WO2016107354A1 - Method and apparatus for providing user personalised resource message pushing - Google Patents

Abstract

Disclosed are a method and apparatus for providing user personalised resource message pushing. The method comprises the steps of: according to the frequency at which a user accesses resources in each date and each time period, determining one or more message pushing dates and time periods suitable for conducting message pushing on the user; under the condition that a current date and time period belong to the message pushing dates and time periods, selecting one or more resources meeting preference characteristics of the user from a resource set; and providing message pushing regarding one or more resources in an operating system.

Description

Method and device for providing user personalized resource message push Technical field

The present invention relates to the field of information processing technologies, and in particular, to a method and apparatus for providing user personalized resource message push.

Background technique

Personalized recommendation is a method of recommending personalized content to users based on the user's historical access record and calculating the user's possible preferences through algorithms such as machine learning. Message push (for example, pop-ups) is an important means of promotion, but it can also cause some level of disruption to users.

In the prior art, the setting is usually made by the product or the operator, the recommendation is made manually, or the message is pushed for all users. This will disturb the user, and can not achieve personalized recommendations for each user, the user experience is not good, affecting the promotion effect of message push.

Summary of the invention

In view of the above problems, the present invention has been made in order to provide a method and apparatus for providing user personalized resource message push that overcomes the above problems or at least partially solves the above problems.

According to a first aspect of the present invention, a method for providing user personalized resource message push is provided, comprising the steps of: determining, according to a frequency of a user accessing a resource on each date and each time period, one or more suitable for pushing a message to the user Multiple message push date and timely question segment; in the case that the current date timely belongs to the message push date and time segment, select one or more resources from the resource set that meet the user's preference characteristics; and in the operating system Provides message pushes about one or more resources.

According to a second aspect of the present invention, an apparatus for providing a user personalized resource message push is provided, including: a message push timing determining module, configured to determine, according to a frequency of a user accessing a resource on each date and each time period, One or more message push date prompting segments sent by the user for the message push; the preference resource selection module is configured to select the user that matches the user from the resource set in the case that the current date timely belongs to the message push date and the timely question segment One or more resources of the preference feature; and a message push module for providing message pushes about the one or more resources in the operating system.

According to a third aspect of the present invention, a method for providing user personalized resource message push is provided, comprising the steps of: utilizing a predefined training model for one or more based on a user, a resource, and a historical feature of the user for message push feedback; Forecasting to provide a score reflecting whether each of the one or more resources is suitable for message push to the user; and providing one or more resources in the operating system based on the score Message push.

According to a fourth aspect of the present invention, there is provided an apparatus for providing user personalized resource message push, comprising: a message push prediction module, configured to utilize a predefined feature according to a user, a resource, and a historical feature of the user for message push feedback The training model predicts one or more resources to give a score reflecting whether each of the one or more resources is suitable for message push to the user; and a message push module for, based on the score, Provides message pushes about one or more resources in the operating system.

According to a fifth aspect of the invention there is provided computer program comprising computer readable code, when said computer readable code is run on a computing device, causing said computing device to perform said providing User personalized resource message push method.

According to a sixth aspect of the invention, a computer readable medium is proposed, wherein the computer program described above is stored.

The present invention provides the above method and apparatus for providing user personalized resource message push. According to the embodiment of the present invention, one or more message push dates suitable for the message push of the user may be determined according to the frequency of the user accessing the resources on each date and each time period, and the current date is timely. In the case where the message push date is a timely segment, one or more resources that match the user's preference characteristics are selected from the resource set, and one or more resources are provided in the operating system. Message push. Thereby, the date and time period suitable for message push to the user can be accurately selected for message push, the probability of disturbing the user is greatly reduced, and personalized resource message push can be performed for the user's preference feature, and the message is improved. Push accuracy and success rate. According to an optional embodiment of the present invention, one or more selected ones may be selected according to a user, a resource, and a historical feature of the user's push feedback for the message, and an online feature of the user's current feedback for the message. The resources are predicted to give a score reflecting whether each of the selected one or more resources is suitable for the user to push the message, and the message is pushed according to the score, thereby further improving the accuracy of the message push. And success rate.

The above description is only an overview of the technical solutions of the present invention, and the above-described and other objects, features and advantages of the present invention can be more clearly understood. Specific embodiments of the invention are set forth below.

DRAWINGS

Various other advantages and benefits will become apparent to those skilled in the art from a The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting. Throughout the drawings, the same reference numerals are used to refer to the same parts. In the drawing:

1 is a flow chart of a method of providing user personalized resource message push in accordance with an embodiment of the present invention;

2 is a flow chart of a method of providing user personalized resource message push in accordance with another embodiment of the present invention;

3 is a schematic structural diagram of an apparatus and a server for providing user personalized resource message push according to an embodiment of the present invention;

4 is a schematic structural diagram of an apparatus for providing user personalized resource message push and a user terminal according to another embodiment of the present invention;

Figure 5 shows schematically a block diagram of a computing device for performing the method according to the invention;

Fig. 6 schematically shows a storage unit for holding or carrying program code implementing the method according to the invention.

Specific embodiment

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While the embodiments of the present invention have been shown in the drawings, the embodiments Rather, these embodiments are provided so that this disclosure will be more fully understood and the scope of the disclosure will be fully disclosed.

According to a first aspect of the present invention, there is provided a method of providing user personalized resource message push. According to an embodiment of the present invention, optionally, the resource may be, for example, a game, music, video, shopping information, etc., and the message push may be in the form of, for example, a pop-up window. In the following description, the principle of the present invention will be described by taking a game pop-up as an example, but this is only to facilitate the reader to understand the principle of the present invention, and is not intended to limit the scope of the present invention. Those skilled in the art will appreciate that various forms of message push can be made for various resources, and such implementations are within the scope of the present invention.

FIG. 1 illustrates a flow diagram of a method 100 of providing user personalized resource message push in accordance with an embodiment of the present invention.

As shown in FIG. 1, the method 100 begins in step S110. In step S110, one or more message push dates suitable for message push to the user are determined according to the frequency of the user accessing the resources on each date and each time period. And time period.

After step S110, step S130 is performed, in which, if the current date and time period belongs to the message push date and time period, one or more resources that match the user's preference characteristics are selected from the resource set.

Taking the game as an example, in step S130, if the current date and time period belong to one or more message push dates and time periods suitable for the user to push the message, the user may be from the full game set according to the user's Preference feature primary tour The play collection selects one or more games that match the user's preferred characteristics.

According to an embodiment of the invention, the user's preference features may include one or more of the following characteristics: the type of the user's preferred resource, the theme, the picture style. In this example, it may be the type, theme, picture style, etc. of the game that the user prefers.

Subsequently, step S150 can be performed in which message pushes regarding one or more resources are provided in the operating system. Still taking a game pop-up as an example, a pop-up window for one or more games can be provided to the user in the operating system for promotion.

Optionally, in an embodiment of the present invention, before step S150, the method 100 may further include the following steps: according to users, resources, and historical features of the user's push feedback for the message, and the user currently pushing the message. An online feature of the feedback, predicting the selected one or more resources using a predefined training model to give a score reflecting whether each of the selected one or more resources is suitable for message push to the user . In the case where the above steps are performed, in step S150, message push regarding one or more resources may be provided in the operating system according to the score.

