CN107977445B

CN107977445B - Application program recommendation method and device

Info

Publication number: CN107977445B
Application number: CN201711307219.6A
Authority: CN
Inventors: 蔡馥励; 王长路; 庞国盛; 李涛
Original assignee: Qilin Hesheng Network Technology Inc
Current assignee: Qilin Hesheng Network Technology Inc
Priority date: 2017-12-11
Filing date: 2017-12-11
Publication date: 2022-05-03
Anticipated expiration: 2037-12-11
Also published as: CN107977445A

Abstract

The embodiment of the application provides an application program recommendation method and device, wherein the method comprises the following steps: dividing a plurality of users into a plurality of user groups according to the related information of each user; when an application search request of a target user is received, judging whether the number of users of a user group where the target user is located is larger than a preset number; if so, determining the industry type of the searched application corresponding to the application search request of the target user; if the industry category is life application, recommending an application program to the target user according to a first application program similarity list corresponding to the current country of the target user, and if the searched application is non-life application, recommending the application program to the target user according to a second application program similarity list corresponding to the user group of the target user. Through the method and the device, the recommendation accuracy of the application program can be improved.

Description

Application program recommendation method and device

Technical Field

The present application relates to the field of application recommendation, and in particular, to an application recommendation method and apparatus.

Background

With the popularization of intelligent terminals and the development of mobile terminal technologies, more and more users choose to download required application software in an application store of a mobile terminal. Statistically, the number of applications in the google play application store exceeds three million for the android system, and it takes much effort for a user to find a desired application among a large number of applications, and therefore, a technology for accurately recommending applications to the user arises.

At present, the application recommendation method for a user mainly includes the steps of counting and sorting the download amount of each application in an application store, generating a total download amount list and download amount list of each type of application, and recommending an application program for the user according to the total download amount list and the download amount list of each type of application.

However, in the above method, the applications are recommended based on the download quantity list, the recommended applications are the same for each user, and personalized differences among different users are not considered, so that the recommendation accuracy of the applications is poor.

Disclosure of Invention

The embodiment of the application aims to provide an application program recommendation method and device so as to improve recommendation accuracy of an application program.

In order to solve the above technical problem, the embodiment of the present application is implemented as follows:

in a first aspect, an embodiment of the present application provides an application program recommendation method, including:

dividing the users of the whole network into a plurality of user groups according to the relevant information corresponding to each user; wherein the related information comprises at least one of religious beliefs, the city, the country, the language used, the stationary country, the stationary city, the income level of the city, and the income level of the stationary city;

when an application search request of a target user is received, judging whether the number of users of a user group where the target user is located is larger than a preset number or not;

if the number of the searched applications is larger than the preset number, determining the industry type of the searched applications corresponding to the application search request of the target user;

if the industry category is life application, recommending an application program to the target user according to a first application program similarity list corresponding to the current country of the target user, and if the industry category is non-life application, recommending an application program to the target user according to a second application program similarity list corresponding to the user group of the target user.

In a second aspect, an embodiment of the present application provides an application recommendation apparatus, including:

the user clustering module is used for dividing the users in the whole network into a plurality of user groups according to the related information corresponding to each user; wherein the related information comprises at least one of religious beliefs, the city, the country, the language used, the stationary country, the stationary city, the income level of the city, and the income level of the stationary city;

the quantity judging module is used for judging whether the quantity of the users of the user group where the target user is located is larger than the preset quantity or not when receiving the application search request of the target user;

the category judging module is used for determining the industry category of the searched application corresponding to the application search request of the target user if the number of the searched applications is larger than the preset number;

and the application recommending module is used for recommending the application program to the target user according to the first application program similarity list corresponding to the current country of the target user if the industry category is life application, and recommending the application program to the target user according to the second application program similarity list corresponding to the user group of the target user if the industry category is non-life application.

In a third aspect, an embodiment of the present application provides an application recommendation device, including: a memory, a processor and a computer program stored on the memory and executable on the processor, the computer program, when executed by the processor, implementing the steps of the application recommendation method as described in the first aspect above.

In a fourth aspect, the present application provides a storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the steps of the application recommendation method according to the first aspect.

By the method and the device, the plurality of users can be divided into the plurality of user groups according to various related information of the users, the application programs are recommended to the users according to the first application program similarity list corresponding to the country where the users are located when the users search the life application programs, and the application programs are recommended to the users according to the second application program similarity list corresponding to the user group where the users are located when the users search the non-life application programs and the number of the users is larger than the preset number, so that the application programs are recommended to the users individually based on the search request of the users, the user group where the users are located and the country where the users are located, and the recommendation accuracy of the application programs is improved. And when the user searches for the life application, the application program is recommended to the user based on the country where the user is located, so that the recommended application program can conform to the living habits of the country where the user is located, and the application search experience of the user is improved. And when the user searches the non-living application, the application program is recommended to the user based on the user group where the user is located, so that the recommended application program can be matched with the user group where the user is located, namely matched with the classification condition of the user, and the application searching experience of the user is improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort.

