CN107729544B

CN107729544B - Method and device for recommending applications

Info

Publication number: CN107729544B
Application number: CN201711055130.5A
Authority: CN
Inventors: 潘岸腾
Original assignee: Alibaba China Co Ltd
Current assignee: Alibaba China Co Ltd
Priority date: 2017-11-01
Filing date: 2017-11-01
Publication date: 2021-06-22
Anticipated expiration: 2037-11-01
Also published as: CN107729544A

Abstract

The invention provides a method and a device for recommending applications. The method comprises the following steps: obtaining the probability of the target application installed by the user by using a prediction method of the probability of the target application installed by the user; and selecting a certain number of applications to be recommended to the user according to a preset mode based on the probability. The prediction method comprises the following steps: predicting the occupation ratio of the application with the label i installed by the user at t +1 time by utilizing a Markov process principle based on the occupation ratio of the application with the label i installed by the user at t time; calculating the maximum likelihood estimation probability of the target application when the same label exists between at least 1 label of any application in the preset application resource library installed by the user and at least 1 label of the target application in the preset application resource library predicted to be installed by the user; and predicting the probability of the user installing the target application in the preset application resource library based on the maximum likelihood estimation probability and the prediction ratio.

Description

Method and device for recommending applications

Technical Field

The invention relates to the technical field of information processing, in particular to a method and a device for predicting the probability of installing a target application by a user and a method and a device for recommending an application based on the probability.

Background

With the development of the internet, the popularization of 3G and 4G mobile communication networks and the popularization of intelligent terminals, a large number of applications are generated. In order to better attract users and improve user experience, many APP stores or APP markets are developed, such as pea pods, PP assistants, APP markets developed by mobile phone manufacturers, and the like, and the APP stores or APP markets are originally designed to show users and provide downloads of various third-party application software (APP) suitable for smart phones, including but not limited to: the system comprises a system tool class, an office business class, a news reading class, a video playing class, a communication social class, a financial and financial class, a life and leisure class, an online shopping class, a game and entertainment class and the like. In order to attract more user attention and provide user loyalty, more new functions have been developed by the powerful application stores or application market providers, rather than just third party application software download functions.

Currently, application stores or application market products basically have an application recommendation function, but the existing application recommendation idea is to recommend popular applications based on counting the number of users who have downloaded and installed a certain type of application, and the existing recommendation method lacks personalization, for example, the game application a with the first popular ranking is not necessarily the game application a which the user B likes to play, for example, the shooting game with the first popular ranking is not necessarily the favorite game type for many female users. Similarly, in many reading applications of the same type and audio/video playing applications of the same type, different users have their favorite personalized application products, which is not necessarily the application product with the highest installation amount. Therefore, if a new application recommendation method is developed to implement personalized recommendation for different users, the user will have a good experience when using the application store or application market product with the personalized recommendation application function, and the satisfaction of the user is improved.

In addition, the maintenance and re-development of various app stores or app markets for applications in the market requires capital investment, and app stores or app market providers can continue to provide better-experienced app store or app market products only if balance is achieved. And the income sources of the application store or application market products are mainly commercial promotion and various advertisement playing of third-party application software. In the process of commercial promotion of third-party application software such as game entertainment popular with users, how to accurately predict the probability of installing a certain third-party application by a user is very important for application stores or application market providers, which is closely related to the commercial promotion effect of the third-party application software and indirectly influences the income of the application stores or the product using the market. In order to maximize income profit in limited resources (including user resources, application display position resources and the like), accurate prediction of the probability of a user installing a certain third-party application is one of key elements.

Disclosure of Invention

The invention aims to provide a method and a device for predicting the probability of installing a target application by a user and a method and a device for recommending an application based on the probability of installing the target application so as to improve the problems.

A first embodiment of the present invention provides a method for predicting a probability of a user installing a target application, where the target application is from a preset application resource library, the method includes:

A) the method comprises the following steps Based on the actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t, predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t +1 by utilizing the Markov process principle;

B) the method comprises the following steps Calculating a ratio of the number of users, in a preset application resource library, who have installed target applications with at least 1 label predicted to be installed by the user u, to the number of users, in the preset application resource library, who have installed applications with at least 1 label identical to the label of the target application, and taking the ratio as a maximum likelihood estimation probability that an application is a target application predicted to be installed when the user u installs any one of the applications in the preset application resource library and the same label exists between at least 1 label of the application and at least 1 label of the target application in the preset application resource library predicted to be installed by the user u;

C) the method comprises the following steps And predicting the probability of the user u installing the target application in the preset application resource library by adopting a total probability method based on the maximum likelihood estimation probability and the prediction ratio of the application with the label i in the preset application resource library installed by the user u at the time of t + 1.

Wherein, in the step A),

predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t by utilizing a Markov process principle based on the occupation ratio of the application with the label i in the preset application resource library already installed by the user u at the time t-1, the occupation ratio of the application with the label a in the preset application resource library newly installed by the user u from the time t-1 to the time t and the occupation ratio of the application with the label d in the preset application resource library unloaded by the user u during the period;

calculating to obtain the required transition probability based on the obtained predicted occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t and the obtained actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t;

and according to the obtained transition probability and the actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t, predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t +1 by utilizing the Markov process principle.

A second embodiment of the present invention provides an application recommendation method, where a probability that a user installs a target application in a preset application resource library is obtained according to the method described in the first embodiment or the method described in the combination of the first embodiment and the preferred embodiment, and a certain number of applications are selected in a preset manner based on the probability to recommend to the user u.

A third embodiment of the present invention provides an application recommendation method, wherein a probability that a user installs a target application in a preset application resource library is obtained according to the method described in the first embodiment or the method described in the combination of the first embodiment and the preferred embodiment, a product of the obtained probability that the user u installs the target application and a charged price of a downloaded application is used as an expected revenue for recommending the target application to the user u, and a certain number of applications are selected to be recommended to the user u in a preset manner based on the expected revenue.

Wherein the charged price for downloading the application comprises: the price charged for downloading each application is the same or different; when the charging price for downloading each application is different, the charging price for downloading the application is the charging price for downloading the target application.