2 shows a flow diagram of a method 200 of providing user personalized resource message push in accordance with another embodiment of the present invention.

As shown in FIG. 2, the method 200 begins in step S210. In step S210, one or more resources are predicted by using a predefined training model according to the user, the resource, and the historical features of the user for the message push feedback. A score is given that reflects whether each of the one or more resources is suitable for message push to the user.

After step S210, step S230 is performed, wherein message pushing about one or more resources is provided in the operating system according to the score. Taking a game pop-up as an example, a pop-up window for one or more games can be provided to the user in the operating system for promotion.

Optionally, in accordance with an embodiment of the present invention, before performing step S210 and step S230, the method 200 may further include the following steps: determining, according to the frequency of the user accessing the resource on each date and each time period, determining that the pair is suitable The one or more message push dates and time periods that the user pushes the message, and in the case that the current date and time period belongs to the message push date and time period, selecting the matching of the user's preference characteristics from the resource set One or more resources.

Optionally, the foregoing step S210 further predicts the one or more resources according to an online feature that the user currently feeds back the message.

In the embodiment shown in Figures 1 and 2, the personalized pop-up window should be pushed at the appropriate time to avoid excessively disturbing the user, enhancing the user experience while increasing the click-through rate. Since the user active data is relatively sparse, in one embodiment of the present invention, the pop-up timing calculation may, for example, select a time period during which the user is accustomed to playing the game in a time window of 7*24 days (i.e., nearly 6 months). Considering that the user's behavior has certain regularity on the same day of different weeks, the timing is defined as "the day of the week", and a specific implementation algorithm is as follows:

First, for example, divide the day into 24 time periods (0 to 23, where 0 means 00:00 to 01:00 and so on) by hour, and calculate the user's day of the week in the specified time window. The frequency of browsing the game website during the time period, such as 12 points on Monday, is 1_12. Then predict the user's pop-up time this week can be converted into: the probability of the user clicking at any time in the case of a given day of the week, that is, the conditional probability:

P(hour|day)=P(hour,day)/P(day)=#(hour,day)/#(day);

Where #(hour,day) indicates the frequency of visits by the user identifier (mid) during a certain time of day of the week, and #(day) indicates the frequency of mid visits in the day of the week, so that the mid-term can be calculated by the Bayesian formula. The timing of the pop-up window on the day of the week.

Then, you can count the number of users accessing the game website in each time period, and arrange the 24 time periods on Sunday according to the number of access users from high to low.

Then, for each mid, the time period in which the game website is browsed in the time window, firstly, the mid-personalization time calculated according to the above step 1 is arranged from high to low, and then the ascending order of the time period is ensured, such as user A's Candidate time The first order of the segments is: 2_20, 2_19, 6_21, 7_23, then the time period of 2_19 is removed, and the time period of the end user A is 2_20, 6_21, 7_23.

According to an embodiment of the present invention, optionally, the method of FIG. 1 and FIG. 2 may further include the steps of: if the determined message push date and the number of time segments are less than the first threshold, The plurality of dates and time periods with the highest frequency of the user accessing the resource are supplemented with a message push date and a time period suitable for the message push of the user, so that the number of the message push date and the time period is equal to the first threshold.

For example, suppose the first threshold is 6, and if the candidate period of the mid is less than 6, the multiple dates and time periods with the highest frequency of accessing the resources of other users are sequentially added (ensure the ascending order of the time segments) until 6 period;

Thus, each mid has a maximum of six candidate time segments. For example, user A has candidate time segments of 1_11, 6_13, 3_15, 1_16, 2_18, and 6_21; by default, when user A accesses, only the current time is in the above. In the six time periods, the pop-up time is satisfied; in addition, in order to increase the recall rate, the time period can be pushed forward for 30 minutes, followed by 30 minutes, and the user 11's 11 time period can be expanded to 10:30. ~12:30.

The operation of the above steps achieves a higher accuracy of user behavior prediction at a lower computational cost, and the prediction efficiency is high. It should be noted that the specific examples of the above steps are only one of the ways to implement the steps, and those skilled in the art can completely adopt other algorithms and operations to achieve the same purpose, as long as the user can be on each date and time. The frequency at which the segment accesses the resource determines the one or more message push dates and time periods suitable for the user to push the message.

The operation of the above steps achieves a higher accuracy of user behavior prediction at a lower computational cost, and the prediction efficiency is high. It should be noted that the specific examples of the above steps are only one of the ways to implement the steps, and those skilled in the art can completely adopt other algorithms and operations to achieve the same purpose, as long as the user can be on each date and time. The frequency at which the segment accesses the resource determines the one or more message push dates and time periods suitable for the user to push the message.

One embodiment of the present invention takes a game as an example. In the above step, in the case that the current date and time period belong to one or more message push dates and time periods suitable for message push to the user, the game may be from the full game set. And selecting one or more games that match the user's preference characteristics according to the user's preference characteristics. According to an embodiment of the invention, the user's preference features may include one or more of the following characteristics: the type of the user's preferred resource, the theme, the picture style. In this example, it may be the type, theme, picture style, etc. of the game that the user prefers.

According to an embodiment of the invention, the historical feature of the user comprises one or more of the following features: a basic feature of the user, a preference feature of the user, a behavioral feature of the user, wherein: the basic feature of the user comprises one of the following features Or multiple: gender, age, occupation; the user's preference characteristics include one or more of the following characteristics: the type of the user's preferred resource, the theme, the picture style; the user's behavior characteristics include one or more of the following characteristics: The situation in which the user accesses the resource running page, the user accesses the resource website, the user accesses the resource payment page, and the user accesses the resource forum.

According to an embodiment of the invention, the historical feature of the resource comprises a basic feature and/or a statistical feature of the resource, wherein: the basic feature of the resource comprises one or more of the following characteristics: a type of the resource, a theme, a picture style, a main role; The statistical characteristics of the resource include one or more of the following characteristics: average click rate of the resource, average starting amount, and average search amount.

According to an embodiment of the present invention, the history feature of the user for message push feedback includes one or more of the following features: a case where the user clicks on the message push, a case where the user does not click the message push, and a case where the user clicks the message after the push is registered.

According to an embodiment of the invention, the feature value (0 or 1) is based on the positive and negative examples. The training data includes two types of data: a positive example (the eigenvalue is 1) and a counterexample (the eigenvalue is 0). Here, a session accessed by a user is used as a window, and the game pop-up window is still taken as an example, such as 2014-05-01. User A clicks on game popup G1, user B clicks on game popup G2, user C does not click on game popup G1, and user A does not click on game popup G2, then divides the four behavior data into two positive examples and Two counterexamples:

Table 1

The game pop-up window is still taken as an example in Table 2 below, showing some examples of the user's historical features, historical features of the resources, and historical features of the user's feedback on the message push to help the reader better understand the principles of the present invention, but Those skilled in the art will appreciate that the scope of the present invention is not limited thereto, and the principles of the present invention are applicable to various historical features.