Fig. 1 is a schematic flowchart of an application recommendation method according to an embodiment of the present application;

fig. 2 is a flowchart illustrating an application recommendation method according to another embodiment of the present application;

fig. 3 is a schematic diagram illustrating a module composition of an application recommendation apparatus according to an embodiment of the present application;

FIG. 4 is a block diagram of an application recommendation apparatus according to another embodiment of the present application;

fig. 5 is a schematic structural diagram of an application recommendation device according to an embodiment of the present application.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the present application, 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 only a part of the embodiments of the present application, and not all of the embodiments. 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 embodiment of the application provides an application program recommending method, an application program recommending device, application program recommending equipment and a storage medium, wherein the application program recommending method can be applied to a mobile terminal side and executed by the mobile terminal, can also be applied to a server side and executed by the server, and the mobile terminal can be a smart phone, a computer, a tablet computer and other equipment.

Fig. 1 is a schematic flowchart of an application recommendation method according to an embodiment of the present application, and as shown in fig. 1, the method includes:

step S102, dividing users of the whole network into a plurality of user groups according to relevant information corresponding to each user, wherein the relevant information comprises at least one item of religious beliefs, the city, the country, the language, the stationary country, the stationary city, the income level of the city and the income level of the stationary city; the whole network users refer to all users in the internet;

step S104, when receiving the application search request of the target user, judging whether the number of users of the user group where the target user is located is larger than the preset number;

step S106, if the quantity is larger than the preset quantity, determining the industry type of the searched application corresponding to the application search request of the target user;

and S108, if the industry category is life application, recommending an application program to the target user according to the first application program similarity list corresponding to the current country of the target user, and if the industry category is non-life application, recommending the application program to the target user according to the second application program similarity list corresponding to the user group of the target user.

According to the method and the device, the plurality of users can be divided into the plurality of user groups according to various related information of the users, the application programs are recommended to the users according to the first application program similarity list corresponding to the country where the users are located when the users search the life application programs, and the application programs are recommended to the users according to the second application program similarity list corresponding to the user group where the users are located when the users search the non-life application programs and the number of the users is larger than the preset number, so that the application programs are recommended to the users individually based on the search request of the users, the user group where the users are located and the country where the users are located, and the recommendation accuracy of the application programs is improved. And when the user searches for the life application, the application program is recommended to the user based on the country where the user is located, so that the recommended application program can conform to the living habits of the country where the user is located, and the application search experience of the user is improved. And when the user searches the non-living application, the application program is recommended to the user based on the user group where the user is located, so that the recommended application program can be matched with the user group where the user is located, namely matched with the classification condition of the user, and the application searching experience of the user is improved.

In this embodiment, it is assumed that the religious belief, the city, the country, the language used, the stationary country, the stationary city, the income level of the city, and the income level of the stationary city may represent a local characteristic of the user, and therefore, the related information includes at least one of the above information, where the local characteristic of the user is, for example, a first-line city living in china and a second-line city living in the united states at present.

In the step S102, the users in the whole network are divided into a plurality of user groups according to the related information corresponding to each user, specifically, the users in the whole network are clustered according to the related information corresponding to each user according to a preset clustering algorithm, so as to obtain a plurality of user groups.

Specifically, the preset clustering algorithm may be a k-means clustering algorithm, and according to the k-means clustering algorithm, the users in the whole network are clustered according to the related information corresponding to each user, so as to obtain a plurality of user groups. The clustering parameter k in the k-means clustering algorithm can be determined according to a manual debugging mode. Based on the principle of a clustering algorithm, after user groups are divided by the clustering algorithm, users in the same user group have larger region similarity, for example, users in the same group are all fixed in a same-line city, and users in different user groups have smaller region similarity.

In this embodiment, the users in the whole network are clustered according to the related information corresponding to each user, so as to obtain a plurality of user groups, and users with greater geographical similarity can be clustered into the same user group, and users with smaller geographical similarity can be clustered into different user groups. Because the user behaviors between users with larger regional similarity are generally similar, and the user behaviors between users with smaller regional similarity are generally smaller, for example, the group buying programs generally installed by the users in the first-line city and the second-line city are American groups, and the group buying programs generally installed by the users in the third-line city are more in number, the user groups are divided for the users, and a basis can be provided for realizing accurate recommendation of the application programs.

When the application recommendation method in this embodiment is executed by the mobile terminal, in step S104, the mobile terminal obtains the application search request of the target user, and determines whether the number of users in the user group where the target user is located is greater than the predetermined number. When the application program recommendation method in this embodiment is executed by the server, in step S104, the mobile terminal obtains an application search request of the target user and sends the application search request to the server, and the server determines whether the number of users in the user group where the target user is located is greater than the predetermined number when obtaining the application search request.