The fourth embodiment of the present invention further provides an apparatus for predicting a probability of a user installing a target application, where the target application is from a preset application resource library, the apparatus includes:

the application installation duty ratio prediction unit is used for predicting the duty ratio of the application with the label i in the preset application resource library installed by the user u at t +1 time by utilizing the Markov process principle based on the actual duty ratio of the application with the label i in the preset application resource library installed by the user u at t time;

a maximum likelihood estimation probability determination unit, configured to calculate a ratio between the number of users, in a preset application resource library, who have installed a target application with at least 1 tag predicted to be installed by the user u, and the number of users, in the preset application resource library, who have installed an application with at least 1 tag identical to the tag of the target application, and use the ratio as a maximum likelihood estimation probability that the application is the target application predicted to be installed when the user u installs any one application in the preset application resource library, and when there is an identical tag between at least 1 tag that the user u has when installing the application and at least 1 tag that the target application has in the preset application resource library predicted to be installed;

and the installation target application probability prediction unit is used for predicting the probability of the user u installing the target application in the preset application resource library by adopting a total probability method based on the maximum likelihood estimation probability and the prediction ratio of the application with the label i in the preset application resource library installed by the user u at the time t + 1.

Wherein the application installation proportion prediction unit further comprises:

a first prediction unit, configured to predict, by using a markov process principle, an occupancy rate of an application with a label i in a preset application resource library installed by a user u at t-1 time based on the occupancy rate of the application with the label i in the preset application resource library already installed by the user u at t-1 time, the occupancy rate of an application with a label a in the preset application resource library newly installed by the user u from t-1 time to t time, and the occupancy rate of an application with a label d from the preset application resource library uninstalled during the period;

and the transition probability obtaining unit is used for calculating and obtaining the required transition probability based on the obtained predicted occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t and the obtained actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t.

The fifth embodiment of the present invention also provides an application recommendation apparatus, which includes:

the apparatus for predicting the probability of the user installing the target application according to the fourth embodiment or the combination of the fourth embodiment and the preferred embodiment thereof;

and the recommending unit is used for selecting a certain number of applications to recommend to the user according to a preset mode based on the probability of the target application in the preset application resource library installed by the user, which is obtained by the device for predicting the probability of the target application installed by the user.

The sixth embodiment of the present invention also provides an application recommendation apparatus, including:

an expected profit obtaining unit configured to take a product of a probability that the user installs a target application in a preset application resource base, which is obtained by the user installation target application probability predicting unit, and a charged price of the downloaded application as an expected profit for recommending the target application to the user;

and the recommending unit is used for selecting a certain number of applications to recommend to the user according to a preset mode based on the expected income.

The seventh embodiment of the present invention also provides a storage device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method as described above in the first embodiment or in combination with the preferred embodiment.

The eighth embodiment of the present invention also provides a storage device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method according to the second embodiment or the method according to the second embodiment in combination with the preferred embodiment.

The ninth embodiment of the present invention also provides a storage device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method as described above in the third embodiment or in combination with the preferred embodiment.

According to the method and the device for predicting the target application installation probability of the user, the probability that the user downloads and installs a certain APP by using an application store or an application market product is predicted by using the Markov process principle, the prediction accuracy is high, the method and the device are also beneficial to more accurately recommending relevant applications to the user, the target application with high installation probability to a certain user is recommended to the user, the purpose of recommending the application to the user in a personalized manner is further realized, and the good experience and satisfaction degree of the user are improved. And accurately recommending the implementation of the related application to the user, which is beneficial to the marketing promotion of the application store or the application market product, is beneficial to the commercial cooperation with the application software product of the third-party developer and various advertisement broadcasts, and when the provider of the application store or the application market product achieves balance through a commercial cooperation mode, the provider of the application store or the application market product is beneficial to the maintenance and the redevelopment of the application store or the application market product, thereby further improving the use experience of the user.

Drawings

FIG. 1 is a flowchart of a method for predicting a probability of a user installing a target application according to a first embodiment of the present invention;

FIG. 2 is a flowchart of a method for predicting a probability of a user installing a target application according to a second embodiment of the present invention;

fig. 3 is a flowchart of an application recommendation method according to a third embodiment of the present invention;

FIG. 4 is a flowchart of an application recommendation method according to a fourth embodiment of the present invention;

fig. 5 is a schematic block diagram of an apparatus for predicting a probability of a user installing a target application according to a fifth embodiment of the present invention;

fig. 6 is a schematic block diagram of an application installation proportion prediction unit according to a sixth embodiment of the present invention;

fig. 7 is a schematic block diagram of an application recommendation apparatus according to a seventh embodiment of the present invention;

fig. 8 is a schematic block diagram of an application recommendation apparatus according to an eighth embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention and the accompanying drawings, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.

As mentioned in the background, many software companies provide users with APP stores or APP market products, such as pea pods or PP assistants, and respective APP market products, such as provided by watson corporation, and users using smart terminals (smartphones, tablets, etc.) are also becoming more accustomed to downloading and installing their own desired third party APP products through various APP stores or APP market products. The downloading and installation of the application can be realized by an application store or an application market APP installed on the terminal, so that the time, labor and convenience are saved. The service providers of the APP products in the APP stores or APP markets can pre-establish an application resource library, namely a preset application resource library, and a large number of various third-party applications which can be downloaded and installed by users are stored in the preset application resource library, wherein the various third-party applications include office business APPs, audio-video playing APPs, communication social APPs, online shopping APPs, game APPs and the like. Different users have different frequencies and numbers of downloading and installing third-party applications through APP products of APP stores or APP markets and APP products, some users may frequently download APP of game types, for example, download and install a new game application product every other week, some users may only download APP of office business types and APP of communication social types, the interval time may be very long, and it may be possible to download and install the APP once every few weeks or months; of course, if the downloading of the APP with the same upgrade is considered as downloading, it may happen within 1 or 2 weeks. When a user downloads and installs a third party application using a different APP store or APP product, their click behavior, download and installation behavior will be logged on the server associated with the APP store or APP product used.

Based on the above, the applications related to the present invention all come from the preset application resource library, that is, the user-installed applications or the applications predicted to be installed by the present invention all refer to application resource libraries (preset application resource libraries) that are installed by downloading from APP product service providers and pre-established on backend servers through APP products of certain APP stores or APP markets installed on the smart terminals used by the users. And the application related to the invention has 1 or more labels from a label library which is established by an application store or an application market APP product service provider used by a user in advance, namely a preset application label library. In addition, in the whole text, the user installation application or the target application refers to that the user installs a certain application APP product on the intelligent terminal used by the user, and the description will not bring misunderstanding to those skilled in the art.