Table 2

According to an embodiment of the present invention, an acquisition method of the above various features is exemplarily given below: a basic feature of the user, for example, may be obtained based on the user's online behavior and registration information; the user's preference feature may be, for example, The user's online behavior and registration information are obtained; the user's behavior characteristics can be obtained, for example, based on the user's search and browsing data, and after the relevant feature data is marked, the user's accurate behavior data can be obtained; the basic characteristics of the resource, The resource entity can be obtained by crawling and manually marking the resource entity, and the statistical characteristics of the resource can be obtained by calculating the startup amount and the search amount of the game through the daily search log and the cloud search log; The historical feature of the message push feedback can be obtained through the feedback log pushed by the resource message, and the user can know whether the user clicks on the resource or not and the behavior data after the click.

Optionally, in an embodiment of the present invention, the method 100 / method 200 may further include the steps of: updating historical features of the user, historical features of the resource, and the same according to the same or different time periods. Describe the historical characteristics of the user's feedback on the message.

According to an embodiment of the present invention, the historical features of the user, the historical features of the resources, and the historical features of the user for message push feedback may be collectively referred to as offline features. For example, the offline feature set can be built every other month. The time attributes of the feature are different. The time window for constructing the feature is, for example, 2014-05-01.

- the user's basic characteristics, for example, can be updated every three months;

- the user's preference feature, the window time may be, for example, 30 days, and the user's preference characteristics are calculated according to the browsing data between the user between 2014-04-01 and 2014-04-30;

- The behavioral characteristics of the user, the window time can be, for example, 15 days, and the feature data of the game official website, post bar, startup page, etc. that the user has visited between 2014-04-15~2014-04-30 is calculated;

- the basic characteristics of the resource, for example, may not be updated after the collection is completed, or updated regularly;

- The statistical characteristics of the resource, the window time can be, for example, 7 days, the average click rate of the statistical resources between 2014-04-24 and 2014-04-30, the average starting amount, and the average retrieval amount;

- For the historical characteristics of the message push feedback, the window time can be, for example, 15 days, and record the user's operation behavior on the game pop-up window within the window time. For example, a user A clicks on the pop-up window of the game G1 in 2014-04-28. , but did not register the game to record; did not click the game G2 pop-up window on 2014-04-29; clicked the game G2 pop-up window on 2014-04-30, the same Register the game at the time; this feedback feature is:

table 3

All of the above are linear features. In order to mine the nonlinear relationship of features, the features can be combined with each other to construct the intersection relationship between the features. Therefore, the feedback features are combined to form a new combined feature. The above feedback adds a new combination feature:

User's historical characteristics of message push feedback	Eigenvalues
3 days ago, the game G1&&2 days ago did not click the game G2	1
Click the game G1&&1 day ago to click the game G3 3 days ago	1
Click the game G1&&1 day ago to click the game G3 and register	1
Didn't click the game 2 days ago G2&&1 day ago click game G3	1
Didn't click the game G2&&1 days ago 2 days ago, click the game G3 and register	1

Table 4

According to an embodiment of the present invention, the online feature of the user's current push feedback for the message is obtained by combining the historical characteristics of the user's feedback for the message and the characteristics of the message push feedback in the user's current online state. The online feature refers to the feature attribute that the user needs to obtain online; in the training data construction, the online feature set can be generated by offline data. Taking the behavior of user A as an example, set 2014-05-01 as the prediction reference point. In the pop-up window of the day, user A clicks on the game pop-up window G1, does not click the game pop-up window G2, and its historical behavior is: 3 Click the game G1 days ago, did not click the game G2 2 days ago, click the game G3 1 day ago, then through the two-two combination, the new online features are as follows:

User's current online characteristics for message push feedback	Eigenvalues
3 days ago, click the game G1&& online click game G1	1
3 days ago, click the game G1&& online without clicking the game G2	1
2 days ago, there is no click game G2&& online click game G1	1
2 days ago, there is no click game G2&& online without clicking game G2	1
1 day ago click game G3&& online click game G1	1
1 day ago click game G3&& online without clicking game G2	1

table 5

Optionally, in an embodiment of the present invention, the operation of feature word expansion may also be included. The feature word expansion is mainly for the user's historical access or search behavior, and the semantic expansion of the search term and the visited web page title is used as a part of the user feature. For example, it may include the following steps:

(11). Training word vector. After the news corpus is segmented, a representation learning can be performed on each word, for example, using a multi-layer neural network, that is, learning another representation of each word, the representation being represented by a real vector of fixed dimensions. The word vector captures the semantic relationship between deeper words. It can be trained, for example, using the open source project word2vec, which uses continuous bag-of-words. The input corpus includes news corpus and user search long words. The processing includes: segmenting and arranging the news corpus (deleting punctuation, modal words, etc.), each segment as a row of training input; and a single search term segmentation as a row of training input data.

(12). Combination. The user's search term or web page title may contain a plurality of short words after the word segmentation, and the word vectors corresponding to each short word are added and multiplied by the normalization coefficient to obtain the word vector of the entire search word or web page title.

(13). Find the most similar words. Based on the word vector representation calculated in step 1, the other two words most similar to the short words after the search word segmentation are searched; then the other two words most similar to the combined search term vector are searched.

(14). Add features. The word of step 3 is added as an extended feature to the user behavior feature, and its weight is set to w/2, and w is the weight of the source word.

In the method 100, according to the operations of the above steps (which may or may not include the above optional steps), the historical features of the user, the historical features of the resources, the historical features of the user's feedback on the message, and the current user's The online feature of message push feedback. Thereafter, one or more resources may be predicted using a predefined training model in accordance with the features described above to give a score reflecting whether each of the one or more resources is suitable for message push to the user.

In the method 200, according to the operations of the above steps (which may or may not include the above optional steps), the historical features of the user, the historical features of the resources, and the historical features of the user's feedback on the message may be obtained. Thereafter, one or more resources may be predicted using a predefined training model in accordance with the features described above to give a score reflecting whether each of the one or more resources is suitable for message push to the user. Further, the one or more resources may be predicted by combining the online features of the user's current feedback for the message.

In one embodiment of the invention, the predefined training model is an L1-logistic regression model. Input Features Using the above-described offline features and online features, the L1-logistic regression model was used to screen for features.

Using one or more of the resources (as an example of a game) to make predictions using the predefined training model described above, a score can be given that reflects whether each of the one or more games is suitable for message push to the user. If the score of the obtained game G1 is, for example, 0.9, the score of the game G2 is, for example, 0.85, the score of the game G3 is, for example, 0.7, and the like. In this case, in step S230 of method 100 of method 100 / step S230 of method 200, message pushes regarding one or more resources may be provided in the operating system based on the scores described above. In one embodiment, only message feeds may be provided regarding one resource with the highest score (game G1 in the above example), message pushes for multiple resources with the highest score may be provided, and thresholds may also be set, for scores Messages above this threshold are pushed by the message.

