B | 15 | 10 | 15 | |
| User C | 9 | 12 | 9 |
TABLE 1 example of user visits to various points of interest
The access probability of the interest point types in each user group is obtained according to the historical access paths of a plurality of users, the interest point types of the interest points can be extracted, the access probability of each interest point type accessed by the users in each user group is respectively counted, and then the user in the user group can be known to be more interested in which type of interest point.
And S25, taking the interest point and/or the interest point type with the highest access probability as a starting point of the predicted access path, specifically referring to the description of the step S13.
Through the steps of the method, the server analyzes a large amount of historical access data, divides the users in the historical access data into a plurality of user groups according to the characteristic parameters of the users, and matches the users into the corresponding user groups, so that the server can only acquire the initial positions of the users, predict the interest points and/or interest point types which the users may access according to the access probability of the users in the user groups matched with the users to different interest points and/or interest point types, and does not need to acquire the real positions of the users all the time, thereby ensuring the privacy of the users and having certain commercial value.
As some optional implementations of this embodiment, as shown in fig. 2, the access path prediction method may further include:
and S26, taking the interest point and/or the interest point type with the highest relevance degree with the starting point in the matched user group as a next interest point and/or a next interest point type of the predicted access path.
In the access path, a certain degree of association exists between the interest points, and the degree of association can be used to define the probability that the interest points simultaneously appear in the same access path. For example, by analyzing the historical access data, it can be found that if a user of a certain user group purchases a beverage after eating and watches a movie after purchasing the beverage, the association among the three points of interest of the restaurant, the beverage shop and the movie theater is relatively high. The degree of association between the interest points may be ordered or unordered. There is a possibility that the users of the user group mostly follow the order of eating first, then buying beverages, then watching movies, rather than eating first, or buying beverages first and then watching meals, in which case it is useful to calculate the relevance between the points of interest in the order in which they are visited, i.e. the relevance from restaurant to beverage store, the relevance from beverage store to movie theater being high, and vice versa the relevance from beverage store to restaurant, the relevance from movie theater to beverage store being low. For other scenarios, the association degree from the cosmetic store 1 to the cosmetic store 2 is not significantly different from the association degree from the cosmetic store 2 to the cosmetic store 1, and the association degree in such scenarios may be set to be unordered, which may be more beneficial for subsequent accurate prediction.
Likewise, in some cases it may not be necessary to focus on a particular point of interest, but rather on a type of point of interest, in which case a degree of association between the various types of points of interest may be calculated. Likewise, the association degree between the interest point types may be ordered or unordered.
For scenes in which the relevance between the interest points and/or the interest point types is ordered, selecting the interest point and/or the interest point type with the highest relevance of the starting point according to the ordered relevance; and for the scene that the association degrees between the interest points and/or the interest point types are unordered, selecting the interest point and/or the interest point type with the highest association degree of the starting point according to the unordered association degree.
And S27, predicting the interest point and/or the interest point type with the highest relevance degree with the currently predicted interest point and/or interest point type in the matched user group as a next interest point and/or a next interest point type of the predicted access path.
The next interest point and/or the next interest point type can be continuously predicted according to the current interest point and/or the current interest point type, so that a predicted access path comprising n interest points and/or interest point types is formed, wherein n is a natural number.
In order to save the computing resources of the server and because of the limited energy and time of the users, the server needs to predict the access paths of the users in time, and the predicted access paths of the users are unlikely to include too many points of interest and/or types of points of interest, that is, n is limited. In some alternative embodiments, n may be set to a fixed value, such as 5. In some alternative embodiments, n may be a value related to a user characteristic parameter, the user may match a different user group, and the value of n may be different, for example, for a user group of a young female, the value of n may be greater than that of a user group of an old male, and n may be set to be equal to or slightly greater than the number of interest points and/or types of interest points included in the historical access path of the user in the user group matched by the user on average. In other alternative embodiments, a more accurate determination of n may be made. For example, when a user watches movies, meals, and haircuts, the stay time of the corresponding interest point is generally longer than that of the clothing store, in order to determine the n value more accurately, the server obtains the average stay time of the user in the matched user group at the interest point and/or the interest point type when predicting one interest point and/or the interest point type, then judges whether the user exceeds the smaller value of the closing time of the store in the business circle after the access of the interest point and/or the interest point type is finished and the average total consumption time of the user in the matched user group on a historical access path, and stops continuing the prediction if the smaller value is exceeded. In this case, the value of n satisfies the following condition:
t1+t2+…+tn≤min{Tb-Ta,Tc}；
wherein t1, t2 and … … tn are respectively the average stay time of the users in the matched user group at n interest points and/or interest point types, Ta represents the current time, Tb represents the closing time of the stores in the target area, Tc represents the average total time consumption of the users in the matched user group on a historical access path, and min { Tb-Ta, Tc } represents the smaller value of Tb-Ta and Tc.