Fig. 1 is a flowchart of a method for predicting a probability of a user installing a target application according to a first embodiment of the present invention. As shown in fig. 1, the method for predicting the probability of installing the target application by the user of the present invention includes the following steps:

s1: and predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t by utilizing a Markov process principle based on the occupation ratio of the application with the label i in the preset application resource library already installed by the user u at the time t-1, the occupation ratio of the application with the label a in the preset application resource library newly installed by the user u from the time t-1 to the time t and the occupation ratio of the application with the label d in the preset application resource library unloaded by the user u during the period.

The third party applications provided by the application store or application marketplace will typically have 1 or more tags from a set of tags preset at the time the application store or application marketplace was developed by the facilitator, i.e., the tag i is from a set of all tags in a library of preset application tags, which is well known to those skilled in the art and is not described in great detail herein.

According to the definition of the Markov process, the inventor generalizes:

the next time state is current state + amount of transfer-amount of transfer, so the predicted share ratio of the application with tag i in the preset application resource library installed by the user u at time t is:

the sum of the occupancy of the application with the label i in the preset application repository already installed by the user u at time t-1 and the occupancy of the application with the label a in the preset application repository newly installed by the user u from time t-1 to time t is subtracted by the value after the occupancy of the application with the label d from the preset application repository unloaded during the period.

Here, the tag i, the tag a, and the tag d may be the same or different.

As introduced above, the number of applications with labels i in the preset application repository that user u has installed at time t-1, the number of applications with labels a in the preset application repository that user u newly installs from time t-1 to time t, and the number of applications with labels d from the preset application repository that user u uninstalls from time t-1 to time t may be obtained by reading the log record, and the obtained number of applications with labels i in the preset application repository that user u has installed at time t-1 is divided by the number of all third party applications that user u installs on the smart terminal that user u uses, and the obtained number of applications with labels a in the preset application repository that user u newly installs from time t-1 to time t is divided by the number of all third party applications that user u installs on the smart terminal that user u uses And dividing the obtained number of the applications with the labels d from the preset application resource library unloaded by the user u from the time t-1 to the time t by the number of all third-party applications installed by the user u on the intelligent terminal used by the user u, thereby obtaining the occupation ratio of the applications with the labels i in the preset application resource library already installed by the user u at the time t-1, the occupation ratio of the applications with the labels a in the preset application resource library newly and additionally installed by the user u from the time t-1 to the time t, and the occupation ratio of the applications with the labels d unloaded by the user u from the preset application resource library from the time t-1 to the time t.

Namely, the proportion of the applications with the labels i in the preset application resource library which is already installed by the user u at the time t-1 is the ratio of the number of the applications with the labels i in the preset application resource library which is already installed by the user u at the time t-1 to the number of all the applications installed by the user u; the proportion of the applications with the labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t is the ratio of the number of the applications with the labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t to the number of all the applications installed by the user u; the proportion of the applications with the labels d from the preset application resource library, which are unloaded by the user u from the time t-1 to the time t, is the ratio of the number of the applications with the labels d, which are unloaded by the user u from the preset application resource library from the time t-1 to the time t, to the number of all the applications installed by the user u.

S2: and calculating to obtain the required transition probability based on the obtained predicted occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t and the obtained actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t.

Obtaining the predicted occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t through the step S1; the number of the applications with the labels i in the preset application resource library installed by the user u at the time t is obtained by reading the log record, and is divided by the number of all third-party applications installed on the intelligent terminal used by the user u, so that the actual occupation ratio of the applications with the labels i in the preset application resource library installed by the user u at the time t is obtained.

According to the definition of the markov process, the method for predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t can also adopt the following formula to calculate:

wherein:

i represents a set of all labels in a preset application label library, wherein I belongs to I, and j belongs to I;

p′_u,t,irepresenting the predicted occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t;

p_u,t-1,irepresenting the actual occupation ratio of the application with the label i in the preset application resource library which is installed by the user u at the time t-1;

q_i,jthe transition probability represents the amount of the application occupation ratio with the label i in the user u installation preset application resource library transferred to the application occupation ratio with the label j in the user u installation preset application resource library;

p_u,t-1,jrepresenting the actual occupation ratio of the application with the label j in the preset application resource library which is installed by the user u at the time t-1;

q_j,ithe transition probability represents the amount of the application proportion with the label j in the preset application resource library installed by the user u compared with the application proportion with the label i in the preset application resource library installed by the user u.

The label i comes from the preset application label library.

Any known method may be used to calculate the required transition probability q_i,jAnd q is_j,i. For example, in a study on a method for solving a transition probability matrix of a markov process, which was written in advance by students of northeast agronomy and published in 2013, "a method for solving a transition probability matrix of a markov process" was discussed in detail.

In addition, the embodiment of the invention provides a method for calculating the required transition probability by a gradient descent method.

With respect to the relationship between the transition probabilities of the markov process and the transition probability matrix, the transition probability matrix is a matrix in which the transition probabilities are arranged in a certain order, which is well known in the art and is not explained herein.

Let transition probability matrix { q_i,jThe objective loss function for I, j ∈ I } is:

wherein q is (q)_1,1,q_1,2,…)

Here, I represents a set of all tags in the preset application tag library, where I ∈ I;

u represents the set of all users installed with the same app store or app market product;

p_u,t,iindicating that user u has installed the actual percentage of applications with labels i in the preset application repository at time t.

A simple explanation is made here for the set U, i.e. the set of all users who have APP products of the same kind, such as APP stores or APP markets, installed on the terminal used by the user. Of course, the set of all users counted here refers to a set of users installed with APP products such as APP stores or APP markets of the same type (which may be different versions), for example, a set of all users installed with PP helper products on the terminals used. Those skilled in the art will appreciate that the application store or application market product referred to in the embodiments described herein is an application store or application market product that uses or integrates the methods and apparatus provided by the present invention.