According to an embodiment of the present invention, using the above model prediction, the probability that each user click (mid) click message is pushed is represented by a probability value of [0-1], and then it is unnecessary to perform a relatively low mid for some predicted values. Message push. In order to automatically control the number of mid-day impacts, while avoiding invalid message pushes to interfere with users, the algorithm of automatically pop-up users is used to achieve the above purposes:

(21). Discretization of predicted probability values, which is divided into 1000 interval segments: that is, each predicted value is accurate to 0.001, then each probability value between 0-1 is an interval segment;

(22). Statistics the number of mid in the above interval segments, such as 0.500: 21000, indicating that the predicted number is 0.5 and the mid number is 21000;

(23). Calculate the cumulative distribution function table of the predicted values based on the number of mids in each of the section segments calculated in step 22. Rcdf(p) =n indicates that there are n mid-numbers whose predicted value is greater than or equal to p;

(24). Assuming that the number of mid-ranges to be affected is m, it is necessary to calculate the threshold p, and the user whose predicted value is above the threshold p satisfies the condition, and exerts a pop-up effect on it. With the cumulative function distribution table of predicted values, the p value can be conveniently calculated. It is only necessary to find the maximum value of rcdf(p) greater than or equal to m in the distribution table, which is the threshold p, ie argmax(rcdf(p)> =m);

For example, in the cumulative distribution function in the table below, to get 23000 mid, according to the above search rule, the probability p should be 0.874.

p	n
...	...
0.876	20000
0.875	22101
0.874	23201
0.873	26001
...	...

Table 6

(25). According to the above step 4 and each mid predicted value s, it is determined whether the mid is for message push. If s>=q, the message is pushed, and if s<q, the message is not pushed.

In accordance with another aspect of the present invention, in accordance with the method 100 described above, the present invention also provides an apparatus 600 for providing user personalized resource message push. FIG. 3 is a schematic structural diagram of an apparatus 600 for providing user personalized resource message push and a server 300 according to an embodiment of the present invention.

As shown in FIG. 3, the apparatus 600 mainly includes a message pushing timing determining module 610, a preference resource selecting module 620, and a message pushing module 630. According to an embodiment of the present invention, the message pushing timing determining module 610 is configured to determine one or more message pushing dates and time periods suitable for message pushing the user according to the frequency of the user accessing the resources on each date and each time period; The resource selection module 620 is configured to select one or more resources that meet the user's preference characteristics from the resource set if the current date and time period belong to the message push date and time period; the message pushing module 630 is configured to operate Message pushes about one or more resources are provided in the system.

According to an embodiment of the present invention, optionally, the resource may be, for example, a game, music, video, shopping information, etc., and the message push may be in the form of, for example, a pop-up window. In the following description, the principle of the present invention will be described by taking a game pop-up as an example, but this is only to facilitate the reader to understand the principle of the present invention, and is not intended to limit the scope of the present invention. Those skilled in the art will appreciate that various forms of message push can be made for various resources, and such implementations are within the scope of the present invention.

First, the message push timing determining module 610 determines one or more message push dates and time periods suitable for the user to perform message push according to the frequency of the user accessing the resources on each date and each time period.

In the case that the current date and time period belongs to the message push date and time period, the preference resource selection module 620 selects one or more resources from the resource set that match the user's preference characteristics.

The message push module 630 can provide message pushes about one or more resources in an operating system. Still taking the game pop-up as an example, the message push module 630 can provide a pop-up window for one or more games to the user in the operating system for promotion.

Optionally, in an embodiment of the present invention, the apparatus 600 may further include an optional module, a message push prediction module 640 (not shown in FIG. 3), in which the message pushing module 630 is operating. One or more resources are available in the system Before the message is pushed, the message push prediction module 640 can use the predefined training model to select the selected one based on the user, the resource, and the historical characteristics of the user's feedback for the message, and the online feature of the user's current feedback for the message. Or multiple resources are predicted to give a score reflecting whether each of the selected one or more resources is suitable for message push to the user. Subsequently, the message push module 630 can provide message pushes about one or more resources in the operating system based on the scores.

In accordance with yet another aspect of the present invention, in accordance with the method 200 described above, the present invention also provides an apparatus 400 for providing user personalized resource message push. FIG. 4 is a schematic structural diagram of an apparatus 400 for providing user personalized resource message push and a user terminal 500 according to an embodiment of the present invention.

As shown in FIG. 4, the apparatus 400 mainly includes a message push prediction module 410 and a message pushing module 420. According to an embodiment of the present invention, the message push prediction module 410 is configured to predict one or more resources by using a predefined training model according to the user, the resource, and the historical feature of the user for the message push feedback, to provide a reflection. Determining whether each of the one or more resources is suitable for a message push for the user; the message push module 420 is configured to provide a message push for one or more resources in the operating system according to the score.

According to an embodiment of the present invention, optionally, the resource may be, for example, a game, music, video, shopping information, etc., and the message push may be in the form of, for example, a pop-up window. In the following description, the principle of the present invention will be described by taking a game pop-up as an example, but this is only to facilitate the reader to understand the principle of the present invention, and is not intended to limit the scope of the present invention. Those skilled in the art will appreciate that various forms of message push can be made for various resources, and such implementations are within the scope of the present invention. Optionally, the one or more resources are further predicted in conjunction with an online feature currently being fed back by the user for the message.

First, the message push prediction module 410 can predict one or more resources by using a predefined training model according to the user, the resource, and the historical features of the user for the message push feedback to give a reflection in the one or more resources. Whether each resource is suitable for the score of the user to push the message. Optionally, the one or more resources are further predicted in conjunction with an online feature currently being fed back by the user for the message.

Subsequently, the message push module 420 can provide message pushes about one or more resources in the operating system based on the scores. Taking the game pop-up window as an example, the message push module 420 can provide a pop-up window for one or more games to the user in the operating system for promotion.

Optionally, in an embodiment of the present invention, the apparatus 400 may further include an optional module, a message push timing determining module 430 (not shown in FIG. 4), for The frequency of accessing resources in each time period determines one or more message push dates and time periods suitable for message push to the user.

Optionally, in an embodiment of the present invention shown in FIG. 4, the apparatus 400 may further include an optional module, the preference resource selection module 440 (not shown in FIG. 4), for When the current date and time period belong to the message push date and time period, one or more resources that match the user's preference characteristics are selected from the resource set.

In the embodiment shown in Figures 3 and 4, the personalized pop-up window should be pushed at the appropriate time to avoid excessively disturbing the user, enhancing the user experience while increasing the click-through rate. Since the user active data is relatively sparse, in one embodiment of the present invention, the calculation of the pop-up timing of the message push timing determination module 610/message push timing determination module 430 can be selected, for example, at 7*24 days (ie, nearly 6). The time window in which the user is used to playing the game in the time window. Considering that the user's behavior has certain regularity on the same day of different weeks, the timing is defined as "the day of the week", and a specific implementation algorithm is as follows:

First, for example, divide the day into 24 time periods (0 to 23, where 0 means 00:00 to 01:00 and so on) by hour, and calculate the user's day of the week in the specified time window. The frequency of browsing the game website during the time period, such as 12 points on Monday, is 1_12. Then predict the user's pop-up time this week can be converted into: in the case of calculating the given day of the week, the user is in office The probability of when to click, that is, the conditional probability:

P(hour|day)=P(hour,day)/P(day)=#(hour,day)/#(day);

Where #(hour,day) indicates the frequency of visits by the user identifier (mid) during a certain time of day of the week, and #(day) indicates the frequency of mid visits in the day of the week, so that the mid-term can be calculated by the Bayesian formula. The timing of the pop-up window on the day of the week.

Then, you can count the number of users accessing the game website in each time period, and arrange the 24 time periods on Sunday according to the number of access users from high to low.