In some cases, there may be two or more points of interest and/or point of interest types that have the same access probability and are close in distance to the user's current location. In this case, as an alternative implementation, the interest points and/or interest point types with the same access probability may be predicted as one interest point and/or interest point type on the access path. As another alternative, the server may select points of interest and/or types of points of interest that the user is more likely to access based on the current time. For example, there are points of interest with the same access probability as movie theaters and restaurants, and the current time is 12:00, the server predicts that the user's access points before this do not include restaurants, and it should be possible to judge that the user should be more interested in restaurants at this time, so that it can be predicted that the user's next point of interest should be restaurants according to the current time. As another alternative, the server may take the current time into account when calculating the association between the respective points of interest and/or the respective types of points of interest accessed by the users in the respective user groups, that is, calculate the association between the respective points of interest and/or the respective types of points of interest accessed by the users in the respective user groups at different times, where the time interval may be flexibly set according to practical situations, for example, may be set to a half-hour or one-hour time interval, or may only set the association between the respective points of interest and/or the respective types of points of interest accessed by the users in the respective user groups at three different time intervals, namely morning, noon, and evening. In addition, the consumption tendencies of the users on weekdays, weekends and holidays can be different, and the relevance between each point of interest and/or each type of point of interest accessed by the users in each user group on different dates can be set, so that the predicted access path can be more accurate.
Through the steps of the method, the server predicts the interest points or interest point types which are possibly accessed by the user subsequently according to the relevance between the interest points and/or interest point types in the matched user group, does not need to acquire the real position of the user all the time, and has certain commercial value while ensuring the privacy of the user.
Fig. 4 illustrates an access path prediction method according to another embodiment of the present invention, which may include the steps of:
and S31, acquiring the current position of the user in the target area and the user characteristic parameters, which refer to the description of the step S11 specifically.
S32, inputting the current position and the user characteristic parameters into a first neural network model to obtain the interest point and/or the interest point type with the highest user access probability, wherein the first neural network model is obtained by training historical access data, the historical access data comprise the collected user characteristic parameters of a plurality of users and historical access paths in the target area, and the historical access paths comprise a starting point and at least one interest point and/or interest point type.
In the case of a large amount of historical visit data, the first neural network model may be trained in a machine learning manner to predict the interest points and/or interest point types with the highest visit probability for the user. The first neural network model may be, for example, a feed-forward neural network, which is trained by using a large amount of historical visit data to obtain the first neural network model, with user characteristic parameters and starting points of a plurality of users as inputs, and with visit probabilities of the initially visited interest points and/or interest point types as outputs. When the server receives the current position of the user and the characteristic parameters of the user, the current position and the characteristic parameters of the user can be input into the trained first neural network model, the first neural network can output the access probability of the interest points and/or the interest point types which are possibly accessed by the user, and therefore the interest points and/or the interest point types with the highest access probability can be obtained. The larger the data volume of the historical access data is, the more accurate the trained first neural network model prediction is.
And S33, taking the interest point and/or the interest point type with the highest access probability as the starting interest point and/or the interest point type of the predicted access path, specifically referring to the description of the step S13.
Through the steps of the method, under the condition that a large amount of historical access data exist, the server can predict the interest points and/or interest point types which the user may visit by using the trained first neural network model only by knowing the initial position of the user, and does not need to acquire the real position of the user all the time, so that the privacy of the user is guaranteed, and a certain commercial value is realized.