This objective loss function is used to represent the error between the actual and said predicted occupation ratio of the application with label i that user u has installed in the preset application repository at time t, with the aim of minimizing this error.

The objective loss function is a function of q ═ q (q)_1,1,q_1,2…) (the first derivative is a monotonic function), and the optimal q-value (q) can be obtained by the gradient descent method_1,1,q_1,2…).

By gradient descentMethod for solving the equation_i,jThe best estimation value of | I, j ∈ I } comprises the following specific steps:

step 1: randomly given a set of transition probability matrices between 0-1 q_i,jI, j belongs to I, and q is set as⁽⁰⁾Initializing the iteration step number k to be 0;

step 2: and (3) iterative calculation:

where θ is the number of steps in the iteration, here taken to be 0.001;

and step 3: judging whether convergence occurs:

Δf(q^(k+1))＝|f(q^(k+1))-f(q^(k))|

if | Δ f (q)^(k+1))-Δf(q^(k))|<α, then q is returned^(k+1)I.e. the estimated transition probability matrix, otherwise, the calculation is continued by returning to step 2, where α is a small value, and may be 0.1 · θ.

Therefore, the required transition probability can be calculated by adopting the gradient descent method introduced by the invention.

S3: and according to the obtained transition probability and the actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t, predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t +1 by utilizing the Markov process principle.

After the required transition probability is obtained, based on the obtained transition probability and the actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t, the occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t +1 can be predicted by utilizing the Markov process principle.

The method for predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at t +1 time by using the Markov process principle comprises the following steps:

wherein:

p′_u,t+1,irepresenting that the user u installs the occupation ratio of the application with the label i in the preset application resource library at the time of t + 1;

p_u,t,irepresenting the actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at the time t;

p_u,t,jrepresenting the actual occupation ratio of the application with the label j in the preset application resource library installed by the user u at the time t;

The application described here has a tag i from the library of preset application tags.

S4: and calculating the ratio of the number of users, in a preset application resource library, of target applications with at least 1 label predicted to be installed by the user u to the number of users, in the preset application resource library, of applications with at least 1 label identical to the label of the target application, and taking the ratio as the maximum likelihood estimation probability that the application is the target application predicted to be installed when the same label exists between at least 1 label of any application in the preset application resource library installed by the user u and at least 1 label of the target application in the preset application resource library predicted to be installed by the user u.

The purpose of this step is to calculate the maximum likelihood estimation probability that an application z is predicted to be installed when the user installs a certain application z in the preset application repository.

For example, assuming that any application b in the preset application resource library is a target application predicted to be installed by the user u, the target application b is provided with n tags,

by c_b,jJ-th tag representing the target application b having n tags, j being 1,2, …, n;

by x_bIndicating that the target application b event is installed on the intelligent terminal used by the user;

with C_b,jIndicates that the label c of the target application b is contained in 1 or more labels of any application in the preset application resource library installed by the user_b,jEvent (2), namely C_b,jIndicating that the same event as the jth label of the target application b exists in 1 or more labels of any application in a preset application resource library installed by a user;

calculating a ratio of a number of users, who have installed a target application b having at least 1 tag in a preset application repository, to a number of users, who have installed an application having at least 1 tag identical to a tag of the target application b in the preset application repository, for a predetermined period of time:

wherein:

InstAppNums(x_b) Representing the number of users who have installed the target application b having at least 1 tag in the preset application repository for a predetermined period of time;

InstCategNums(C_b,j) Indicating the number of users having at least 1 application with the same tag as the target application b in the preset application repository installed during the predetermined period of time.

Here, at least 1 tag in the preset application repository is installed to be identical to the tag of the target application bThe application of (A) is as follows: when any application z in the preset application resource library is installed, the same label exists between 1 or more labels of the application z and 1 or more labels of the target application b, the same label may be 1 or more, and even the plurality of labels of the application z may include all the labels of the target application b. Apparently, InstAppnums (x)_b) Is InstCategNums (C)_b,j) A subset of (a).

For example, assume that the target APP b predicted to be installed by the user u is a flag novel APP provided in a PP assistant, which has 2 tags: news reading and novels; when any user downloads and installs the palm reading APP through the PP assistant, the tags of the APP are as follows: novels and electronic books; the palm-reading APP installed has the same tag "novel" as the target application b has.

The obtained ratio is the probability that the installed application is the application b when the user installs the application having at least 1 label identical to the label of the application b in the preset application resource library. With P (x)_b|C_b,j) When this probability is expressed, then

In addition, the predetermined period of time is only a predetermined period of time for calculating the ratio, and in practice, a suitable period of time may be selected according to the balance point between the required accuracy and the calculation amount, and may be, for example, but not limited to, 7 days, 15 days, 30 days, and the like.

The ratio P (x) to be obtained_b|C_b,j) And the maximum likelihood estimation probability that an application y is the target application b predicted to be installed when the same label exists between at least 1 label of any application y in the preset application resource library installed by the certain user u and at least 1 label of the target application b in the preset application resource library predicted to be installed by the user u. Setting the maximum likelihood estimation probability to P_u(x_b|C_b,j) Indicates that P (x) is to be_b|C_b,j) As P_u(x_b|C_b,j) An estimate of (d).

S5: and predicting the probability of the user u installing the target application in the preset application resource library by adopting a total probability method based on the maximum likelihood estimation probability and the prediction ratio of the application with the label i in the preset application resource library installed by the user u at the time of t + 1.

After obtaining the maximum likelihood estimation probability that the application is the target application predicted to be installed when the same label exists between at least 1 label of any application in the preset application resource library installed by the user u and at least 1 label of the target application in the preset application resource library predicted to be installed by the user u, and the prediction occupation ratio of the third-party application with the label i in the preset application resource library installed by the user u at t +1 time, the probability that the target application in the preset application resource library is installed by the user u can be predicted by using a total probability method.

The method for predicting the probability of the user u to install the target application in the preset application resource library comprises the following steps:

wherein:

P_u(x_b) Representing the probability of the user u installing the target application b in the preset application resource library;

n represents the number of labels of the target application b;

c_b,jj-th tag representing a target application b having n tags, j being 1,2, …, n;

representing the occupation ratio of the application with the jth label of the target application b in a preset application resource library installed by the user u at t +1 time;

P(x_b|C_b,j) Representing the maximum likelihood estimation probabilityThat is, the maximum likelihood estimation probability represents that when the user installs any application in the preset application resource library, the application has at least 1 tag, and when the same tag exists between at least 1 tag that the target application b in the preset application resource library predicted to be installed by the user has, the application is the target application b predicted to be installed.

In addition, as described by way of example above, with x_bRepresenting the user to install the target application b event, using C_b,jIndicating that the same event as the jth label of the target application b exists in the labels of any application in the preset application resource library installed by the user.

Concerning obtaining calculation P_u(x_b) The deduction process of the formula (1) is as follows:

for any user u mentioned in the present invention, according to the total probability formula, the probability of the user u installing the target application b in the preset application resource library can be written as follows:

wherein:

P_u(C_b,j) Representing the probability that the user u installs the application containing at least 1 label of the target application b in a preset application resource library; in other words, the probability that the same event as the jth tag of the target application b exists in 1 or more tags of any application installed in a preset application resource library by the user u is represented;

n represents the number of labels of the target application b;

P_u(x_b|C_b,j) Representing the maximum likelihood estimation probability, that is, at least 1 label of any application in the preset application resource library when the user u installs the application and at least one label of the target application b in the preset application resource library predicted to be installed by the user uThe maximum likelihood estimation probability that the application is the target application b predicted to be installed when the same label exists between 1 label.

With P (x) obtained above_b|C_b,j) As P_u(x_b|C_b,j) Replacing the estimated value of (A);

by using

As P_u(C_b,j) Replacing the estimated value of (a) to obtain:

in a preferred embodiment, the time t mentioned throughout the present invention can be in the time units of days, weeks, months, preferably in the time units of weeks, i.e. a week of 7 days. While t-1 time represents the previous time unit and t +1 time represents the next time unit.

Thus, according to the method for predicting the probability of the user installing the target application, which is described above, the probability of the user downloading and installing a certain APP application by using an application store or an application market product can be predicted by using the Markov process principle, and the prediction accuracy is high.

Although the above provides a specific implementation method for predicting the occupation ratio of an application with a label i in a preset application resource library installed by a user u at t +1 time based on the actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at t time, the markov process principle is proposed by russian mathematician markov in 1907, and through research and development for many years, more derivative principles and implementation methods have been deduced, and all that is applicable to the present invention is predicting the occupation ratio of an application with a label i in a preset application resource library installed by the user u at t +1 time by using the markov process principle and the derivative principles.

To this end, fig. 2 is a flowchart of a method for predicting a probability of a user installing a target application according to a second embodiment of the present invention. As shown in fig. 2, the method for predicting the probability of installing the target application by the user of the present invention includes:

The implementation process of step a herein may adopt any method for predicting the occupation ratio of the application with the label i in the preset application resource library installed by the user u at time t +1 based on the well-known markov process principle and its derivative principle.

Here, the implementation process of step B is completely the same as the implementation process of step S4 in the first embodiment, and is not described here again.

Here, the implementation process of step C is completely the same as the implementation process of step S5 in the first embodiment, and is not described here again.

Thus, according to the prediction method for the user installation target application probability, the probability that the user downloads and installs a certain APP by using an application store or an application market product can be predicted, and the prediction accuracy is high.

Fig. 3 is a flowchart of an application recommendation method according to a third embodiment of the present invention; as shown in fig. 3, the application recommendation method of the present invention includes:

according to the method described in the first embodiment, or the method described in the combination of the first embodiment and the preferred embodiment thereof, or the method described in the second embodiment, the probability that the user u installs the target application in the preset application resource library is obtained, and based on the probability, a certain number of applications are selected in a preset manner to be recommended to the user u.

The step of selecting a certain number of applications to recommend to the user u according to the probability in a preset mode comprises the following steps: selecting a certain number of applications to recommend to a user u based on the probability from big to small; or randomly selecting a certain number of applications from the applications corresponding to the probability greater than or equal to the preset threshold value to recommend the applications to the user u.

And the user can know which applications are more interested by the user through the obtained installation target application probability, so that a certain number of applications are selected to be recommended to the user according to the installation target application probability in a preset mode. Preferably, a certain number of applications are selected from the large to the small based on the probability to recommend to the user. The specific number can be arbitrarily selected according to the actual application scenario, for example, the specific number may be 5 to 20, or 20 to 50, or may be a greater number of applications, such as 100, 200, and so on. In addition, a threshold may be preset, and a certain number of applications may be randomly selected from the applications corresponding to the probability greater than or equal to the preset threshold to be recommended to the user, for example, 5 to 20, or 20 to 50, or a greater number of applications may be selected, such as 100, 200, and the like.

According to the application recommendation method provided by the invention, the probability of installing the target application by the user with high accuracy is used, and the target application with high installation probability is recommended to a certain user, so that the purpose of recommending the application to the user in a personalized manner is realized, and the good experience and satisfaction of the user are improved.

FIG. 4 is a flowchart of an application recommendation method according to a fourth embodiment of the present invention; as shown in fig. 4, the application recommendation method of the present invention includes:

according to the method described in the first embodiment, or the method described in the combination of the first embodiment and the preferred embodiment thereof, or according to the method described in the second embodiment, the probability that the user u installs the target application in the preset application resource library is obtained, the product of the obtained probability that the user u installs the target application and the charged price of the downloaded application is used as the expected revenue of recommending the target application to the user u, and a certain number of applications are selected to be recommended to the user u in a preset manner based on the expected revenue.

The method comprises the following steps of selecting a certain number of applications to recommend to a user u according to a preset mode based on the expected income, wherein the steps comprise: selecting a certain number of applications to recommend to a user u based on the expected income from big to small; or randomly selecting a certain number of applications from the applications corresponding to the expected income which is greater than or equal to the preset threshold value to recommend the applications to the user u.

Wherein, the charging price for downloading the application comprises 2 cases that the charging price for downloading each application is the same or different; when the charging price for downloading each application is different, the charging price for downloading the application is the charging price for downloading the target application. When the charge price for downloading each application is the same, the charge price is determined by a service provider who provides an application store or application market product; when the price charged for downloading each application is different, the price charged is determined by the agreement of the application store or the service provider of the application market product with the provider of the application or with the agreement of the advertiser who binds the application to play the advertisement.