After that, for each mid, the time period in which the game website is browsed in the time window, firstly, the mid-personalization time calculated according to the above operation is arranged from high to low, and then the ascending order of the time period is ensured, for example, the candidate of user A. The first time of the time period is: 2_20, 2_19, 6_21, 7_23, then the time period of 2_19 is removed, and the time period of the end user A is 2_20, 6_21, 7_23.

According to the embodiment shown in FIG. 3/FIG. 4, optionally, in the case that the determined number of message push dates and time periods is less than the first threshold, the message push timing determination module 610/message push timing is determined. The module 430 supplements the plurality of dates and time periods with the highest frequency of accessing the resources by other users as the message push date and time period suitable for the message push of the user, so that the number of the message push date and the time period is equal to the first Threshold.

For example, suppose the first threshold is 6, and if the candidate period of the mid is less than 6, the multiple dates and time periods with the highest frequency of accessing the resources of other users are sequentially added (ensure the ascending order of the time segments) until 6 period;

Thus, each mid has a maximum of six candidate time segments. For example, user A has candidate time segments of 1_11, 6_13, 3_15, 1_16, 2_18, and 6_21; by default, when user A accesses, only the current time is in the above. In the six time periods, the pop-up time is satisfied; in addition, in order to increase the recall rate, the time period can be pushed forward for 30 minutes, followed by 30 minutes, and the user 11's 11 time period can be expanded to 10:30. ~12:30.

The operation of the message push timing determination module 610 / the message push timing determination module 430 achieves a higher accuracy rate of user behavior prediction at a lower computational cost, and the prediction efficiency is high. It should be noted that the above specific example is only one of the modes of implementing the message push timing determining module 610 / the message pushing timing determining module 430. Those skilled in the art can completely adopt other algorithms and operations to achieve the same purpose, as long as It is possible to determine one or more message push dates and time periods suitable for the user to push the message according to the frequency of the user accessing the resources on each date and each time period.

In one embodiment of the present invention, in the case of a game, in the case where the current date and time period belong to one or more message push dates and time periods suitable for message push to the user, the preferred resource selection module 620/preference The resource selection module 440 can select one or more games that match the user's preferred characteristics from the full game set based on the user's preferred features.

According to an embodiment of the invention, the user's preference features may include one or more of the following characteristics: the type of the user's preferred resource, the theme, the picture style. In this example, it may be the type, theme, picture style, etc. of the game that the user prefers.

According to an embodiment of the invention, the historical feature of the user comprises one or more of the following features: a basic feature of the user, a preference feature of the user, a behavioral feature of the user, wherein: the basic feature of the user comprises one of the following features Or multiple: gender, age, occupation; the user's preference characteristics include one or more of the following characteristics: the type of the user's preferred resource, the theme, the picture style; the user's behavior characteristics include one or more of the following characteristics: The situation in which the user accesses the resource running page, the user accesses the resource website, the user accesses the resource payment page, and the user accesses the resource forum.

According to an embodiment of the invention, the historical feature of the resource comprises a basic feature and/or a statistical feature of the resource, wherein: the basic feature of the resource comprises one or more of the following characteristics: a type of the resource, a theme, a picture style, a main role; The statistical characteristics of the resource include one or more of the following characteristics: average click rate of the resource, average starting amount, and average search amount.

According to an embodiment of the invention, the historical characteristics of the user for message push feedback include one or more of the following features: The situation in which the user clicks on the message push, the user does not click on the message push, and the user clicks on the message to push and register.

According to an embodiment of the invention, the feature value (0 or 1) is based on the positive and negative examples. The training data includes two types of data: a positive example (the eigenvalue is 1) and a counterexample (the eigenvalue is 0). Here, a session accessed by a user is used as a window, and the game pop-up window is still taken as an example, such as 2014-05-01. User A clicks on game popup G1, user B clicks on game popup G2, user C does not click on game popup G1, and user A does not click on game popup G2, then divides the four behavior data into two positive examples and Two counterexamples are shown in Table 1 above.

In the foregoing Table 2, the game pop-up window is taken as an example, and some examples of the historical features of the user, the historical features of the resources, and the historical features of the user for the feedback of the message are shown to help the reader better understand the principle of the present invention, but Those skilled in the art will appreciate that the scope of the present invention is not limited thereto, and the principles of the present invention are applicable to various historical features.

According to an embodiment of the present invention, optionally, the manner in which the message push prediction module 640/message push prediction module 410 in FIG. 3/FIG. 4 obtains the above various features is exemplarily given: a basic feature of the user, The message push prediction module 640 / the message push prediction module 410 can be obtained, for example, based on the user's online behavior and registration information; the user's preference feature, the message push prediction module 640 / the message push prediction module 410 can be based, for example, on the user's online behavior And the registration information is obtained; the message behavior prediction feature of the user, the message push prediction module 640 / the message push prediction module 410 can be obtained, for example, based on the user's search and browsing data, and after the related feature data is marked, the user's accuracy can be obtained. Behavior data; a basic feature of the resource, the message push prediction module 640 / the message push prediction module 410 may label the resource entity by crawl crawling and manual labeling, thereby obtaining; statistical characteristics of the resource, the message push prediction module 640/message push prediction module 410 can pass daily search logs The cloud scans the log, calculates the startup amount and the search amount of the game to obtain the historical feature of the message push feedback, and the message push prediction module 640/message push prediction module 410 can learn the user through the feedback log pushed by the resource message. Obtained by clicking on the resource or not and the behavior data after the click.

Optionally, in an embodiment of the present invention shown in FIG. 3, the apparatus 300 may further include an optional module-history feature update module 650 (not shown in FIG. 3) for using the same or different The time period updates the historical characteristics of the user, the historical characteristics of the resource, and the historical characteristics of the user for feedback of the message push.

Also optionally, in an embodiment of the present invention shown in FIG. 4, the apparatus 400 may further include an optional module-history feature update module 450 (not shown in FIG. 4) for using the same or The historical characteristics of the user, the historical characteristics of the resource, and the historical characteristics of the user for feedback of the message are updated in different time periods.

According to an embodiment of the present invention, the historical features of the user, the historical features of the resources, and the historical features of the user for message push feedback may be collectively referred to as offline features. For example, the offline feature set can be built every other month. The time attributes of the feature are different. The time window for constructing the feature is, for example, 2014-05-01.

- the user's basic characteristics, for example, can be updated every three months;

- the user's preference feature, the window time may be, for example, 30 days, and the user's preference characteristics are calculated according to the browsing data between the user between 2014-04-01 and 2014-04-30;

- The behavioral characteristics of the user, the window time can be, for example, 15 days, and the feature data of the game official website, post bar, startup page, etc. that the user has visited between 2014-04-15~2014-04-30 is calculated;

- the basic characteristics of the resource, for example, may not be updated after the collection is completed, or updated regularly;

- The statistical characteristics of the resource, the window time can be, for example, 7 days, the average click rate of the statistical resources between 2014-04-24 and 2014-04-30, the average starting amount, and the average retrieval amount;

- For the historical characteristics of the message push feedback, the window time can be, for example, 15 days, and record the user's operation behavior on the game pop-up window within the window time. For example, a user A clicks on the pop-up window of the game G1 in 2014-04-28. But not registered The game is recorded; there is no pop-up window of the game G2 on 2014-04-29; the game G2 pop-up window is clicked on 2014-04-30, and the game is registered at the same time; the feedback characteristics are as shown in Table 3 above.