As some optional implementations of this embodiment, as shown in fig. 4, the access path prediction method may further include:
and S34, inputting the starting interest point and/or the interest point type and the user characteristic parameter into a second neural network model to obtain the subsequent interest point and/or the interest point type of the predicted access path, wherein the second neural network model is obtained by training historical access data.
In the case of a large amount of historical visit data, the second neural network model may be trained in a machine learning manner to predict subsequent points of interest and/or point of interest types for the predicted visit path. The second Neural Network model may be, for example, a Recurrent Neural Network (RNN) or a long short-Term Memory Network (LSTM), which is adapted to predict events in a time series. Inputting user characteristic parameters and initial interest points and/or interest point types of a plurality of users to output for subsequent interest points and/or interest point types of the predicted access path, and training the neural network by using a large amount of historical access data to obtain the second neural network model. When the server receives the initial interest points and/or interest point types and the user characteristic parameters, the initial interest points and/or interest point types and the user characteristic parameters can be input into a trained second neural network model, and the second neural network can output interest points and/or interest point type sequences which are possibly visited by the user subsequently, so that a predicted visit path can be obtained. Likewise, the larger the data size of the historical access data, the more accurate the trained second neural network model prediction can be.
Likewise, the number n of the points of interest and/or the types of the points of interest included in the predicted access path may be set according to the relevant method in step S27, so that the computing resources of the server may be saved.
Through the steps of the method, under the condition that a large amount of historical access data exist, the server predicts the interest points or interest point types which are possibly accessed by the user subsequently by using the trained second neural network model, the real position of the user does not need to be obtained all the time, and the method has certain commercial value while the privacy of the user is guaranteed.
In the above-described embodiment, the server may predict the interest points and/or interest point types that the user may access, according to the historical access data, by only knowing the initial location of the user. However, there may be many noisy data in the historical visit data, which may affect the accuracy of the prediction if the server also takes into account these noisy data when predicting. For this reason, in some optional implementations of this embodiment, the server also needs to filter the historical access data.
In an optional implementation manner, the step of filtering the historical access data may include:
a) respectively acquiring the staying time of each user accessing each interest point and/or each interest point type in the historical access data;
b) and removing the interest points and/or the interest point types with the stay time less than the preset time.
If each point of interest visited by the user is simply collected, a large amount of interference data may be brought, and subsequent access path prediction is not facilitated. The interest points can be screened according to the stay time of the user for visiting each interest point, when the stay time of the user for visiting a certain interest point and/or interest point type is short, probably because the user is not interested in the interest point and/or interest point type, the interest point and/or interest point type should be screened from the historical visit path of the user, for example, the time of the user visiting a certain cosmetic shop does not exceed 3 minutes, the user is judged to be not interested in the cosmetic shop to a great extent, and therefore the cosmetic shop can be deleted from the historical visit path of the user.
As another alternative, the step of filtering the historical access data may include:
c) respectively counting the access probability of each user accessing each interest point and/or each interest point type in the target area;
d) points of interest and/or types of points of interest having a visit probability below a predetermined threshold are filtered out.
Similarly, if historical access data of a certain user in a certain business district is collected, for example, it may be found that the user accesses the point of interest 1 10 times and accesses the point of interest 2 9 times, and only accesses the point of interest 31 time, it may be found that the access probability of the user for the point of interest 1 is 50%, the access probability of the point of interest 2 is 45%, and the access probability of the point of interest 3 is 5%, since the access probability of the user for the point of interest 3 is lower than a certain set value, such as 10%, it may be determined that the user is not interested in the point of interest 3, and thus the user may be deleted from the historical access path of the user, and the same is true for the type of the point of interest. By screening out the interest points and/or interest point types with the access probability lower than the preset threshold value, interference data can be removed to a certain extent, and subsequent access path prediction is facilitated.
Fig. 5 shows an information pushing method according to an embodiment of the present invention, which may include the following steps:
s41, when the current position of the user is received, the predicted access path of the user is obtained.