In the step S106, if the number of users in the user group where the target user is located is greater than the predetermined number, the industry category of the searched application corresponding to the application search request of the target user is determined.

In one embodiment, if the application search request includes a name of an application, the application corresponding to the name of the application is determined as the searched application, and if the application search request includes a keyword of the application, the mobile terminal or the server determines the application including the keyword in the name, and uses the application with the largest installation amount in the determined applications as the searched application.

After the searched application corresponding to the application search request is determined, the industry type of the searched application can be determined according to the description information of the searched application, wherein the industry type of the application program comprises life type, game type, video and audio type and the like. In a specific embodiment, the description information of the searched application records the industry category of the searched application.

In this embodiment, a first application similarity list corresponding to a current country where a target user is located is predetermined, where the first application similarity list is used to record a plurality of established applications and similar applications corresponding to each established application, and the similar applications corresponding to each established application recorded in the first application similarity list are all applications installed by the user in the current country. In this embodiment, the user in each country refers to a user who has long lived in the country.

Correspondingly, in step S108, recommending an application program to the target user according to the first application program similarity list corresponding to the current country of the target user, specifically:

(a1) searching at least one first similar application corresponding to the searched application in the first application similar list, and determining at least one second similar application corresponding to the searched application according to an application recommendation strategy based on a collaborative filtering algorithm;

(a2) the method comprises the steps of sequencing at least one first similar application and at least one second similar application through a pre-trained sequencing model, recommending the sequenced at least one first similar application and at least one second similar application to a target user, wherein the sequencing model is obtained through training based on installation records of the target user on historically recommended application programs.

In the action (a1), the similar application corresponding to the searched application is searched in the first application similarity list, and the searched similar application is used as the first similar application, and since the similar application of each given application recorded in the first application similarity list is the application installed by the user in the current country, the first similar application is also the application installed by the user in the current country.

In the above action (a1), at least one second similar application corresponding to the searched application is determined according to an application recommendation policy based on a collaborative filtering algorithm, which may be an SVD (Singular value decomposition) based collaborative filtering algorithm. In the SVD-based collaborative filtering algorithm, the ID (identity) of each user is used as a row of a two-dimensional matrix, and the ID of each application program is used as a column of the two-dimensional matrix to form a super-large two-dimensional matrix. If a certain user installs a certain application program, the value corresponding to the row of the user and the column of the application program is marked as 1, otherwise, the value is 0. And performing SVD on the two-dimensional matrix to obtain two vectors which are respectively a user vector and an application program vector, and determining similar users and similar applications through the user vector and the application program vector so as to determine at least one second similar application corresponding to the searched application. In another embodiment, a User-based collaborative filtering algorithm and an Item-based collaborative filtering algorithm may be combined instead of the SVD-based collaborative filtering algorithm.

Since the searched application is a life-class application, the first similar application and the second similar application are also life-class applications.

In the step (a2), the information related to the at least one first similar application and the information related to the at least one second similar application are input to a pre-trained ranking model, an application ranking result output by the model is obtained, the at least one first similar application and the at least one second similar application are ranked according to the application ranking result, and the ranked at least one first similar application and the ranked at least one second similar application are recommended to the target user.

The ranking model is obtained by training based on the installation records of the target user on the historically recommended application programs. Specifically, an installation record of a target user for a historically recommended application program is obtained, the target user is determined to select the installed application program and the uninstalled application program from the historically recommended application programs according to the installation record, relevant information of the selectively installed application program is used as a positive sample, relevant information of the uninstalled application program is used as a negative sample, the positive sample and the negative sample are input into a logistic regression model for training, and the trained model is used as the ranking model. The related information of the application program includes description information of the application program, and may further include user evaluation, user rating, and the like.

In this embodiment, the similar applications of each given application recorded in the first application program similar list are all applications installed by the user in the current country, so that the first similar applications can be the applications installed by the user in the current country, and therefore, the similar applications installed by the user in the country are recommended to the user under the condition that the user searches for the life application program, so that the user installs the life application program conforming to the living environment of the user, and the accuracy of recommending the application program is improved.

In the embodiment, the first application program similarity list and the application program recommendation strategy based on the collaborative filtering algorithm are combined to recommend the application program to the user, and the method has the advantage of improving recommendation accuracy.

Because the sequencing model is obtained by training based on the installation record of the target user on the historically recommended application program, after the first similar application and the second similar application are sequenced through the sequencing model, the application program which is sequenced more forwards can be more consistent with the behavior characteristics of the historically installed application program of the target user, so that the application program which is consistent with the installation habit of the application program of the user is preferentially recommended to the user, and the application program searching experience of the user is improved.

In this embodiment, a second application program similarity list corresponding to a user group where a target user is located is predetermined, where the second application program similarity list is used to record a plurality of established applications and similar applications corresponding to each established application, and the similar applications corresponding to each established application recorded in the second application program list are all applications installed in the user group where the target user is located, that is, applications installed by users in the user group where the target user is located.