All of the above are linear features. In order to mine the nonlinear relationship of features, the features can be combined with each other to construct the cross relationship between the features. Therefore, the feedback features are combined to form a new combined feature. Table 4 above shows:

According to the embodiment shown in FIG. 3 and FIG. 4, the message push prediction module 640/message push prediction module 410 combines the historical features of the user for the message push feedback and the characteristics of the user's current online state for the message push feedback. To obtain the online characteristics of the user's current feedback on the message. The online feature refers to the feature attribute that the user needs to obtain online; in the training data construction, the online feature set can be generated by offline data. Taking the behavior of user A as an example, set 2014-05-01 as the prediction reference point. In the pop-up window of the day, user A clicks on the game pop-up window G1, does not click the game pop-up window G2, and its historical behavior is: 3 Click the game G1 a day ago, click the game G2 2 days ago, click the game G3 1 day ago, then add the online features through the two-two combination, as shown in Table 5 above.

Optionally, in an embodiment of the present invention, the operation of feature word expansion may also be included. The feature word expansion is mainly for the user history access or search behavior, and the search term, the accessed web page title, and the like are semantically expanded as part of the user feature, which may include, for example, the aforementioned steps (11)-(14).

According to the above operation of the message push prediction module 640, the historical features of the user, the historical features of the resources, the historical features of the user's feedback on the message, and the online features of the user's current feedback on the message can be obtained. Thereafter, the message push prediction module 640 can predict one or more resources using a predefined training model according to the above features to provide a message reflecting whether each of the one or more resources is suitable for the user. Push score.

According to the above operation of the message push prediction module 410, the historical features of the user, the historical features of the resources, and the historical features of the user's feedback on the message can be obtained. Thereafter, the message push prediction module 410 can predict one or more resources using a predefined training model according to the above features to provide a message reflecting whether each of the one or more resources is suitable for the user. Push score. Optionally, the one or more resources are further predicted in conjunction with an online feature currently being fed back by the user for the message.

In one embodiment of the invention, the predefined training model is an L1-logistic regression model. Input Features Using the above-described offline and online features, L1-logistic regression models were used to screen for features:

The message push prediction module 640 / message push prediction module 410 predicts one or more resources (for example, a game) using the above-described predefined training model, and may give a reflection reflecting each of the one or more games Whether the game is suitable for the score of the user to push the message. If the score of the obtained game G1 is, for example, 0.9, the score of the game G2 is, for example, 0.85, the score of the game G3 is, for example, 0.7, and the like. In this case, the message push module 630/message push module 420 can provide message pushes about one or more resources in the operating system based on the scores described above. In one embodiment, the message push module 630/message push module 420 may only provide message pushes regarding one resource with the highest score (game G1 in the above example), and may provide a plurality of resources with the highest score. The message push can also set a threshold, and the message is pushed for resources whose score is higher than the threshold.

According to an embodiment of the present invention, using the above model prediction, the probability that each user click (mid) click message is pushed is represented by a probability value of [0-1], and then it is unnecessary to perform a relatively low mid for some predicted values. Message push. In order to automatically control the number of mid-day impacts while avoiding invalid message pushes to interfere with the user, the algorithm for automatically popping up the user is automatically used to achieve the above purpose: as shown in (21)-(25) above, it will not be repeated here.

The present invention provides the above-described method 100 and apparatus 600 for providing user personalized resource message push. According to the embodiment shown in FIG. 1 and FIG. 3, one or more message push dates and time periods suitable for message pushing to the user may be determined according to the frequency of the user accessing the resources on each date and each time period. The current date and time period belong to the message push date and In the case of a time period, one or more resources that match the user's preferred characteristics are selected from the set of resources and message pushes about one or more resources are provided in the operating system. Thereby, the date and time period suitable for message push to the user can be accurately selected for message push, the probability of disturbing the user is greatly reduced, and personalized resource message push can be performed for the user's preference feature, and the message is improved. Push accuracy and success rate. According to an optional embodiment of the present invention, one or more selected ones may be selected according to a user, a resource, and a historical feature of the user's push feedback for the message, and an online feature of the user's current feedback for the message. The resources are predicted to give a score reflecting whether each of the selected one or more resources is suitable for the user to push the message, and the message is pushed according to the score, thereby further improving the accuracy of the message push. And success rate.

The present invention provides the above-described method 200 and apparatus 400 for providing user personalized resource message push. According to the embodiment shown in FIG. 2 and FIG. 4, one or more resources can be predicted by using a predefined training model based on the user, the resource, and the historical features of the user for the message push feedback. A score reflecting whether each of the one or more resources is suitable for message push to the user, and providing a message push for one or more resources in the operating system based on the score. Thereby, personalized resource message push can be performed for the user's preference feature, which improves the accuracy and success rate of message push. According to an optional embodiment of the present invention, one or more message push dates and time periods suitable for message pushing to the user may be determined according to the frequency of the user accessing the resources on each date and each time period, and the current date and time are When the time period belongs to the message push date and time period, one or more resources that match the user's preference characteristics are selected from the resource set. Thereby, the date and time period suitable for the message push to the user can be accurately selected for message push, which greatly reduces the probability of disturbing the user.

The methods and apparatus provided herein are not inherently related to any particular computer, virtual system, or other device. Various general purpose systems can also be used with the teaching based on the teachings herein. The structure required to construct such a device is apparent from the above description. Moreover, the invention is not directed to any particular programming language. It is to be understood that the invention may be embodied in a variety of programming language, and the description of the specific language has been described above in order to disclose the preferred embodiments of the invention.

In the description provided herein, numerous specific details are set forth. However, it is understood that the embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of the description.

Similarly, the various features of the invention are sometimes grouped together into a single embodiment, in the above description of the exemplary embodiments of the invention, Figure, or a description of it. However, the method disclosed is not to be interpreted as reflecting the intention that the claimed invention requires more features than those recited in the claims. Rather, as the following claims reflect, inventive aspects lie in less than all features of the single embodiments disclosed herein. Therefore, the claims following the specific embodiments are hereby explicitly incorporated into the embodiments, and each of the claims as a separate embodiment of the invention.

Those skilled in the art will appreciate that the modules in the apparatus of the embodiments can be adaptively changed and placed in one or more different devices than the embodiment. Several of the modules in the embodiments may be combined into one module or unit or component, and further, they may be divided into a plurality of sub-modules or sub-units or sub-components. In addition to the fact that at least some of such features and/or processes or modules are mutually exclusive, any combination of the features disclosed in the specification, including the accompanying claims, the abstract and the drawings, and any methods so disclosed or All processes or units of the device are combined. Each feature disclosed in this specification (including the accompanying claims, the abstract, and the drawings) may be replaced by the alternative features that provide the same, equivalent or similar purpose.