For example, the access path prediction method in the embodiment shown in fig. 1 to 4 may be adopted to predict at least one point of interest and/or a type of point of interest that the user may access only according to the obtained current point of interest of the user and the user characteristic parameter, and the specific content of which may be referred to in the embodiment shown in fig. 1 to 4.
And S42, pushing corresponding information to the user according to the predicted access path.
After predicting the interest points which the user may visit, the server may push information related to the interest points to the user, such as store addresses, new merchandise information, discount information, coupons, and the like. The information can be pushed in various ways, for example, the information can be used as a notification of an APP, can be in a form of a short message, and can also be directly displayed on an APP home page, and the APP home pages opened by different users are different and are customized individually for different users.
In some optional embodiments of this embodiment, the information of each point of interest included in the predicted access path may be sequentially pushed to the user, for example, in a push manner in which the information of the point of interest is directly displayed on a top page of the APP, the information of each point of interest of the predicted access path is sequentially displayed in the top page. However, in a notification or short message push manner, if the information of each point of interest is continuously and sequentially pushed to the user, the user may receive a lot of information at a time, the user often only notices the most recently pushed information, that is, the information of the last point of interest, and the point of interest may be far from the current location of the user, so that the user may ignore all pushed information, which may result in poor information delivery effect. As an optional implementation manner, the server only pushes the predicted first point of interest information to the user, and pushes the information of the second point of interest to the user after delaying for a period of time according to the average staying time of the user at the point of interest in the user group matched with the user, so as to continue to push subsequent point of interest information, so that the latest pushed information received by the user is the information of the point of interest that the user will access next, and the effect of information delivery is enhanced.
For the case that the predicted access path includes each interest point type, as an optional implementation manner, the information of the interest points of the interest point type in the target area may be pushed to the user, for example, it is predicted that the user is currently interested in watching movies, information of all movie theaters in the business circle, such as address information, ticket price information, etc., may be displayed on the APP top page, the user may select a movie theater to be visited according to the actual situation, for example, a user who pays attention to the price may select a movie theater with low price, a user who wants to watch the movie as soon as possible may select a movie theater closest to the user, a user who only wants to watch a certain movie may select a movie theater with the most suitable play time, compared with the manner of only pushing specific interest points, pushing multiple interest points of the same category enriches the user's selections, and also may play a good information delivery effect. As an alternative, when there are more points of interest of that type of point of interest in the target area, it may be difficult for the user to make a decision, such as predicting that the user is currently interested in dining, however there are often many restaurants in the business district, and even if the type of points of interest is more refined, such as to western meals, there may still be many points of interest of that type. At this time, the server may screen out a part of the interest points according to the consumption tendency of the user in the user group matched with the user, for example, the server may obtain the consumption price of the user in the user group matched with the user, so that the interest points which are not in the consumption price range may be screened out, thereby facilitating the user to make a decision.
Correspondingly, as shown in fig. 6, an embodiment of the present invention further provides an access path prediction apparatus, which may include:
an obtaining unit 51, configured to obtain a current location of a user in a target area and a user characteristic parameter;
the prediction unit 52 is configured to obtain an interest point and/or a type of the interest point with the highest user access probability according to the current location, the user characteristic parameter, and the historical access data;
and a path unit 53, configured to use the interest point and/or the interest point type with the highest access probability as a starting interest point and/or an interest point type of the predicted access path.
The details of the obtaining unit 51, the predicting unit 52 and the path unit 53 may be understood by referring to the corresponding related descriptions and effects in the embodiments shown in fig. 1 to fig. 4, and are not described herein again.
Correspondingly, as shown in fig. 7, an embodiment of the present invention further provides an information pushing apparatus, which may include:
an access path predicting device 61, configured to obtain a predicted access path of the user when receiving the current location of the user, where the access path predicting device 61 may be, for example, an access path predicting device in the embodiment shown in fig. 6;
and a pushing unit 62, configured to push corresponding information to the user according to the predicted access path.