Correspondingly, in step S108, recommending an application program to the target user according to the second application program similarity list corresponding to the user group where the target user is located, specifically:

(b1) at least one third similar application corresponding to the searched application is searched in the second application program similar list, and at least one fourth similar application corresponding to the searched application is determined according to an application program recommendation strategy based on a collaborative filtering algorithm;

(b2) and sequencing at least one third similar application and at least one fourth similar application through a pre-trained sequencing model, and recommending the sequenced at least one third similar application and at least one fourth similar application to the target user, wherein the sequencing model is obtained by training based on the installation record of the target user on the historically recommended application program.

In the action (b1), the similar application corresponding to the searched application is searched in the second application program similar list, and the searched similar application is used as the third similar application, where the similar application of each given application recorded in the second application program similar list is an application installed by the user in the user group, and therefore the third similar application is also an application installed by the user in the user group.

In the action (b1), at least one fourth similar application corresponding to the searched application is determined according to the application recommendation policy based on the collaborative filtering algorithm. The process can refer to the description of (a1) above, and is not repeated here.

Since the searched application is a non-life application, the third similar application and the fourth similar application are also non-life applications.

In the step (b2), the related information of the at least one third similar application and the related information of the at least one fourth similar application are input to a pre-trained ranking model, an application ranking result output by the model is obtained, the at least one third similar application and the at least one fourth similar application are ranked according to the application ranking result, and the ranked at least one third similar application and the ranked at least one fourth similar application are recommended to the target user. Reference may be made to the foregoing description for the ranking model, which is not repeated here.

In this embodiment, the similar applications of each given application recorded in the second application program similar list are all applications installed by the user in the user group, and the third similar application can be the application installed by the user in the user group, so that the application program conforming to the user classification condition of the user is recommended to the user when the user searches for the non-living application program, so that the user installs the application program conforming to the user classification condition of the user, and the accuracy of recommending the application program is improved.

In the embodiment, the application program is recommended to the user by combining the second application program similarity list and the application program recommendation strategy based on the collaborative filtering algorithm, and the beneficial effect of improving the recommendation accuracy is achieved.

Because the sequencing model is obtained by training based on the installation record of the target user on the historically recommended application program, after the third similar application and the fourth similar application are sequenced through the sequencing model, the application program which is sequenced more forwards can be more consistent with the behavior characteristics of the historically installed application program of the target user, so that the application program which is consistent with the installation habit of the application program of the user is preferentially recommended to the user, and the application program searching experience of the user is improved.

Since the living application program (such as a food takeout program) needs to be matched with the current living environment so as to be convenient for the user to use the living application program, and the non-living application program (such as a reading program) does not need to be matched with the current living environment and needs to be matched with the behavior habit of the user to improve the use experience of the user, in the embodiment, the searching living application program and the non-living application program are treated separately, and the application program is recommended according to the situation. When the life application is searched, program recommendation is carried out based on a first application program similarity list corresponding to the current country of the target user, so that the life application program matched with the living environment of the user can be installed by the user, and the recommendation accuracy of the application program is improved; when searching for non-living applications, program recommendation is performed based on the second application program similarity list corresponding to the user group where the target user is located, so that the user can install the application program matched with the user behavior of the user, and the accuracy of application program recommendation is improved.

In this embodiment, before step S108, it is further required to determine a first application similarity list corresponding to a country where the target user is currently located, where the specific determination process is as follows:

(c1) acquiring a predetermined application program similarity list; each application in the whole network and the similar application corresponding to each application are recorded in the application program similarity list; wherein, each application in the whole network refers to all applications in the Internet;

(c2) deleting the application programs which are not installed by the user in the country where the application programs are located in the similar application corresponding to each application recorded in the similar application program list;

(c3) and taking the deleted application program similarity list as a first application program similarity list corresponding to the current country.

Specifically, a predetermined application similarity list is obtained, optionally, the application similarity list includes all the applications existing in the entire network so as to ensure comprehensiveness of the applications, the application similarity list further records similar applications corresponding to each application existing in the entire network, and partial contents of the application similarity list may be as shown in table 1 below.

TABLE 1

Application program	Similar applications
		WeChat	Microblog, facebook
Love art	Youke, youtube

In act (c2), among the similar applications corresponding to each application recorded in the application similarity list, the applications that are not installed by the user in the country where the user currently exists are deleted. Specifically, the application installation list of each country may be counted in advance, where the application installation list is used to record popular application programs installed by users in the corresponding country, the application installation list of the current country is compared with the application similarity list, and in the similar application corresponding to each application recorded in the application similarity list, the application not recorded in the application installation list of the current country is deleted, so that the application not installed by the user in the current country is deleted.