In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features that are included in other embodiments and not in other features, combinations of features of different embodiments are intended to be within the scope of the present invention. Within and shape In different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

The various device embodiments of the present invention may be implemented in hardware, or in a software module running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that a microprocessor or digital signal processor (DSP) may be used in practice to implement some of some or all of the means for providing user personalized resource message push in accordance with embodiments of the present invention or All features. The invention can also be implemented as a device program (e.g., a computer program and a computer program product) for performing some or all of the methods described herein. Such a program implementing the invention may be stored on a computer readable medium or may be in the form of one or more signals. Such signals may be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

For example, Figure 5 schematically illustrates a block diagram of a computing device for performing the method in accordance with the present invention. The computing device conventionally includes a processor 510 and a computer program product or computer readable medium in the form of a memory 520. The memory 520 may be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), an EPROM, a hard disk, or a ROM. Memory 520 has a memory space 530 for program code 531 for performing any of the method steps described above. For example, storage space 530 for program code may include various program code 531 for implementing various steps in the above methods, respectively. The program code can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks. Such computer program products are typically portable or fixed storage units as described with reference to FIG. The storage unit may have storage segments, storage spaces, and the like that are similarly arranged to memory 520 in the computing device of FIG. The program code can be compressed, for example, in an appropriate form. In general, the storage unit comprises computer readable code 531 ' for performing the steps of the method according to the invention, ie code that can be read by a processor such as 510, which when executed by the computing device causes the calculation The device performs the various steps in the methods described above.

It is to be noted that the above-described embodiments are illustrative of the invention and are not intended to be limiting, and that the invention may be devised without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as a limitation. The word "comprising" does not exclude the presence of the elements or steps that are not recited in the claims. The word "a" or "an" The invention can be implemented by means of hardware comprising several distinct elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means can be embodied by the same hardware item. The use of the words first, second, and third does not indicate any order. These words can be interpreted as names.

In addition, it should be noted that the language used in the specification has been selected for the purpose of readability and teaching, and is not intended to be construed or limited. Therefore, many modifications and changes will be apparent to those skilled in the art without departing from the scope of the invention. The disclosure of the present invention is intended to be illustrative, and not restrictive, and the scope of the invention is defined by the appended claims.

The present invention is applicable to computer systems/servers that can operate with numerous other general purpose or special purpose computing system environments or configurations. Examples of well-known computing systems, environments, and/or configurations suitable for use with computer systems/servers include, but are not limited to, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, based on Microprocessor systems, set-top boxes, programmable consumer electronics, networked personal computers, small computer systems, mainframe computer systems, and distributed cloud computing technology environments including any of the above, and the like.

The computer system/server can be described in the general context of computer system executable instructions (such as program modules) being executed by a computer system. Generally, program modules may include routines, programs, target programs, components, logic, data structures, and the like that perform particular tasks or implement particular abstract data types. Computer system/server can be in a distributed cloud computing ring In the context of implementation, in a distributed cloud computing environment, tasks are performed by remote processing devices that are linked through a communications network. In a distributed cloud computing environment, program modules may be located on a local or remote computing system storage medium including storage devices.

&quot;an embodiment,&quot; or &quot;an embodiment,&quot; or &quot;an embodiment,&quot; In addition, it is noted that the phrase "in one embodiment" is not necessarily referring to the same embodiment.

The above is only a part of the embodiments of the present invention, and it should be noted that those skilled in the art can also make several improvements and retouchings without departing from the principles of the present invention. It should be considered as the scope of protection of the present invention.

Claims (39)

1. A method for providing user personalized resource message push, comprising the steps of:

   Determining one or more message push dates and time periods suitable for message push to the user according to the frequency of the user accessing the resources on each date and each time period;

   When the current date and time period belong to the message push date and time period, one or more resources that match the user's preference characteristics are selected from the resource set;

   Provides message pushes about one or more resources in the operating system.
2. The method of claim 1 wherein prior to said step of providing a message push for one or more resources in an operating system, said method further comprising the steps of:

   Determining the selected one or more resources based on the user, the resource, and the historical characteristics of the user's feedback on the message, and the online feature of the user's current feedback on the message, using a predefined training model to predict the selected one or more resources Whether each of the one or more resources is suitable for the message push for the user, and

   In the step of providing a message push for one or more resources in an operating system, a message push for one or more resources is provided in the operating system based on the score.
3. A method according to any of claims 1-2, further comprising the steps of:

   When the determined number of message push dates and time periods is less than the first threshold, the plurality of dates and time periods with the highest frequency of accessing the resources of other users are supplemented with the message push date suitable for the message push of the user and The time period is such that the number of the message push date and time period is equal to the first threshold.
4. A method according to any one of claims 1 to 3, further comprising the steps of: updating historical characteristics of the user, historical characteristics of the resource, and feedback of the user to the message in accordance with the same or different time periods Historical characteristics.
5. A method according to any one of claims 1 to 4, wherein the historical characteristics of the user comprise one or more of the following: a basic feature of the user, a preference feature of the user, a behavioral characteristic of the user, wherein:

   The user's basic characteristics include one or more of the following characteristics: gender, age, occupation;

   The user's preference characteristics include one or more of the following characteristics: the type of the user's preferred resource, the theme, and the picture style;

   The behavior characteristics of the user include one or more of the following characteristics: a situation in which the user accesses the resource running page, a situation in which the user accesses the resource website, a situation in which the user accesses the resource payment page, and a situation in which the user accesses the resource forum.
6. The method of any of claims 1-5, wherein the historical characteristics of the resource comprise a base feature and/or a statistical feature of the resource, wherein:

   The basic characteristics of the resource include one or more of the following characteristics: the type of the resource, the theme, the style of the picture, and the main role;

   The statistical characteristics of the resource include one or more of the following characteristics: average click rate of the resource, average starting amount, and average search amount.
7. A method according to any one of claims 1 to 6, wherein the historical characteristics of the user's push feedback for the message include one or more of the following features: a situation in which the user clicks on the message push, a situation in which the user does not click on the message push, The user clicks on the message to post the registration.
8. A method according to any one of claims 1-7, wherein the online feature of the user's current push feedback for the message is by combining the historical characteristics of the user's push feedback for the message and the characteristics of the user's current online state for message push feedback. owned.
9. The method of any of claims 1-8, wherein the predefined training model is an L1-logistic regression model.
10. A device for providing user personalized resource message push, comprising:

    a message pushing timing determining module, configured to determine one or more message pushing dates and time segments suitable for the user to perform message pushing according to the frequency of the user accessing the resources on each date and each time period;

    a preference resource selection module, configured to select one or more resources from the resource set that meet the user's preference characteristics if the current date and time period belong to the message push date and time period;

    A message push module for providing message pushes about one or more resources in an operating system.
11. The apparatus of claim 10 further comprising a message push prediction module for utilizing a predefined training model based on user, resource, and historical characteristics of the user's push feedback for the message, and the online characteristics of the user's current push feedback for the message Predicting the selected one or more resources to give a score reflecting whether each of the selected one or more resources is suitable for message push to the user,

    The message pushing module provides message pushing about one or more resources in the operating system according to the score.
12. The apparatus according to any one of claims 10-11, wherein, in the case that the determined number of message push dates and the number of time periods is less than the first threshold, the message push timing determination module has the highest frequency of accessing resources by other users. The plurality of dates and time periods are supplemented with a message push date and time period suitable for message push to the user, such that the number of message push dates and time periods is equal to the first threshold.
13. The apparatus of any of claims 10-12, further comprising:

    The historical feature update module is configured to update the historical feature of the user, the historical feature of the resource, and the historical feature of the user for feedback of the message according to the same or different time periods.
14. The apparatus of any of claims 10-13, wherein the historical characteristics of the user comprise one or more of the following: a user's basic characteristics, a user's preferred characteristics, and a user's behavioral characteristics, wherein:

    The user's basic characteristics include one or more of the following characteristics: gender, age, occupation;

    The user's preference characteristics include one or more of the following characteristics: the type of the user's preferred resource, the theme, and the picture style;

    The behavior characteristics of the user include one or more of the following characteristics: a situation in which the user accesses the resource running page, a situation in which the user accesses the resource website, a situation in which the user accesses the resource payment page, and a situation in which the user accesses the resource forum.
15. Apparatus according to any of claims 10-14, wherein the historical characteristics of the resource comprise a base feature and/or a statistical feature of the resource, wherein:

    The basic characteristics of the resource include one or more of the following characteristics: the type of the resource, the theme, the style of the picture, and the main role;

    The statistical characteristics of the resource include one or more of the following characteristics: average click rate of the resource, average starting amount, and average search amount.
16. Apparatus according to any one of claims 10-15, wherein the historical characteristics of the user's push feedback for the message include one or more of the following features: the user clicks on the message push, the user does not click on the message push, The user clicks on the message to post the registration.
17. The apparatus according to any one of claims 10-16, wherein the message push prediction module obtains the user's current message push by combining the historical characteristics of the user's push feedback for the message and the characteristics of the message push feedback in the user's current online state. Online characteristics of feedback.
18. The apparatus of any one of claims 10-17, wherein the predefined training model is an L1-logistic regression model.
19. A method for providing user personalized resource message push, comprising the steps of:

    Determining one or more resources using a predefined training model to provide a reflection of whether each of the one or more resources is suitable for the user, resource, and historical characteristics of the user for message push feedback The score that the user pushes the message; and

    Based on the score, message pushes about one or more resources are provided in the operating system.
20. The method of claim 19, wherein prior to each step, the method comprises:

    Determining one or more message push dates and time periods suitable for message push to the user according to the frequency of the user accessing the resources on each date and each time period;

    In a case where the current date and time period belong to the message push date and time period, the one or more resources that match the user's preference characteristics are selected from the resource set.
21. A method according to any of claims 19-20, further comprising the steps of:

    When the determined number of message push dates and time periods is less than the first threshold, the plurality of dates and time periods with the highest frequency of accessing the resources of other users are supplemented with the message push date suitable for the message push of the user and The time period is such that the number of the message push date and time period is equal to the first threshold.
22. The method of any one of claims 19-21, further comprising the step of updating historical characteristics of the user, historical characteristics of the resource, and feedback of the user to the message in accordance with the same or different time periods Historical characteristics.
23. A method according to any one of claims 19-22, wherein the historical characteristics of the user comprise one or more of the following: a user's basic characteristics, a user's preferred characteristics, a user's behavioral characteristics, wherein:

    The user's basic characteristics include one or more of the following characteristics: gender, age, occupation;

    The user's preference characteristics include one or more of the following characteristics: the type of the user's preferred resource, the theme, and the picture style;

    The behavior characteristics of the user include one or more of the following characteristics: a situation in which the user accesses the resource running page, a situation in which the user accesses the resource website, a situation in which the user accesses the resource payment page, and a situation in which the user accesses the resource forum.
24. The method of any of claims 19-23, wherein the historical characteristics of the resource comprise a base feature and/or a statistical feature of the resource, wherein:

    The basic characteristics of the resource include one or more of the following characteristics: the type of the resource, the theme, the style of the picture, and the main role;

    The statistical characteristics of the resource include one or more of the following characteristics: average click rate of the resource, average starting amount, and average search amount.
25. A method according to any one of claims 19 to 24, wherein the historical characteristics of the user's push feedback for the message include one or more of the following features: a situation in which the user clicks on the message push, a situation in which the user does not click on the message push, The user clicks on the message to post the registration.
26. A method according to any one of claims 19 to 25, wherein the one or more resources are further predicted based on an online feature of the user's current push feedback for the message.
27. A method according to any one of claims 19-26, wherein the online feature of the user's current push feedback for the message is by combining the historical characteristics of the user's push feedback for the message and the characteristics of the user's current online state for message push feedback. owned.
28. The method of any one of claims 19-27, wherein the predefined training model is an L1-logistic regression model.
29. A device for providing user personalized resource message push, comprising:

    a message push prediction module, configured to predict one or more resources by using a predefined training model according to a user, a resource, and a historical feature of the user for message push feedback, to provide a reflection in the one or more resources Every resource Is it appropriate to rate the message for this user; and

    A message pushing module, configured to provide a message push about one or more resources in an operating system according to the score.
30. The apparatus of claim 29, further comprising:

    a message pushing timing determining module, configured to determine one or more message pushing dates and time segments suitable for the user to perform message pushing according to the frequency of the user accessing the resources on each date and each time period;

    And a preference resource selection module, configured to select, from the resource set, the one or more resources that meet the user's preference feature if the current date and time period belong to the message push date and time period.
31. The apparatus according to any one of claims 29-30, wherein the message push timing determination module accesses resources of other users if the determined message push date and the number of time periods are less than the first threshold The plurality of dates and time periods with the highest frequency are supplemented with a message push date and time period suitable for message push to the user, such that the number of message push dates and time periods is equal to the first threshold.
32. The apparatus of any of claims 29-31, further comprising:

    The historical feature update module is configured to update the historical feature of the user, the historical feature of the resource, and the historical feature of the user for feedback of the message according to the same or different time periods.
33. The apparatus of any of claims 29-32, wherein the historical characteristics of the user comprise one or more of the following: a user's base feature, a user's preference feature, a user's behavioral characteristics, wherein:

    The user's basic characteristics include one or more of the following characteristics: gender, age, occupation;

    The user's preference characteristics include one or more of the following characteristics: the type of the user's preferred resource, the theme, and the picture style;

    The behavior characteristics of the user include one or more of the following characteristics: a situation in which the user accesses the resource running page, a situation in which the user accesses the resource website, a situation in which the user accesses the resource payment page, and a situation in which the user accesses the resource forum.
34. Apparatus according to any one of claims 29-33, wherein the historical characteristics of the resource comprise a base feature and/or a statistical feature of the resource, wherein:

    The basic characteristics of the resource include one or more of the following characteristics: the type of the resource, the theme, the style of the picture, and the main role;

    The statistical characteristics of the resource include one or more of the following characteristics: average click rate of the resource, average starting amount, and average search amount.
35. A device according to any one of claims 29-34, wherein the historical characteristics of the user for feedback of the message include one or more of the following features: a user clicks on a message push, a user does not click on a message push, The user clicks on the message to post the registration.
36. The apparatus according to any one of claims 29-35, wherein the message push prediction module obtains the user's current message push by combining the historical characteristics of the user's push feedback for the message and the characteristics of the message push feedback in the user's current online state. Online characteristics of feedback.
37. The apparatus of any one of claims 29-36, wherein the predefined training model is an L1-logistic regression model.
38. A computer program comprising computer readable code, when said computer readable code is run on a computing device, causing said computing device to perform the provision of user personalized resources according to any of claims 1-9 A method of message pushing, or causing the computing device to perform the method of providing user personalized resource message pushing according to any one of claims 19-28.
39. A computer readable medium storing the computer program of claim 38.

Images