The details of the access path predicting device 61 and the pushing unit 62 may be understood by referring to the corresponding related description and effects in the embodiment shown in fig. 5, and are not described herein again.
As shown in fig. 8, an embodiment of the present invention further provides a server, which may include a processor 71 and a memory 72, where the processor 71 and the memory 72 may be connected by a bus or by other means, and fig. 8 illustrates the connection by the bus as an example.
The processor 71 may be a Central Processing Unit (CPU). The Processor 71 may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or combinations thereof.
The memory 72, which is a non-transitory computer readable storage medium, may be used for storing non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions corresponding to the method for launching an application program in the embodiment of the present invention. The processor 71 executes various functional applications of the processor and data processing by executing non-transitory software instructions stored in the memory 72, namely, implements the access path prediction method or the information push method in the above-described method embodiments.
The memory 72 may include high speed random access memory and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 72 may optionally include memory located remotely from the processor 71, and such remote memory may be connected to the processor 71 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The details of the server may be understood by referring to the corresponding descriptions and effects in the embodiments shown in fig. 1 to fig. 5, and are not described herein again.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk (Hard Disk Drive, abbreviated as HDD) or a Solid State Drive (SSD), etc.; the storage medium may also comprise a combination of memories of the kind described above.
Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.
Claims (17)
1. An access path prediction method, comprising:
acquiring the current position of a user in a target area and user characteristic parameters;
analyzing a large amount of historical access data, dividing users in the historical access data into a plurality of user groups according to the characteristic parameters of the users, and matching the users into the corresponding user groups;
predicting the interest point and/or the interest point type with the highest user access probability according to the current position, the user characteristic parameters and historical access data;
taking the interest point and/or the interest point type with the highest access probability as a starting interest point and/or an interest point type of a predicted access path;
calculating the relevance between the interest points according to the sequence of the accessed interest points;
taking the interest point and/or the interest point type with the highest relevance degree with the starting interest point and/or the interest point type in the matched user group as a next interest point and/or an interest point type of the predicted access path; and/or
And taking the interest point and/or the interest point type with the highest relevance with the currently predicted interest point and/or interest point type in the matched user group as the next interest point and/or interest point type of the predicted access path.
2. The method of claim 1, wherein the historical visiting data comprises collected user characteristic parameters of a plurality of users and a historical visiting path in the target area, the historical visiting path comprises at least one interest point and/or interest point type, and the obtaining of the interest point and/or interest point type with the highest user visiting probability according to the current position, the user characteristic parameters and the historical visiting data comprises:
matching corresponding user groups for the users according to the user characteristic parameters, wherein the user groups are obtained according to the user characteristic parameters of the users;
obtaining interest points and/or interest point types within a preset distance from the current position;
and selecting the interest point and/or the interest point type with the highest user access probability in the matched user group from the obtained interest points and/or the interest point types, wherein the access probability of the interest point and/or the interest point type in each user group is obtained according to the historical access paths of the plurality of users.
3. The method of claim 2, wherein the step of deriving the user group according to the user characteristic parameters of the plurality of users comprises:
calculating the similarity between the user characteristic parameters of each user;
and dividing the plurality of users into a plurality of user groups according to the similarity.
4. The method of claim 2, wherein the step of deriving the user group according to the user characteristic parameters of the plurality of users comprises:
and clustering the user characteristic parameters of all users by using an unsupervised learning mode to form a plurality of user groups.
5. The method according to claim 2, wherein said matching the user with the corresponding user group according to the user characteristic parameter comprises:
respectively calculating the similarity between the user characteristic parameters of the users and the user characteristic parameters of the users in the user groups;
and selecting the user group with the highest similarity to be matched as the user group corresponding to the user.
6. The method according to any of claims 2-5, wherein when there are points of interest and/or types of points of interest with the same access probability or when there are points of interest and/or types of points of interest with the same relevance,
and selecting the interest points and/or interest point types which are more likely to be accessed by the user according to the current time.