The application installation list of each country can be obtained by adopting the following method: for a certain country, counting the applications installed by each user in the country, determining the applications installed by the users in the country, determining n applications with the highest installation amount, wherein n can be 5 thousands or other numerical values, and sorting the n applications with the highest installation amount in a descending order according to the installation amount to obtain an application installation list of the country.

In the action (c3), the deleted application similarity list is used as the first application similarity list corresponding to the current country. Since the applications that are not installed by the user in the current country are deleted through the action (c2), the similar applications recorded in the first application similarity list are all the applications installed by the user in the current country.

In this embodiment, by obtaining the first application similarity list corresponding to the current country, the application installed by the user in the country can be recommended to the user conveniently, so that a situation that the application recommended to the user does not conform to the living environment of the user is avoided. For example, if the country where a certain user is currently located is the united states, facebook is recommended to the user when the user searches for the WeChat, and if the country where the user is currently located is the china, microblog is recommended to the user when the user searches for the WeChat.

In this embodiment, the application installation list of each country may be counted in advance, and the first application similarity list corresponding to each country is determined according to the application installation list of each country and the application similarity list, so that the application is conveniently and quickly recommended to the user.

Accordingly, in this embodiment, before step S108, it is further required to determine a second application similarity list corresponding to the user group where the target user is located, where the specific determination process is as follows:

(d1) acquiring a predetermined application program similarity list; each application in the whole network and the corresponding similar application of each application are recorded in the application program similarity list; the application similarity list is the application similarity list in (c1) above;

(d2) deleting the application programs which are not installed by the user in the user group in the similar applications corresponding to each application recorded in the application program similar list;

(d3) and taking the deleted application program similarity list as a second application program similarity list corresponding to the user group.

Specifically, as shown in the above (c1), a predetermined application similarity list is obtained, optionally, the application similarity list includes all the applications currently existing in the whole network to ensure the comprehensiveness of the applications, the application similarity list further records similar applications corresponding to each application existing in the whole network, and partial contents of the application similarity list may be as shown in the above table 1.

In act (d2), the applications not installed by the user in the user group are deleted from the similar applications corresponding to each application recorded in the application similarity list. Specifically, the application installation list of each user group may be counted in advance, where the application installation list is used to record popular application programs installed by users in the corresponding user group, the application installation list of the user group is compared with the application similarity list, and in the similar application corresponding to each application recorded in the application similarity list, the application program that is not recorded in the application installation list of the user group is deleted, so that the application program that is not installed by the users in the user group is deleted.

The application installation list of each user group can be obtained by adopting the following modes: for a certain user group, counting the application programs installed by each user of the user group, so as to determine the application programs installed by the users in the user group, determine m application programs with the highest installation amount, wherein m can be 5 thousands or other numerical values, and sort the m application programs with the highest installation amount in a descending order according to the installation amount to obtain an application program installation list of the user group.

In act (d3), the deleted application similarity list is used as the second application similarity list corresponding to the user group. Since the applications not installed by the user in the user group are deleted through the action (d2), the similar applications recorded in the second application similarity list are all the applications installed by the user in the user group.

In this embodiment, by obtaining the second application program similarity list corresponding to the user group, the application programs installed by the users in the user group can be recommended to the users conveniently, so that the situation that the application programs recommended to the users do not conform to the classification situation of the users is avoided. For example, for fitness applications, a user group where a target user is located mostly uses a keep application, and therefore, when the target user searches for the fitness applications, the keep application is recommended to the target user.

In this embodiment, the application installation list of each user group may be counted in advance, and the second application similarity list corresponding to each user group is determined according to the application installation list of each user group and the application similarity list, so that the application is conveniently and quickly recommended to the user.

In this embodiment, when recommending an application program to a user, the application program may be recommended to the user according to a search request of the user, and may also be recommended to the user according to an application program recently installed by the user. For example, detecting application programs installed by a user in the last week, when a life-class application program is detected, searching for at least one first-class similar application corresponding to the detected application program in a first application program similar list, determining at least one second-class similar application corresponding to the detected application program according to an application program recommendation strategy based on a collaborative filtering algorithm, sequencing the at least one first-class similar application and the at least one second-class similar application through a pre-trained sequencing model, and recommending the sequenced at least one first-class similar application and the at least one second-class similar application to the user. Similarly, when detecting the non-living application, according to the same manner, recommending the application to the user according to the second application similarity list and the application recommendation policy based on the collaborative filtering algorithm, so as to increase the diversity of the application recommendation scene, so that the user can browse the recommended application each time the user opens the application download software (such as a software store or Google Play).

In the embodiment of the application, the recommended application information is displayed on an interface of a mobile terminal of a user, so that the application is recommended to the user.