The applications which are more interesting to the user can be known through the obtained expected income, and simultaneously, income profit maximization can be realized, so that a certain number of applications are selected to be recommended to the user according to a preset mode based on the expected income. Preferably, a certain number of applications are selected to be recommended to the user in the descending order based on the expected profit value. The specific number can be arbitrarily selected according to the actual application scenario, for example, the specific number may be 5 to 20, or 20 to 50, or may be a greater number of applications, such as 100, 200, and so on. In addition, a threshold may be preset, and a certain number of applications may be randomly selected from the applications corresponding to the expected profit greater than or equal to the preset threshold to be recommended to the user, for example, 5 to 20, or 20 to 50, or a greater number of applications may be selected, such as 100, 200, and the like.

According to the application recommendation method provided by the invention, the probability of installing the target application by the user with high accuracy is used, and the target application with high installation probability is recommended to a certain user, so that the purpose of recommending the application to the user in a personalized manner is realized, and the good experience and satisfaction of the user are improved. In addition, the probability of installing the target application is multiplied by the unit price of the commercial download of the application, so that the expected income of the commercial application can be shown, the commercial application with high expected income is recommended to the user, or the commercial application with high expected income is placed in a high-quality position to preferentially show the advertisement bound with the commercial application, and the goal of maximizing income profit can be realized. In this way, it is advantageous for marketing of application stores or application market products, facilitating business collaboration with third party application software developers for commercial promotion of their application software products and various advertising, both from the perspective of accurately recommending relevant applications to users and from the perspective of maximizing revenue margins. When the providers of the application stores or the application market products achieve balance through the business cooperation mode, the maintenance and the re-development of the application stores or the application market products are facilitated, and therefore the use experience of the users is further improved.

Fig. 5 is a schematic block diagram of an apparatus for predicting a probability of a user installing a target application according to a fifth embodiment of the present invention. As shown in fig. 5, in the apparatus for predicting a probability of a user installing a target application, the target application being from a preset application resource library, the apparatus includes:

wherein:

i represents a set of all labels in a preset application label library, wherein I belongs to I;

q_i,jto transition probabilities, itRepresenting the amount by which user u installs the application proportion with label i to transfer to user u installs the application proportion with label j;

q_j,ia transition probability, which represents the amount by which user u installs the application having label j to transition to user u installs the application having label i;

the label i comes from the preset application label library.

The method for predicting the probability of the user u installing the target application in the preset application resource library by the installation target application probability prediction unit comprises the following steps:

wherein:

n represents the number of labels of the target application b;

P(x_b|C_b,j) And the maximum likelihood estimation probability is expressed, namely the maximum likelihood estimation probability of the application which is the target application b predicted to be installed when the same label exists between at least 1 label of any application in the preset application resource library and at least 1 label of the target application b predicted to be installed when the user installs the application.

It will be clear to those skilled in the art that for the convenience and brevity of description, the specific operation of the apparatus described above may be referred to the corresponding operation of the method of implementation described in the foregoing second embodiment, and the examples and related descriptions listed in the foregoing second embodiment are also applicable to the operation of the apparatus, and will not be repeated here.

Just as the first embodiment described above provides a specific implementation method for predicting the occupation ratio of an application with a label i in a preset application resource library installed by a user u at t +1 time by using the markov process principle based on the actual occupation ratio of the application with the label i in the preset application resource library installed by the user u at t time. In accordance with the sixth embodiment of the present invention, a description is also provided regarding a specific configuration of the application installation duty prediction unit.

Fig. 6 is a schematic block diagram of an application installation proportion prediction unit according to a sixth embodiment of the present invention. As illustrated in fig. 6, the application installation duty prediction unit may include:

In this way, the application installation duty prediction unit can predict the duty of the application with the label i in the preset application resource library installed by the user u at the time t +1 by using the markov process principle according to the obtained transition probability and the actual duty of the application with the label i in the preset application resource library installed by the user u at the time t.

In the first prediction unit, the proportion of the applications with the labels i in the preset application resource library which is already installed by the user u at the time t-1 is the ratio of the number of the applications with the labels i in the preset application resource library which is already installed by the user u at the time t-1 to the number of all the applications installed by the user u; the proportion of the applications with the labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t is the ratio of the number of the applications with the labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t to the number of all the applications installed by the user u; the proportion of the applications with the labels d from the preset application resource library, which are unloaded by the user u from the time t-1 to the time t, is the ratio of the number of the applications with the labels d, which are unloaded by the user u from the preset application resource library from the time t-1 to the time t, to the number of all the applications installed by the user u.

The method for predicting the proportion of the applications with the labels i in the preset application resource library installed by the user u at the time t by the first prediction unit comprises the following steps:

the proportion of the application with the label i in the preset application resource library which is already installed by the user u at the time t-1 is added with the proportion of the application with the label a in the preset application resource library which is newly installed by the user u from the time t-1 to the time t, and then the proportion of the application with the label d from the preset application resource library which is unloaded in the period is subtracted.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working process of the application installation duty prediction unit described above can refer to the corresponding process of the implementation method described in the foregoing first embodiment, and the example and the related description listed in the foregoing first embodiment are also applicable to explain the working process of the application installation duty prediction unit, and will not be described repeatedly herein.

Thus, according to the device for predicting the probability of the user installing the target application, which is described above, the probability that the user downloads and installs a certain APP by using an application store or an application market product can be predicted by using the Markov process principle, and the prediction accuracy is high.

Fig. 7 is a schematic block diagram of an application recommendation apparatus according to a seventh embodiment of the present invention. As shown in fig. 7, the application recommendation apparatus of the present invention includes:

the apparatus for predicting the probability of the user installing the target application according to the fifth embodiment or the fifth embodiment in combination with the preferred embodiment or the sixth embodiment thereof;

Wherein selecting a number of applications in a preset manner based on the expected revenue comprises: selecting a certain number of applications from large to small based on the expected income; or randomly selecting a certain number of applications from the applications corresponding to the expected income greater than or equal to the preset threshold value.

It will be clear to those skilled in the art that for the convenience and brevity of description, the specific operation of the apparatus described above can be realized by referring to the corresponding operation of the method described in the foregoing third embodiment, and the example and the related description listed in the foregoing third embodiment are also applicable to the operation of the apparatus, and will not be repeated here.

According to the application recommendation device provided by the invention, the probability of installing the target application by the user with high accuracy is used, and the target application with high installation probability is recommended to a certain user, so that the purpose of recommending the application to the user in a personalized manner is realized, and the good experience and satisfaction of the user are improved.

Fig. 8 is a schematic block diagram of an application recommendation apparatus according to an eighth embodiment of the present invention. As shown in fig. 8, the application recommendation apparatus of the present invention includes:

Wherein, the charge price of the downloaded application comprises 2 conditions that the charge price for downloading each application is the same or different; when the charging price for downloading each application is different, the charging price for downloading the application is the charging price for downloading the target application. When the charge price for downloading each application is the same, the charge price is determined by a service provider who provides an application store or application market product; when the price charged for downloading each application is different, the price charged is determined by the agreement of the application store or the service provider of the application market product with the provider of the application or with the agreement of the advertiser who binds the application to play the advertisement.

It is clear to those skilled in the art that for the convenience and brevity of description, the specific working process of the above-described apparatus can be referred to the corresponding process of the implementation method described in the foregoing fourth embodiment, and the examples and related description listed in the foregoing fourth embodiment are also applicable to the working process of the apparatus, and will not be repeated herein.

According to the application recommendation device provided by the invention, the probability of installing the target application by the user with high accuracy is used, and the target application with high installation probability is recommended to a certain user, so that the purpose of recommending the application to the user in a personalized manner is realized, and the good experience and satisfaction of the user are improved. In addition, the probability of installing the target application is multiplied by the unit price of the commercial download of the application, so that the expected income of the commercial application can be shown, the commercial application with high expected income is recommended to the user, or the commercial application with high expected income is placed in a high-quality position to preferentially show the advertisement bound with the commercial application, and the goal of maximizing income profit can be realized. In this way, it is advantageous for marketing of application stores or application market products, facilitating business collaboration with third party application software developers for commercial promotion of their application software products and various advertising, both from the perspective of accurately recommending relevant applications to users and from the perspective of maximizing revenue margins. When the providers of the application stores or the application market products achieve balance through the business cooperation mode, the maintenance and the re-development of the application stores or the application market products are facilitated, and therefore the use experience of the users is further improved.

The embodiment of the present invention further provides a computer program product for executing the method for predicting the probability of the user installing the target application and a computer program product for executing the application recommendation method, where the computer program product includes a computer-readable storage medium storing program codes, and instructions included in the program codes may be used to execute the method described in the foregoing method embodiment, and specific implementation may refer to the method embodiment, and details are not described herein.

To this end, the invention also provides a storage device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method as described above in the first embodiment or in combination with the preferred embodiments thereof.

To this end, the invention also provides a storage device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method as described above in the second embodiment or in combination with the preferred embodiments thereof.

To this end, the invention also provides a storage device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method as described above in the third embodiment or in combination with its preferred embodiments.

To this end, the invention also provides a storage device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method as described above in the fourth embodiment or in combination with the preferred embodiments thereof.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a smart tablet, a smart phone, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a U disk, a removable hard disk, a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.

Claims

1. A method for predicting the probability of a user installing a target application, wherein the target application is from a preset application resource library, the method comprises the following steps:

A) the method comprises the following steps Based on the actual occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t, predicting the occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t +1 by utilizing the Markov process principle;

C) the method comprises the following steps And predicting the probability of the user u installing the target application in the preset application resource library by adopting a total probability method based on the maximum likelihood estimation probability and the prediction ratio of the application with the first label i in the preset application resource library installed by the user u at the time of t + 1.

2. The method of claim 1, wherein in step a), the method for predicting the occupation ratio of the application with the first label i in the preset application resource library installed by the user u at t +1 time by using markov process principle comprises:

wherein:

p′_u，t+1，irepresenting that the user u installs the occupation ratio of the application with the first label i in the preset application resource library at the time of t + 1;

p_u，t，irepresenting the actual occupation ratio of an application with a first label i in a preset application resource library installed by a user u at time t;

q_i，ja transition probability, which represents the amount by which user u installs an application having the first label i to transition to user u installing an application having label j;

p_u，t，jrepresenting the actual occupation ratio of the application with the label j in the preset application resource library installed by the user u at the time t;

q_j，ia transition probability, which represents the amount by which user u installs the application having label j to transition to user u installing the application having the first label i;

the first tag i is from the library of preset application tags.

3. The method according to claim 1 or 2, characterized in that in step A),

predicting the occupation ratio of the application with the first label i in the preset application resource library installed by the user u at t time by utilizing a Markov process principle based on the occupation ratio of the application with the first label i in the preset application resource library already installed by the user u at t-1 time, the occupation ratio of the application with the second label a in the preset application resource library newly installed by the user u from t-1 time to t time and the occupation ratio of the application with the third label d in the preset application resource library unloaded by the user u during the period;

calculating to obtain a required transition probability based on the obtained predicted occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t and the obtained actual occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t;

and according to the obtained transition probability and the actual occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t, predicting the occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t +1 by utilizing the Markov process principle.

4. The method according to claim 3, wherein the proportion of the applications with the first label i in the preset application resource library already installed by the user u at time t-1 is the ratio of the obtained number of the applications with the first label i in the preset application resource library already installed by the user u at time t-1 to the number of all the applications installed by the user u; the proportion of the applications with the second labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t is the ratio of the number of the applications with the second labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t to the number of all the applications installed by the user u; the proportion of the applications with the third label d unloaded by the user u from the time t-1 to the time t from the preset application resource library is the ratio of the number of the applications with the third label d unloaded by the user u from the preset application resource library from the time t-1 to the time t to the number of all the applications installed by the user u.

5. The method of claim 3, wherein the predicted percentage of time t that the user u installs the application with the first tag i in the preset application resource library is:

the sum of the occupancy of the application with the first label i in the preset application repository that the user u has installed at time t-1 and the occupancy of the application with the second label a in the preset application repository that the user u has newly installed during time t-1 to time t minus the occupancy of the application with the third label d from the preset application repository that was uninstalled during that period.

6. The method of claim 2, wherein predicting the probability that user u installs the target application in the pre-configured application resource pool comprises:

wherein:

n represents the number of labels of the target application b;

c_b，ja jth tag representing a target application b having n tags, j ═ 1, 2.., n;

P(x_b|C_b，j) To indicate the poleThe maximum likelihood estimation probability represents the maximum likelihood estimation probability that an application is a target application b predicted to be installed when the same label exists between at least 1 label of any application in the preset application resource library and at least 1 label of the target application b predicted to be installed when the user installs the application.

7. An application recommendation method, characterized in that the method according to any one of claims 1 to 6 obtains the probability of installing a target application in a preset application resource library by a user u, and selects a certain number of applications to recommend to the user u in a preset manner based on the probability.

8. The application recommendation method according to claim 7, wherein the step of selecting a certain number of applications to recommend to the user u in a preset manner based on the probability comprises: selecting a certain number of applications to recommend to a user u based on the probability from big to small; or randomly selecting a certain number of applications from the applications corresponding to the probability greater than or equal to the preset threshold value to recommend the applications to the user u.

9. An application recommendation method, characterized in that the method according to any one of claims 1 to 6 obtains the probability of installing a target application in a preset application resource library by user u, takes the obtained product of the probability of installing the target application by user u and the charging price of downloading the application as an expected revenue for recommending the target application to user u, and selects a certain number of applications to recommend to user u in a preset manner based on the expected revenue.

10. The application recommendation method of claim 9, wherein the step of selecting a certain number of applications to recommend to the user u in a preset manner based on the expected revenue comprises: selecting a certain number of applications to recommend to a user u based on the expected income from big to small; or randomly selecting a certain number of applications from the applications corresponding to the expected income which is greater than or equal to the preset threshold value to recommend the applications to the user u.

11. The application recommendation method according to claim 9, wherein the charged price for downloading the application comprises: the price charged for downloading each application is the same or different; when the charging price for downloading each application is different, the charging price for downloading the application is the charging price for downloading the target application.

12. A memory device having stored therein a plurality of instructions adapted to be loaded by a processor and to perform the method of any of claims 1 to 6.

13. A memory device having stored therein a plurality of instructions adapted to be loaded by a processor and to carry out the method of one of claims 7 to 11.

14. An apparatus for predicting a probability of a user installing a target application, wherein the target application is from a preset application resource library, the apparatus comprising:

the application installation duty ratio prediction unit is used for predicting the duty ratio of the application with the first label i in the preset application resource library installed by the user u at t +1 time by utilizing the Markov process principle based on the actual duty ratio of the application with the first label i in the preset application resource library installed by the user u at t time;

and the installation target application probability prediction unit is used for predicting the probability of the user u installing the target application in the preset application resource library by adopting a total probability method based on the maximum likelihood estimation probability and the prediction ratio of the application with the first label i in the preset application resource library installed by the user u at the time t + 1.

15. The apparatus according to claim 14, wherein the method for predicting the occupation ratio of the application with the first label i installed in the preset application resource library by the user u at t +1 time using markov process principle comprises:

wherein:

the first tag i is from the library of preset application tags.

16. The apparatus according to claim 15, wherein the application installation proportion prediction unit further comprises:

a first prediction unit, configured to predict, by using a markov process principle, an occupancy of an application with a first tag i in a preset application repository installed by a user u at time t-1 based on the occupancy of the application with the first tag i in the preset application repository already installed by the user u at time t-1, the occupancy of an application with a second tag a in the preset application repository newly installed by the user u from time t-1 to time t, and the occupancy of an application with a third tag d from the preset application repository uninstalled during the period;

and the transition probability obtaining unit is used for calculating to obtain the required transition probability based on the obtained predicted occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t and the obtained actual occupation ratio of the application with the first label i in the preset application resource library installed by the user u at the time t.

17. The apparatus according to claim 16, wherein in the first prediction unit, the ratio of the applications with the first label i in the preset application resource library already installed by the user u at time t-1 is the ratio of the obtained number of applications with the first label i in the preset application resource library already installed by the user u at time t-1 to the number of all applications installed by the user u; the proportion of the applications with the second labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t is the ratio of the number of the applications with the second labels a in the preset application resource library newly installed by the user u from the time t-1 to the time t to the number of all the applications installed by the user u; the proportion of the applications with the third label d unloaded by the user u from the time t-1 to the time t from the preset application resource library is the ratio of the number of the applications with the third label d unloaded by the user u from the preset application resource library from the time t-1 to the time t to the number of all the applications installed by the user u.

18. The apparatus of claim 16, wherein the first prediction unit is configured to predict the percentage of the user u installing the application with the first tag i in the preset application resource pool at time t by:

the proportion of the application with the first label i in the preset application resource library which is already installed by the user u at the time t-1 is added to the proportion of the application with the second label a in the preset application resource library which is newly installed by the user u from the time t-1 to the time t, and then the proportion of the application with the third label d from the preset application resource library which is unloaded during the period is subtracted.

19. The apparatus of claim 15, wherein the means for predicting the probability of the user u installing the target application in the preset application resource library comprises:

wherein:

n represents the number of labels of the target application b;

P(x_b|C_b，j) The maximum likelihood estimation probability is represented, namely at least 1 label of any application in the preset application resource library and at least 1 label of the target application b in the preset application resource library predicted to be installed are represented when the user installs the applicationThe maximum likelihood estimation probability that the application is the target application b predicted to be installed when the same label exists between labels.

20. An application recommendation device, comprising:

means for predicting a probability of a user installing a target application according to one of claims 14-19;

21. The application recommendation device of claim 20, wherein selecting a number of applications in a preset manner based on the probability comprises: selecting a certain number of applications from large to small based on the expected income; or randomly selecting a certain number of applications from the applications corresponding to the expected income greater than or equal to the preset threshold value.

22. An application recommendation device, comprising:

23. The application recommendation device of claim 22, wherein selecting a number of applications in a preset manner based on the expected revenue comprises: selecting a certain number of applications from large to small based on the expected income; or randomly selecting a certain number of applications from the applications corresponding to the expected income greater than or equal to the preset threshold value.

24. The application recommendation device of claim 22, wherein the charged price for downloading an application comprises: the price charged for downloading each application is the same or different; when the charging price for downloading each application is different, the charging price for downloading the application is the charging price for downloading the target application.