7. The method according to claim 1, wherein the historical visiting data includes collected user characteristic parameters of a plurality of users and a historical visiting path in the target area, the historical visiting path includes a starting point and at least one interest point and/or interest point type, and the obtaining of the interest point and/or interest point type with the highest user visiting probability according to the current position and the user characteristic parameters includes:
and inputting the current position and the user characteristic parameters into a first neural network model to obtain the interest point and/or the interest point type with the highest user visit probability, wherein the first neural network model is obtained by training the historical visit data.
8. The method of claim 7, further comprising:
inputting the starting interest point and/or interest point type and the user characteristic parameter into a second neural network model to obtain a subsequent interest point and/or interest point type of the predicted visit path, wherein the second neural network model is obtained by training the historical visit data.
9. The method according to claim 1, wherein the predicted access path comprises n points of interest and/or point of interest types, n satisfying the following condition
t1+t2+…+tn≤min{Tb-Ta,Tc}；
Wherein t1, t2 and … … tn are respectively the average stay time of the users in the matched user group at the n interest points and/or interest point types, Ta represents the current time, Tb represents the closing time of the stores in the target area, Tc represents the average total time consumption of the users in the matched user group on a historical access path, min { Tb-Ta, Tc } represents the smaller value of Tb-Ta and Tc, and n is a natural number.
10. The method of claim 1, further comprising:
and screening the historical access data.
11. The method of claim 10, wherein the filtering the historical access data comprises:
respectively acquiring the staying time of each user accessing each interest point and/or each interest point type in the historical access data;
and removing the interest points and/or the interest point types with the stay time less than the preset time.
12. The method of claim 10, wherein the filtering the historical access data comprises:
respectively counting the access probability of each user for accessing each interest point and/or each interest point type;
points of interest and/or types of points of interest having a visit probability below a predetermined threshold are removed.
13. An information pushing method, comprising:
upon receiving a current location of a user, obtaining a predicted access path of the user according to the method of any one of claims 1-12;
and pushing corresponding information to the user according to the predicted access path.
14. An access path prediction apparatus, comprising:
the acquisition unit is used for acquiring the current position of a user in the target area and the characteristic parameters of the user;
the grouping matching unit is used for analyzing a large amount of historical access data, dividing users in the historical access data into a plurality of user groups according to the user characteristic parameters, and matching the users into corresponding user groups;
the prediction unit is used for acquiring the interest point and/or the type of the interest point with the highest user access probability according to the current position, the user characteristic parameters and historical access data;
the path unit is used for taking the interest point and/or the interest point type with the highest access probability as a starting interest point and/or an interest point type of a predicted access path;
the relevance calculating unit is used for calculating the relevance among the interest points according to the sequence of the accessed interest points;
the path unit is further configured to use the point of interest and/or the point of interest type with the highest degree of association with the starting point of interest and/or the point of interest type in the matched user group as a next point of interest and/or the point of interest type of the predicted access path; and/or
The path unit is further configured to use the point of interest and/or the point of interest type with the highest degree of association with the currently predicted point of interest and/or point of interest type in the matched user group as a next point of interest and/or point of interest type of the predicted access path.
15. An information pushing apparatus, comprising:
the access path prediction device of claim 14, configured to, upon receiving a current location of a user, obtain a predicted access path of the user;
and the pushing unit is used for pushing corresponding information to the user according to the predicted access path.
16. A server, comprising: a memory and a processor communicatively coupled to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the method of any of claims 1-13.
17. A computer-readable storage medium having stored thereon computer instructions for causing a computer to perform the method of any one of claims 1-13.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811284144.9A CN109614556B (en) | 2018-10-31 | 2018-10-31 | Access path prediction and information push method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811284144.9A CN109614556B (en) | 2018-10-31 | 2018-10-31 | Access path prediction and information push method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109614556A CN109614556A (en) | 2019-04-12 |
| CN109614556B true CN109614556B (en) | 2020-03-13 |
Family
ID=66002097
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811284144.9A Active CN109614556B (en) | 2018-10-31 | 2018-10-31 | Access path prediction and information push method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109614556B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112231594B (en) * | 2019-06-27 | 2024-04-12 | 北京百度网讯科技有限公司 | Information processing method and device |
| CN110377825A (en) * | 2019-07-15 | 2019-10-25 | 腾讯科技（深圳）有限公司 | Content providing, device, equipment and storage medium |
| CN110781415B (en) * | 2019-09-16 | 2022-09-30 | 腾讯科技（深圳）有限公司 | Interest point recommendation method, device, equipment and medium |
| CN110782284A (en) * | 2019-10-24 | 2020-02-11 | 腾讯科技（深圳）有限公司 | Information pushing method and device and readable storage medium |
| CN112883725A (en) * | 2020-12-29 | 2021-06-01 | 上海讯飞瑞元信息技术有限公司 | File generation method and device, electronic equipment and storage medium |
| CN112667726A (en) * | 2021-01-04 | 2021-04-16 | 拉卡拉支付股份有限公司 | Data extraction method, data extraction device, electronic equipment, storage medium and program product |
| CN112902964B (en) * | 2021-01-22 | 2024-02-23 | 南京邮电大学 | Path recommendation method and device based on interest points |
| CN112989207B (en) * | 2021-04-27 | 2021-08-27 | 武汉卓尔数字传媒科技有限公司 | Information recommendation method and device, electronic equipment and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107092629A (en) * | 2017-01-18 | 2017-08-25 | 北京小度信息科技有限公司 | Recommend method and device |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI252985B (en) * | 2004-12-15 | 2006-04-11 | Inst Information Industry | Guide route generation method and system |
| CN107491474A (en) * | 2017-06-26 | 2017-12-19 | 北京三快在线科技有限公司 | Information recommendation method and device |
| CN107798557B (en) * | 2017-09-30 | 2020-09-25 | 平安科技（深圳）有限公司 | Electronic device, service place recommendation method based on LBS data and storage medium |
-
2018
- 2018-10-31 CN CN201811284144.9A patent/CN109614556B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107092629A (en) * | 2017-01-18 | 2017-08-25 | 北京小度信息科技有限公司 | Recommend method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109614556A (en) | 2019-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109614556B (en) | Access path prediction and information push method and device | |
| US9183497B2 (en) | Performance-efficient system for predicting user activities based on time-related features | |
| CN106570722B (en) | Intelligent recommendation system and intelligent recommendation method | |
| Cui et al. | Personalized travel route recommendation using collaborative filtering based on GPS trajectories | |
| Zhao et al. | Discovering different kinds of smartphone users through their application usage behaviors | |
| US9436695B2 (en) | Dynamically expiring crowd-sourced content | |
| US9330174B1 (en) | Determining topics of interest | |
| US9348898B2 (en) | Recommendation system with dual collaborative filter usage matrix | |
| US20170364933A1 (en) | User maintenance system and method | |
| WO2016206196A1 (en) | Method and device for obtaining user attribute information, and server | |
| CN107291888B (en) | Machine learning statistical model-based living recommendation system method near living hotel | |
| US20150112963A1 (en) | Time and location based information search and discovery | |
| US9208444B1 (en) | Determining happiness indices | |
| WO2019056661A1 (en) | Search term pushing method and device, and terminal | |
| US20100082427A1 (en) | System and Method for Context Enhanced Ad Creation | |
| KR20180006875A (en) | Methods and systems for providing information for on-demand services | |
| JP6937776B2 (en) | Methods and devices for identifying geographic location attribute information | |
| WO2016015444A1 (en) | Target user determination method, device and network server | |
| CN108416616A (en) | The sort method and device of complaints and denunciation classification | |
| Marakkalage et al. | Understanding the lifestyle of older population: Mobile crowdsensing approach | |
| KR20140071417A (en) | Mobile advertising using data networks based on intelligence data associated with internet-connectable devices derived using graph models | |
| CN109903086B (en) | Similar crowd expansion method and device and electronic equipment | |
| US20150248436A1 (en) | Methods, Circuits, Devices, Systems and Associated Computer Executable Code for Assessing a Presence Likelihood of a Subject at One or More Venues | |
| CN113454669A (en) | Characterizing a place by user visited features | |
| CN104867015B (en) | A kind of article delivery person recommendation method based on user's moving projection |
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 |