As can be seen from the above-mentioned process for generating the first application program similarity list and the second application program similarity list, the generation of the first application program similarity list and the second application program similarity list both depend on the application program similarity list, and therefore, in order to ensure that the application program similarity list contains more application programs, the application program similarity list preferably includes all application programs existing in the current network, so as to ensure the comprehensiveness of the application programs. Before the first application similarity list and the second application similarity list are generated, the application similarity list may be generated by:

(e1) determining all application programs installed by a network-wide user, capturing a fifth similar application list corresponding to each application program in a Google Play application store according to mark information in the Google Play application store, generating a fifth list, sequencing the application programs in the fifth list according to the sequence of the installation amount of each application program from high to low, and recording the similar application of each application program;

(e2) in the GooglePlay application store, determining description information of each application program in the GooglePlay application store, performing semantic extraction on the description information of each application program by using algorithms such as textrank, tfidf and rake, performing NER named entity recognition, labeling named entities such as product names in the description information, combining semantic extraction results and labeling results to obtain labels of each application program, performing vectorization expression on the labels of each application program by using a word embedding algorithm, determining semantic similarity among the labels of each application program according to the vectorization expression results of the labels of each application program, determining similar applications of each application program according to the semantic similarity among the labels of each application program, generating a sixth list, wherein the installation amount of each application program is in a descending order, sequencing the application programs and recording similar applications of each application program;

(e3) and combining the fifth list and the sixth list to obtain an application program similarity list. Specifically, the p application programs (or all applications) with the highest installation amount in the fifth list and the similar applications of each application program are taken, similarly, the p application programs (or all applications) with the highest installation amount in the sixth list and the similar applications of each application program are taken, the application programs taken out from the fifth list and the application programs taken out from the sixth list are cross-sorted to obtain an application program similar list, for example, A, B, C is taken out from the fifth list, and a, B and C are taken out from the sixth list, and then the application program similar list is a, B, C and C. The applications in the application similarity list are sorted from top to bottom according to installation amount, similar applications of the applications are recorded, p can take 5 ten thousand or other numerical values, and the format of data taken out from the fifth or sixth list can be specifically A1-A2 and A3, wherein A1 is the identifier of the application, and A3 and A2 are similar applications of A1.

In this embodiment, the application similarity list is generated by combining all the applications installed by the network-wide user and all the applications in the google play application store, and the application similarity list can be ensured to include all the applications existing in the network-wide environment, so as to ensure the comprehensiveness of the applications.

In this embodiment, before step S102, the country where the user is located may be determined according to the positioning information of the mobile terminal of the user, and the country where the user is located may be determined by:

(1) acquiring an application program installation list of a current country of a user; the application program installation list records a plurality of applications installed by users in the current country;

(2) judging whether the current country of the user is the resident country of the user or not according to the matching degree between the application program installation list and the application program installation list of the user; the application program installation list of the user comprises all the application programs installed by the user;

(3) and if so, determining the current country of the user as the living country of the user, otherwise, determining the living country of the user according to the matching degree between the application program installation list of other countries and the application program installation list of the user.

As described above, in the present embodiment, the application installation list for each country for recording the popular applications installed by the users in the corresponding country may be counted in advance. In this embodiment, an application installation list of a country where a user is currently located is obtained, a matching degree between the application installation list and an application installation list of the user is determined, where the matching degree refers to a ratio of the number of applications that are overlapped in the application installation list and the application installation list of the user to the number of applications in the application installation list of the user, and if the matching degree is greater than a preset value, the country where the user is currently located is determined as a living country of the user, and the country where the user is currently located is determined as a living country of the user, otherwise, the matching degrees between the application installation lists of other countries and the application installation list of the user are respectively calculated, and a country corresponding to the application installation list with the largest matching degree is determined as the living country of the user. In this way, the target user's country of residence may be determined.

In the above process, the matching degree is calculated, for example, when the number of the applications overlapped in the application installation list and the application installation list of the user is 20, and the number of the applications in the application installation list of the user is 30, the matching degree is 20/30 — 66.7%.

Since the application program used by the user who lives in a certain country is more consistent with the application program used by the user in the country, in this embodiment, the living country of the user is determined according to the matching degree between the application program installation list of the user and the application program installation list of the country, which has an advantage of accurate determination effect. By determining the living country of the user and the country where the user is located, the condition that the user lives in a certain country in a short term can be distinguished, and therefore accurate user group division is carried out.

In this embodiment, when the number of users in the user group where the target user is located is less than or equal to the predetermined number, recommending an application program installation list corresponding to all the user groups to the target user; the method comprises the steps that a plurality of application programs installed by users in all user groups are recorded in a descending order according to installation amount of the application programs in an application program installation list corresponding to all the user groups.

In this embodiment, for users in all user groups, according to an application installed by each user, all installed applications are determined, an installation amount of each application is determined, 5 ten thousand (certainly, other predetermined values) application programs with the highest installation amount are determined, the application programs are arranged in a descending order according to the installation amount, an application installation list corresponding to all the user groups is obtained, and when the number of users in a user group where a target user is located is less than or equal to a predetermined number (for example, 1 ten thousand), the application installation list is recommended to the target user, so that the application programs are recommended to the target user based on program installation conditions of the users in all the user groups.

Fig. 2 is a schematic flowchart of an application recommendation method according to another embodiment of the present application, and as shown in fig. 2, the method includes the following steps:

step S202, an application search request of a user is obtained.

Step S204, determining whether the number of users in the user group is greater than a preset number.

If so, go to step S206, otherwise, go to step S212.

Step S206, determining whether the category of the searched application corresponding to the application search request is a life category.

If so, go to step S208, otherwise, go to step S210.

And step S208, recommending the application program to the user according to the first application program similarity list corresponding to the current country of the user and the application program recommendation strategy based on the collaborative filtering algorithm.

And step S210, recommending the application program to the user according to the second application program similarity list corresponding to the user group where the user is located and the application program recommendation strategy based on the collaborative filtering algorithm.

In step S212, the application installation list corresponding to all the user groups is recommended to the user.

In summary, the embodiment of the present application has the following advantages:

(1) according to various kinds of relevant information of the users, the users are divided into groups, so that the recommendation accuracy of the application program is improved, and the situations that the whats app is recommended in a three-four-wire city, the live broadcast application is recommended in a religious area and the like are avoided;

(2) the current country and the resident country of the user are taken as one of the factors of user clustering, so that the application program can be more accurately recommended to the user living in the country other than the resident country in a short term;

(3) the life application and the non-life application are treated in a distinguishing mode, and the application program recommendation experience of the user is improved.

Corresponding to the foregoing method, an embodiment of the present application further provides an application recommendation device, and fig. 3 is a schematic diagram of module compositions of the application recommendation device provided in an embodiment of the present application, and as shown in fig. 3, the device includes:

a user clustering module 41, configured to divide the users in the whole network into a plurality of user groups according to the related information corresponding to each user; wherein the related information comprises at least one of religious beliefs, the city, the country, the language used, the stationary country, the stationary city, the income level of the city, and the income level of the stationary city;

a quantity judging module 42, configured to, when receiving an application search request of a target user, judge whether the number of users in a user group where the target user is located is greater than a predetermined number;

a category determination module 43, configured to determine an industry category of the searched application corresponding to the application search request of the target user if the number of the searched applications is greater than the predetermined number;

and the application recommending module 44 is configured to recommend an application program to the target user according to the first application program similarity list corresponding to the country where the target user is currently located if the industry category is life-class application, and recommend an application program to the target user according to the second application program similarity list corresponding to the user group where the target user is located if the industry category is non-life-class application.

By the method and the device, the application program can be recommended to the user in a personalized manner based on the search request of the user, the user group where the user is located and the country where the user is located, and the recommendation accuracy of the application program is improved. And when the user searches for the life application, the application program is recommended to the user based on the country where the user is located, so that the recommended application program can conform to the living habits of the country where the user is located, and the application search experience of the user is improved. And when the user searches the non-living application, the application program is recommended to the user based on the user group where the user is located, so that the recommended application program can be matched with the user group where the user is located, namely matched with the classification condition of the user, and the application searching experience of the user is improved.

Optionally, the user clustering module 41 is specifically configured to:

and clustering the users of the whole network according to a preset clustering algorithm and the related information corresponding to each user respectively to obtain a plurality of user groups.

Optionally, the first application program similarity list is used for recording a plurality of established applications and similar applications corresponding to each established application respectively; wherein, the similar applications corresponding to the established application are all the applications installed by the user in the current country;

the application recommendation module 44 is specifically configured to:

searching at least one first similar application corresponding to the searched application in the first application program similar list, and determining at least one second similar application corresponding to the searched application according to an application recommendation strategy based on a collaborative filtering algorithm;

sequencing the at least one first similar application and the at least one second similar application through a pre-trained sequencing model, and recommending the sequenced at least one first similar application and at least one second similar application to the target user; and the sequencing model is obtained by training based on the installation record of the target user on the historically recommended application program.

Optionally, the second application program similarity list is used for recording a plurality of established applications, and similar applications respectively corresponding to each established application; wherein, the similar applications corresponding to the established application are all the applications installed in the user group;

the application recommendation module 44 is specifically configured to:

at least one third similar application corresponding to the searched application is searched in the second application program similar list, and at least one fourth similar application corresponding to the searched application is determined according to an application recommendation strategy based on a collaborative filtering algorithm;

sequencing the at least one third similar application and the at least one fourth similar application through a pre-trained sequencing model, and recommending the sequenced at least one third similar application and at least one fourth similar application to the target user; and the sequencing model is obtained by training based on the installation record of the target user on the historically recommended application program.

Fig. 4 is a schematic diagram of module components of an application recommendation apparatus according to another embodiment of the present application, as shown in fig. 4, optionally, further including a first list determining module 51, configured to:

acquiring a predetermined application program similarity list; each application in the whole network and the similar application corresponding to each application are recorded in the application program similarity list;

deleting the application programs which are not installed by the user in the current country from the similar applications corresponding to each application recorded in the application program similarity list;

and taking the deleted application program similarity list as a first application program similarity list corresponding to the current country.

As shown in fig. 4, optionally, a second list determining module 52 is further included for:

deleting the application programs which are not installed by the users in the user group in the similar applications corresponding to each application recorded in the application program similar list;

and taking the deleted application program similarity list as a second application program similarity list corresponding to the user group.

Optionally, the system further comprises a country of residence determining module, configured to:

acquiring an application program installation list of a current country of a user; the application program installation list records a plurality of applications installed by the user in the current country;

judging whether the current country of the user is the resident country of the user according to the matching degree between the application program installation list and the application program installation list of the user;

and if so, determining the current country of the user as the living country of the user, otherwise, determining the living country of the user according to the matching degree between the application program installation list of other countries and the application program installation list of the user.

Further, based on the foregoing method, an embodiment of the present application further provides an application recommendation device, and fig. 5 is a schematic structural diagram of the application recommendation device provided in the embodiment of the present application.

As shown in fig. 5, the application recommendation device may have a relatively large difference due to different configurations or performances, and may include one or more processors 701 and a memory 702, where one or more stored applications or data may be stored in the memory 702. Memory 702 may be, among other things, transient storage or persistent storage. The application stored in memory 702 may include one or more modules (not shown), each of which may include a series of computer-executable instructions in a recommendation device for the application. Still further, the processor 701 may be configured to communicate with the memory 702 to execute a series of computer-executable instructions in the memory 702 on the application recommendation device. The application recommendation apparatus may also include one or more power supplies 703, one or more wired or wireless network interfaces 704, one or more input-output interfaces 705, one or more keyboards 706, and the like.

In a specific embodiment, the application recommendation device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and when executed by the processor, the computer program can implement the process as described in the above application recommendation method embodiment, and specifically includes the following steps:

Optionally, when the computer program is executed by the processor, dividing the users in the whole network into a plurality of user groups according to the related information corresponding to each user, including:

Optionally, when the computer program is executed by the processor, the first application program similarity list is used for recording a plurality of established applications and similar applications corresponding to each established application respectively; wherein, the similar applications corresponding to the established application are all the applications installed by the user in the current country;

recommending the application program to the target user according to the first application program similarity list corresponding to the current country of the target user, wherein the recommending comprises the following steps:

Optionally, when the computer program is executed by the processor, the second application program similarity list is used for recording a plurality of established applications and similar applications corresponding to each established application respectively; wherein, the similar applications corresponding to the established application are all the applications installed in the user group;

recommending the application program to the target user according to the second application program similarity list corresponding to the user group where the target user is located, wherein the recommending comprises the following steps:

Optionally, the computer program, when executed by the processor, further comprises:

Further, an embodiment of the present application further provides a storage medium for storing computer-executable instructions, and in a specific embodiment, the storage medium may be a usb disk, an optical disk, a hard disk, or the like, and when the computer-executable instructions stored in the storage medium are executed by a processor, the processes described in the foregoing embodiment of the application program recommendation method can be implemented, and the same effect is achieved, which is not described herein again.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. An application recommendation method, comprising:

2. The method of claim 1, wherein dividing users of the whole network into a plurality of user groups according to the related information corresponding to each user respectively comprises:

3. The method of claim 1, wherein the first application similarity list is used for recording a plurality of predetermined applications, and each predetermined application corresponds to a similar application; wherein, the similar applications corresponding to the established application are all the applications installed by the user in the current country;

4. The method of claim 1, wherein the second application similarity list is used for recording a plurality of predetermined applications, and each predetermined application corresponds to a similar application; wherein, the similar applications corresponding to the established application are all the applications installed in the user group;

5. The method of claim 1, further comprising:

6. The method of claim 1, further comprising:

7. The method of any of claims 1 to 6, further comprising:

8. An application recommendation apparatus, comprising:

9. The apparatus of claim 8, wherein the user clustering module is specifically configured to:

10. The apparatus of claim 8, wherein the first application affinity list is used to record a plurality of predetermined applications and each predetermined application corresponds to a similar application; wherein, the similar applications corresponding to the established application are all the applications installed by the user in the current country;

the application recommendation module is specifically configured to:

11. The apparatus of claim 8, wherein the second application similarity list is used for recording a plurality of predetermined applications and each predetermined application corresponds to a similar application; wherein, the similar applications corresponding to the established application are all the applications installed in the user group;

the application recommendation module is specifically configured to:

12. The apparatus of claim 8, further comprising a first list determination module to:

13. The apparatus of claim 8, further comprising a second list determination module to:

14. The apparatus of any one of claims 8 to 13, further comprising a country of residence determination module to: