CN110188282A

CN110188282A - The method and apparatus for determining preference-score

Info

Publication number: CN110188282A
Application number: CN201910472339.4A
Authority: CN
Inventors: 葛莉
Original assignee: Guangzhou Huaduo Network Technology Co Ltd
Current assignee: Guangzhou Huaduo Network Technology Co Ltd
Priority date: 2019-05-31
Filing date: 2019-05-31
Publication date: 2019-08-30

Abstract

The disclosure is directed to a kind of method and apparatus of determining preference-score, belong to Internet technical field.The described method includes: obtaining a variety of behavioral datas with reference to behavior type that target account corresponds to target chatroom；Behavioral data based on every kind with reference to behavior type calculates separately every kind with reference to the corresponding participation score of behavior type；Based on every kind with reference to the corresponding weight of behavior type and the corresponding participation score of every kind of reference behavior type, calculate the corresponding participation score of goal behavior type, as target account to the preference-score of target chatroom, wherein weight is the weight obtained from linear regression model (LRM) trained in advance.Using the disclosure, account can be conveniently calculated to the preference-score of chatroom, and calculated account is more bonded reality to the preference-score of chatroom, relationship that can more between actual response account and chatroom.

Description

The method and apparatus for determining preference-score

Technical field

The disclosure is directed to Internet technical fields, especially with respect to a kind of method and apparatus of determining preference-score.

Background technique

User can enter virtual chatroom by terminal, realize friend-making by way of speech exchange in chatroom Interaction.There is different chat theme and chat style, server can calculate different user and chat target for different chatrooms The preference-score of its room can then calculate all users to the preference total score of target chatroom, finally can be based on this partially Good total score, determines the pouplarity of target chatroom.

During calculating preference-score of the target user to target chatroom, firstly, server needs to count target The behavioral data collection that user generates daily in target chatroom, behavioral data collection is made of the behavioral data of multiple dimensions, multiple The behavioral data of dimension includes speech number, speech total duration, virtual present amount of submitting etc..Then, server needs base In the behavioral data of each dimension, target account is calculated separately to the preference-score of each dimension of target chatroom.Finally, clothes Business device needs for the preference-score of all dimensions to be weighted summation and obtains target user to the preference-score of target chatroom.

In practical applications, it is referred to aforesaid way and calculates different user to the preference-score of target chatroom, then All users can be calculated to the preference total score of target chatroom, the preference total score can be finally based on, determine that target is chatted The pouplarity of its room.

In implementing the present disclosure, inventor discovery the prior art has at least the following problems:

During calculating preference-score of the target user to target chatroom, need the preference-score of all dimensions It is weighted summation and show that target user to the preference-score of target chatroom, therefore just needs to preset each dimension corresponding Weight.The setting of weight is usually what setting person provided by feat of experience.Often, it can be seen that due to weight in subsequent practice What is set is unreasonable, and preference-score is caused to be unable to actual response target user to the preference of target chatroom.

Summary of the invention

In order to overcome the problems, such as that present in the relevant technologies, present disclose provides following technical schemes:

According to the first aspect of the embodiments of the present disclosure, a kind of method of determining preference-score is provided, which comprises

Obtain a variety of behavioral datas with reference to behavior type that target account corresponds to target chatroom；

Behavioral data based on every kind with reference to behavior type calculates separately every kind and participates in reference to behavior type is corresponding Point；

Based on every kind with reference to the corresponding weight of behavior type and the corresponding participation score of every kind of reference behavior type, mesh is calculated The corresponding participation score of behavior type is marked, as the target account to the preference-score of the target chatroom, wherein described Weight is the weight obtained from linear regression model (LRM) trained in advance, and the linear regression model (LRM) trained in advance is with every kind It is sample input, using the corresponding sample preference-score of goal behavior type as sample with reference to the corresponding sample preference-score of behavior type What this output was trained, described every kind is corresponding based on sample account with reference to the corresponding sample preference-score of behavior type What a variety of behavioral datas with reference to behavior type of sample chatroom determined, the corresponding sample preference of the goal behavior type obtains Point be based on sample account correspond to sample chatroom goal behavior type behavioral data determination.

Optionally, described a variety of to include whether the target account has paid close attention to the target chatroom with reference to behavior type Number, the target in the preset duration for the chatroom that house-owner's account, the target account enter in preset duration Account the resident total duration of the target chatroom, in the preset duration target account in the target chatroom In speech total duration, in the preset duration target account sent in the target chatroom word message and The total number of default expression image message, the target account sends out void in the target chatroom in the preset duration The number of quasi- present or the target account invitation and the associated good friend's account of the target account in the preset duration At least two into the number of the target chatroom.

Optionally, described to refer to behavior type including first with reference to behavior type, the first reference behavior type is institute The house-owner's account whether target account has paid close attention to the target chatroom is stated, it is described based on every kind of behavior number with reference to behavior type According to calculating separately every kind with reference to the corresponding participation score of behavior type, comprising:

If the behavioral data is house-owner's account that the target account has paid close attention to the target chatroom, it is determined that institute It states first and is scored at the first default value with reference to the corresponding participation of behavior type；

If the behavioral data is house-owner's account that the target account does not pay close attention to the target chatroom, it is determined that institute It states first and is scored at the second default value with reference to the corresponding participation of behavior type.

Optionally, described to refer to behavior type including second with reference to behavior type, the reference behavior type is the mesh The number for the chatroom that mark account enters in preset duration, it is described based on every kind of behavioral data with reference to behavior type, respectively Every kind is calculated with reference to the corresponding participation score of behavior type, comprising:

If the number for the chatroom that the target account enters in preset duration is less than or equal to third present count Value, it is determined that described second chats with reference to what the corresponding participation of behavior type was scored at that the target account enters in preset duration The difference of the number of its room and the 4th default value；

If the number for the chatroom that the target account enters in preset duration be greater than the third default value and Less than or equal to the 5th default value, it is determined that described second is scored at the 6th present count with reference to the corresponding participation of behavior type Value；

If the number for the chatroom that the target account enters in preset duration is greater than the 5th default value, Determine that described second is scored at the 7th default value with reference to the corresponding participation of behavior type.

Optionally, the behavior type that refers to includes third with reference to behavior type, and the reference behavior type is described In preset duration the target account the target chatroom resident total duration, it is described based on every kind with reference to behavior type Behavioral data calculates separately every kind with reference to the corresponding participation score of behavior type, comprising:

The target user described in the preset duration is determined in the resident total duration of other multiple chatrooms, according to resident The sequence of total duration from high to low uses the target account in the resident total duration of the target chatroom and the target Resident total duration of the family in other multiple chatrooms is ranked up, and obtains the corresponding sequence serial number in different chatrooms；

Based on the corresponding sequence serial number in the difference chatroom, the corresponding sequence serial number in the target chatroom is returned One change processing, the sequence serial number after obtaining normalized；

Based on the sequence serial number after the normalized, calculates the third and participate in reference to behavior type is corresponding Point.

Optionally, described to refer to behavior type including the 4th with reference to behavior type, the reference behavior type is described Speech total duration of the target account in the target chatroom in preset duration, it is described to refer to behavior type based on every kind Behavioral data, calculate separately every kind with reference to the corresponding participation score of behavior type, comprising:

Speech total duration of the target user described in the preset duration in other multiple chatrooms is determined, according to hair Say total duration sequence from high to low to speech total duration of the target account in the target chatroom and the mesh It marks speech total duration of the user in other multiple chatrooms to be ranked up, obtains the corresponding sequence serial number in different chatrooms；

Based on the corresponding sequence serial number in the target chatroom, calculates the described 4th and participate in reference to behavior type is corresponding Point.

Optionally, described to refer to behavior type including the 5th with reference to behavior type, the reference behavior type is described The target account sends the total of word message and default expression image message in the target chatroom in preset duration Item number, the behavioral data based on every kind with reference to behavior type calculate separately every kind and participate in reference to behavior type is corresponding Point, comprising:

Obtain in the preset duration target account is sent in other multiple chatrooms word message and in advance If the total number of facial expression image message, according to the sequence of total number from high to low to the target account in the target chatroom The middle total number for sending word message and default expression image message and the target account are in other multiple chatrooms The total number for sending word message and default expression image message is ranked up, and obtains the corresponding sequence sequence in different chatrooms Number；

Based on the corresponding sequence serial number in the target chatroom, calculates the described 5th and participate in reference to behavior type is corresponding Point.

Optionally, described to refer to behavior type including the 6th with reference to behavior type, the reference behavior type is described The target account sends out the number of virtual present in the target chatroom in preset duration, described to be based on every kind of reference line For the behavioral data of type, every kind is calculated separately with reference to the corresponding participation score of behavior type, comprising:

If the target account sends out the number of virtual present in the target chatroom in the preset duration More than or equal to the 8th default value, it is determined that the described 6th is scored at the 9th present count with reference to the corresponding participation of behavior type Value；

If the target account sends out the number of virtual present in the target chatroom in the preset duration Less than the 8th default value, it is determined that the described 6th is scored at the tenth default value with reference to the corresponding participation of behavior type.

Optionally, described to refer to behavior type including the 7th with reference to behavior type, the reference behavior type is described The target account invites time for entering the target chatroom with the associated good friend's account of the target account in preset duration Number, it is described that the corresponding participation score of every kind of reference behavior type is calculated separately based on every kind of behavioral data with reference to behavior type, Include:

Enter if the target account is invited in the preset duration with the associated good friend's account of the target account The number of the target chatroom is more than or equal to the 11st default value, it is determined that the described 7th is corresponding with reference to behavior type Participation be scored at the 12nd default value；

Enter if the target account is invited in the preset duration with the associated good friend's account of the target account The number of the target chatroom is less than the 11st default value, it is determined that the described 7th with reference to the corresponding ginseng of behavior type Be scored at the 13rd default value.

Optionally, the method also includes:

Determine the target account to the preference-score of multiple chatrooms；

It determines the highest chatroom of preference-score, obtains the corresponding target chatroom in the highest chatroom of the preference-score Type information；

In preset multiple chatrooms to be recommended, chat room-type information and the target chat room-type information are determined The the first preset number target chatroom to be recommended to match；

The terminal logged in the target account sends pushing away for the chatroom information comprising target chatroom to be recommended Recommend message.

Optionally, the method also includes:

Determine multiple accounts respectively to the preference-score of each chatroom；

For each chatroom, determine the multiple account to the preference-score summation of the preference-score of the chatroom；

The second highest chatroom of preset number preference-score summation is determined, as chatroom to be recommended.

According to the second aspect of an embodiment of the present disclosure, a kind of device of determining preference-score is provided, described device includes:

Module is obtained, a variety of behavioral datas with reference to behavior type of target chatroom are corresponded to for obtaining target account；

Computing module, for calculating separately every kind with reference to behavior type based on every kind of behavioral data with reference to behavior type Corresponding participation score；

Grading module, for based on the corresponding weight of every kind of reference behavior type and the corresponding ginseng of every kind of reference behavior type With score, the corresponding participation score of goal behavior type is calculated, as the target account to the preference of the target chatroom Score, wherein the weight is the weight obtained from linear regression model (LRM) trained in advance, and the linear of training in advance returns Returning model is with reference to the corresponding sample preference-score of behavior type using every kind as sample input, with the corresponding sample of goal behavior type This preference-score is that sample output is trained, and described every kind is base with reference to the corresponding sample preference-score of behavior type It is determined in a variety of behavioral datas with reference to behavior type that sample account corresponds to sample chatroom, the goal behavior type pair The sample preference-score answered is that the behavioral data for the goal behavior type that sample chatroom is corresponded to based on sample account is determined.

Optionally, described to refer to behavior type including first with reference to behavior type, the computing module is used for:

Optionally, described to refer to behavior type including second with reference to behavior type, the reference behavior type is the mesh The number for the chatroom that mark account enters in preset duration, the computing module are used for:

Optionally, the behavior type that refers to includes third with reference to behavior type, and the reference behavior type is described In the resident total duration of the target chatroom, the computing module is used for the target account in preset duration:

Optionally, described to refer to behavior type including the 4th with reference to behavior type, the reference behavior type is described Speech total duration of the target account in the target chatroom, the computing module are used in preset duration:

Optionally, described to refer to behavior type including the 5th with reference to behavior type, the reference behavior type is described The target account sends the total of word message and default expression image message in the target chatroom in preset duration Item number, the computing module, is used for:

Optionally, described to refer to behavior type including the 6th with reference to behavior type, the reference behavior type is described The target account sends out the number of virtual present in the target chatroom in preset duration, and the computing module is used for:

Optionally, described to refer to behavior type including the 7th with reference to behavior type, the reference behavior type is described The target account invites time for entering the target chatroom with the associated good friend's account of the target account in preset duration Number, the computing module are used for:

Optionally, described device further include:

Determining module, for determining the target account to the preference-score of multiple chatrooms；Determine preference-score highest Chatroom, obtain the corresponding target chat room-type information in the highest chatroom of the preference-score；It is preset it is multiple to Recommend to determine the first preset number mesh that chat room-type information and the target chat room-type information match in chatroom Mark chatroom to be recommended；

Recommending module, the terminal for logging in the target account send chatting comprising target chatroom to be recommended The recommendation message of its room information.

Optionally, the determining module is also used to determine multiple accounts respectively to the preference-score of each chatroom；For every A chatroom determines the multiple account to the preference-score summation of the preference-score of the chatroom；

The recommending module is also used to determine the second highest chatroom of preset number preference-score summation, as wait push away Recommend chatroom.

According to the third aspect of an embodiment of the present disclosure, a kind of server is provided, the server includes that processor, communication connect Mouth, memory and communication bus, in which:

The processor, the communication interface and the memory complete mutual communication by the communication bus；

The memory, for storing computer program；

The processor, for executing the program stored on the memory, to realize above-mentioned determining preference-score Method.

According to a fourth aspect of embodiments of the present disclosure, a kind of computer readable storage medium is provided, it is described computer-readable Computer program is stored in storage medium, the computer program realizes above-mentioned determining preference-score when being executed by processor Method.

The technical scheme provided by this disclosed embodiment can include the following benefits:

In the embodiments of the present disclosure, a variety of behavioral datas with reference to behavior type of chatroom can be corresponded to according to account, The high linear regression model (LRM) of degree of fitting is established, using the weight in the high linear regression model (LRM) of degree of fitting, refers to behavior to every kind The preference-score or behavioural characteristic data of type are weighted summation process.And then account can be conveniently calculated to merely The preference-score of its room, avoid by virtue of experience provide weight cause occur calculate target account the preference of target chatroom is obtained Divide the problem of inaccuracy.The method that the embodiment of the present disclosure provides can make calculated account to the preference-score of chatroom more Stick on that conjunction is practical, relationship that can more between actual response account and chatroom.

It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not The disclosure can be limited.

Detailed description of the invention

The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure Example, and together with specification for explaining the principles of this disclosure.In the accompanying drawings:

Fig. 1 is a kind of application schematic diagram of the method for determining preference-score shown according to an exemplary embodiment；

Fig. 2 is a kind of flow diagram of the method for determining preference-score shown according to an exemplary embodiment；

Fig. 3 is the structural schematic diagram shown according to an exemplary embodiment with reference to behavior type；

Fig. 4 is a kind of application schematic diagram of the method for determining preference-score shown according to an exemplary embodiment；

Fig. 5 is a kind of structural schematic diagram of the device of determining preference-score shown according to an exemplary embodiment；

Fig. 6 is a kind of structural schematic diagram of server shown according to an exemplary embodiment.

Through the above attached drawings, it has been shown that the specific embodiment of the disclosure will be hereinafter described in more detail.These attached drawings It is not intended to limit the scope of this disclosure concept by any means with verbal description, but is by referring to specific embodiments Those skilled in the art illustrate the concept of the disclosure.

Specific embodiment

Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.

The embodiment of the present disclosure provides a kind of method of determining preference-score, and this method can be realized by server.Service Device can determine that target account to the preference-score of target chatroom, can also determine different accounts to the preference of different chatrooms Score.In turn, different application can be extended based on the preference-score between account and chatroom.Wherein, chatroom can be respectively The chatroom of seed type may include text chat room, voice-enabled chat room, Video chat room etc..

As shown in Figure 1, based on the preference-score between account and chatroom, extending difference by taking voice-enabled chat room as an example and answering With may include (1) chatroom content quality Assessment for classification: being chatted according to the different accounts for entering different phonetic chatroom voice The predilection grade of its room counts different accounts to the distribution situation of the predilection grade of voice-enabled chat room, determines voice-enabled chat room Rank is divided, the quality of voice-enabled chat room can be intuitively observed by division rank belonging to voice-enabled chat room；(2) account preference Chatroom analysis is recommended: predilection grade reflects account to the preference of voice-enabled chat room, it is possible thereby to account recommendation It has generated interest or the voice-enabled chat room there may be interest；(3) potentiality house-owner account is excavated: by predilection grade, It can be from horizontal lower house-owner's account (the creation voice-enabled chat room DAU (Daily Active User, day any active ues quantity) Account) in, identify high-quality house-owner's account；(4) the higher voice-enabled chat room of predilection grade is recommended or is reminded not closing The account of note is paid close attention to: to the higher voice-enabled chat room of account recommendation predilection grade, recommending the account that do not pay close attention to concern preference Score house-owner's account of higher voice-enabled chat room.

An exemplary embodiment of the present disclosure provides a kind of methods of determining preference-score, as shown in Fig. 2, the place of this method Managing process may include following step:

Step S210 obtains a variety of behavioral datas with reference to behavior type that target account corresponds to target chatroom.

In an implementation, during target account enters target chatroom and carries out friend-making interaction, target account can be generated The behavioral data of corresponding target chatroom, behavior data can be the different behavioral datas with reference to behavior type.Server can To count these behavioral datas, the behavioral data based on statistics extracts behavioural characteristic data.

Step S220, the behavioral data based on every kind with reference to behavior type calculate separately every kind and correspond to reference to behavior type Participation score.

In an implementation, preference-score is also believed to behavioural characteristic data, can be to every kind of behavior with reference to behavior type Data are processed, and obtain behavioural characteristic data.

Step S230 is participated in reference to the corresponding weight of behavior type and every kind with reference to behavior type is corresponding based on every kind Point, the corresponding participation score of goal behavior type is calculated, as target account to the preference-score of target chatroom.

Wherein, weight is the weight obtained from linear regression model (LRM) trained in advance, in advance trained linear regression mould Type is to be inputted with reference to the corresponding sample preference-score of behavior type as sample using every kind, is inclined with the corresponding sample of goal behavior type It is scored at what sample output was trained well, every kind is based on sample account with reference to the corresponding sample preference-score of behavior type What a variety of behavioral datas with reference to behavior type that family corresponds to sample chatroom determined, the corresponding sample preference of goal behavior type Score is that the behavioral data for the goal behavior type that sample chatroom is corresponded to based on sample account is determined.

In an implementation, it can be weighted summation with reference to the corresponding participation score of behavior type by every kind, obtains target account Preference-score of the family to target chatroom.The weight used during weighted sum can be returns from the linear of training in advance Return the weight obtained in model.Since weight is got from linear regression model (LRM) trained in advance, which can be more preferable The different accounts of fitting are to the behavioral data between different chatrooms and output is better able to the preference-score of reaction truth.

Trained linear regression model (LRM) is explained below, and obtains the process of weight therein.

Optionally, a variety of to can include but is not limited to whether target account has paid close attention to target chatroom with reference to behavior type The number for the chatroom that house-owner's account, target account enter in preset duration, target account is chatted in target in preset duration The resident total duration of its room, in preset duration speech total duration of the target account in target chatroom, in preset duration Target account sends total number, the mesh in preset duration of word message and default expression image message in target chatroom Mark account sends out the number of virtual present or target account invitation and target account in preset duration in target chatroom Associated good friend's account enters the number of target chatroom.

Wherein, preset duration can be a consecutive days, and a consecutive days are a statistics days.As shown in figure 3, server It can determine whether (action1) target account has paid close attention to house-owner's account of target chatroom, determine (action2) target account The number of the chatroom entered within a consecutive days determines that (action3) target account within a consecutive days is chatted in target The resident total duration of its room determines (action4) when speech of the target account in target chatroom is total within a consecutive days It is long, determine that (action5) target account within a consecutive days sends word message and default expression in target chatroom The total number of image message determines that (action6) target account within a consecutive days sends out virtual gift in target chatroom The number of object determines that (action7) target account within a consecutive days is invited and enters with the associated good friend's account of target account The number of target chatroom.Seven kinds are provided in the embodiments of the present disclosure refers to behavior type, it certainly can also be according to actual needs Increase other behavioral datas for referring to behavior type.

For in preset duration target account can be calculated according to minute, such as in the resident total duration of target chatroom Fruit is resident the minute numerical value that total duration is not integer form, can round up.For the target account in preset duration Speech total duration in target chatroom, can calculate according to minute, if speech total duration is not the minute of integer form Numerical value can round up.

Since target account is in the same consecutive days, different chatrooms can be entered, be referred to aforesaid way, count mesh A variety of behavioral datas with reference to behavior type that mark account is generated when the same consecutive days entering different chatrooms.For example, For consecutive days D1, can count on that target account generates when D1 enters different chatrooms is a variety of with reference to behavior type Behavioral data.Then, target account can be counted when different consecutive days entering different chatrooms with continued reference to aforesaid way The a variety of behavioral datas with reference to behavior type generated.For example, consecutive days D1 can be counted on target account and be entered in D1 The a variety of behavioral datas with reference to behavior type generated when different chatroom.For consecutive days D2, target account can be counted on A variety of behavioral datas with reference to behavior type that family is generated when D2 enters different chatrooms.Repeat aforesaid operations, Zhi Daotong Count a variety of behavioral datas with reference to behavior type that target account is generated when D1 to Dn enters different chatrooms.

A variety of preference-scores with reference to behavior type that a large amount of account corresponds to chatroom constitute the sample of linear regression model (LRM) This input.It is explained below and behavioral data is processed, obtain a variety of participation scores with reference to behavior type and (be referred to as Behavioural characteristic data) mode.This mode can be used for obtaining the process of sample input, can be used for subsequent in calculating mesh Every kind used when marking account to target chatroom is with reference to the corresponding process for participating in score (behavioural characteristic data) of behavior type.

It optionally, include first with reference to behavior type with reference to behavior type, the first reference behavior type is for target account The no house-owner's account for having paid close attention to target chatroom, the behavioral data based on every kind with reference to behavior type, calculates separately every kind of reference If the step of behavior type corresponding participation score may include: that behavioral data is that target account has paid close attention to target chatroom House-owner's account, it is determined that first is scored at the first default value with reference to the corresponding participation of behavior type；If behavioral data is mesh Mark account does not pay close attention to house-owner's account of target chatroom, it is determined that first is scored at second in advance with reference to the corresponding participation of behavior type If numerical value.

In an implementation, server may determine that whether target account has paid close attention to house-owner's account of target chatroom, can be Same day concern, it is paid close attention to before can also be.If target account has paid close attention to house-owner's account of target chatroom, can determine First, with reference to the corresponding participation score feature_action1 of behavior type, remembers feature_action1=1.If target account House-owner's account of target chatroom is not paid close attention at family, then can determine the corresponding participation score feature_ of the first reference behavior type Action1 remembers feature_action1=0.

Optionally, include second with reference to behavior type with reference to behavior type, be target account default with reference to behavior type The number of the chatroom entered in duration, the behavioral data based on every kind with reference to behavior type calculate separately every kind with reference to behavior If type it is corresponding participate in score the step of may include: the chatroom that target account enters in preset duration number it is small In or equal to third default value, it is determined that second is scored at target account when default with reference to the corresponding participation of behavior type The number of the chatroom entered in long and the difference of the 4th default value；If the chatroom that target account enters in preset duration Number be greater than third default value and be less than or equal to the 5th default value, it is determined that second with reference to behavior type it is corresponding Participation is scored at the 6th default value；If it is default that the number for the chatroom that target account enters in preset duration is greater than the 5th Numerical value, it is determined that second is scored at the 7th default value with reference to the corresponding participation of behavior type.

In an implementation, server can count the number for the chatroom that target account enters within a consecutive days.Service Device can achieve the effect that branch mailbox with demarcation interval, and the number for the chatroom that target account enters within a consecutive days is put into and is drawn In by stages, second is obtained with reference to the corresponding participation score feature_action2 of behavior type.If target account is at one The number of the chatroom entered in consecutive days is 1 or 2, then feature_action2=target account can be remembered one The number -1 of the chatroom entered in a consecutive days.If the number for the chatroom that target account enters within a consecutive days is 3 or 4, then it can remember feature_action2=2.If the chatroom that target account enters within a consecutive days Number be greater than 4, then can remember feature_action2=3.

Optionally, include third with reference to behavior type with reference to behavior type, be the mesh in preset duration with reference to behavior type It marks account and every seed ginseng is calculated separately based on the behavioral data of every kind of reference behavior type in the resident total duration of target chatroom The step of examining behavior type corresponding participation score may include: to determine that target user is in other multiple chats in preset duration The resident total duration of room, according to the sequence of resident total duration from high to low to target account target chatroom it is resident total when Resident total duration of the long and target user in other multiple chatrooms is ranked up, and obtains the corresponding sequence in different chatrooms Serial number；Based on the corresponding sequence serial number in different chatrooms, sequence serial number corresponding to target chatroom is normalized, and is obtained Sequence serial number after to normalized；Based on the sequence serial number after normalized, it is corresponding with reference to behavior type to calculate third Participation score.

In an implementation, server can count within consecutive days target account target chatroom it is resident total when It is long, target account can also be counted within the consecutive days in the resident total duration of other multiple chatrooms.According to resident total duration Sequence from high to low to target account target chatroom resident total duration and target account in other multiple chatrooms Resident total duration is ranked up, and obtains the corresponding sequence serial number order_action3 in different chatrooms.It should be noted that tool There is the chatroom of identical resident total duration that can assign identical sequence serial number, therefore, a sequence serial number can correspond to multiple Chatroom.Order_action3 is 1, corresponds to the longest chatroom of resident total duration.In addition, only when target account enters When into chatroom, friend-making interaction just will do it, therefore can filter out the chatroom that resident total duration is 0.

Then, server can determine that maximum order_action3 is denoted as max_order_ in all sequence serial numbers Action3 can be based on max_order_action3, return to the sequence serial number of each chatroom for each chatroom One change is handled, and the sequence serial number after obtained normalized can be denoted as norm_order_action3.If max_ Order_action3 be greater than 1, then norm_order_action3=(max_order_action3-order_action3)/ (max_order_action3-1).If max_order_action3 is equal to 1, norm_order_action3=1.Most Eventually, the sequence serial number after normalized is located in section [0,1].

Finally, server can achieve the effect that branch mailbox with demarcation interval, determines and drawn belonging to norm_order_action3 By stages obtains third with reference to the corresponding participation score feature_action3 of behavior type.If norm_order_ Action3 is located at section [0,0.75], then feature_action3=ceiling (norm_order_action3/0.25), Wherein ceiling () is the function that rounds up.If norm_order_action3 is located at section (0.75,1), then Feature_action3=4.If norm_order_action3=1, feature_action3=5.

Optionally, include the 4th with reference to behavior type with reference to behavior type, be the mesh in preset duration with reference to behavior type Speech total duration of the account in target chatroom is marked, the behavioral data based on every kind with reference to behavior type calculates separately every kind The step of participation score corresponding with reference to behavior type may include: to determine that target user is multiple merely at other in preset duration Speech total duration in its room, according to speech of the sequence to target account in target chatroom of speech total duration from high to low The speech total duration of total duration and target user in other multiple chatrooms is ranked up, and it is corresponding to obtain different chatrooms Sequence serial number；Based on the corresponding sequence serial number in target chatroom, the 4th is calculated with reference to the corresponding participation score of behavior type.

In an implementation, when server can count that speech of the target account in target chatroom is total in a consecutive days It is long, speech total duration of the target account in other multiple chatrooms in a consecutive days can also be counted.When according to making a speech total Long sequence from high to low is to the speech of speech total duration and target user in other multiple chatrooms in target chatroom Total duration is ranked up, and obtains the corresponding sequence serial number order_action4 in different chatrooms.Order_action4 is 1, right Answer the longest chatroom of speech total duration.It, can be without statistics if target user does not make a speech in chatroom.In reality In the application of border, most of account, 70 or more percent account, the number for the chatroom made a speech within a consecutive days Within 3.

Then, server can determine demarcation interval belonging to order_action4 with demarcation interval, obtain the 4th reference The corresponding participation score feature_action4 of behavior type.If order_action4 is equal to 1, feature_ Action4=3.If order_action4 is equal to 2 or 3, feature_action4=2.If order_action4 More than or equal to 4, then feature_action4=1.It does not make a speech in addition, replenishing target user in a consecutive days Chatroom feature_action4 be 0.

Optionally, include the 5th with reference to behavior type with reference to behavior type, be the mesh in preset duration with reference to behavior type Mark account sends the total number of word message and default expression image message in target chatroom, refers to behavior based on every kind The behavioral data of type, calculating separately every kind with reference to the step of behavior type corresponding participations score may include: acquisition pre- If target account sends the total number of word message and default expression image message in other multiple chatrooms in duration, press Word message is sent in target chatroom to target account according to the sequence of total number from high to low and default facial expression image disappears The total number and target account of breath send the total of word message and default expression image message in other multiple chatrooms Item number is ranked up, and obtains the corresponding sequence serial number in different chatrooms；Based on the corresponding sequence serial number in target chatroom, the is calculated Five with reference to the corresponding participation score of behavior type.

In an implementation, server, which can count the target account within a consecutive days and send text in target chatroom, disappears The total number of breath and default expression image message can also count within a consecutive days target account in other multiple chats The total number that word message and default expression image message are sent in room, according to the sequence of total number from high to low to target account Family sent in target chatroom word message and default expression image message total number and target account it is multiple at other Word message is sent in chatroom and the total number of default expression image message is ranked up, and it is corresponding to obtain different chatrooms Sort serial number order_action5.

Then, server can determine demarcation interval belonging to order_action5 with demarcation interval, obtain the 5th reference The corresponding participation score feature_action5 of behavior type.If order_action5 is equal to 1, feature_ Action5=3.If order_action5 is equal to 2 or 3, feature_action5=2.If order_action5 More than or equal to 4, then feature_action5=1.

Optionally, include the 6th with reference to behavior type with reference to behavior type, be the mesh in preset duration with reference to behavior type Mark account sends out the number of virtual present in target chatroom, and the behavioral data based on every kind with reference to behavior type is counted respectively If calculate every kind with reference to the step of behavior type corresponding participation score may include: in preset duration target account in target The number that virtual present is sent out in chatroom is more than or equal to the 8th default value, it is determined that the 6th is corresponding with reference to behavior type Participation be scored at the 9th default value；If target account sends out virtual present in target chatroom in preset duration Number is less than the 8th default value, it is determined that the 6th is scored at the tenth default value with reference to the corresponding participation of behavior type.

In an implementation, server can count the target account within a consecutive days and send out virtual gift in target chatroom The number of object.In statistics one consecutive days of discovery there is the behavior for sending out virtual present in only fraction account.If target account The number that virtual present is sent out at family in target chatroom is greater than 0 time, then feature_action6=1.If target account exists The number that virtual present is sent out in target chatroom is equal to 0 time, then feature_action6=0.

Optionally, include the 7th with reference to behavior type with reference to behavior type, be the mesh in preset duration with reference to behavior type Mark account invites the number for entering target chatroom with the associated good friend's account of target account, based on every kind with reference to behavior type Behavioral data, if the step of calculating separately every kind of participation score corresponding with reference to behavior type may include: in preset duration Interior target account invites the number for entering target chatroom with the associated good friend's account of target account to be more than or equal to the 11st Default value, it is determined that the 7th is scored at the 12nd default value with reference to the corresponding participation of behavior type；If in preset duration Interior target account, which is invited, enters the number of target chatroom less than the 11st default value with the associated good friend's account of target account, Then determine that the 7th is scored at the 13rd default value with reference to the corresponding participation of behavior type.

In an implementation, server can count the target account within a consecutive days and invite and the associated good friend of target account Account enters the number of target chatroom.Statistics discovery one consecutive days in only fraction account exist invite good friend's account into Enter the behavior of chatroom.If target account invites the number for entering target chatroom with the associated good friend's account of target account big In 0 time, then feature_action7=1.If target account invitation enters target with the associated good friend's account of target account and chats The number of its room is equal to 0 time, then feature_action7=0.

After obtaining for training the sample of linear regression model (LRM) to input, it can also obtain for training linear regression mould The sample of type exports.During the output of selected sample, sample output can be selected by target of business revenue.When account is being chatted When the amount of room consumption is higher, it can indicate that the account is higher to the preference of the chatroom.

In sample input, there are a variety of behavioral datas with reference to behavior type that different accounts correspond to different chatrooms, Each pair of account and chatroom can be denoted as a binary group, for example, target account and target chatroom can be used as one Binary group.Early period can acquire different accounts in a variety of behaviors with reference to behavior type of different chatrooms in D1 to Dn consecutive days Data can therefrom determine multiple binary groups.For the binary group that each consecutive days generate, for example, in D1, target account Three chatrooms are entered, then obtain 3 binary groups, including the chatroom target account-A, the chatroom target account-B, target account The family chatroom-C.It can be based on same mode, determined in D1, the binary group of other accounts and other chatrooms composition.With Afterwards, can be within the consecutive days of the preset number after D1, such as in 14 days after D1, i.e. in D2 to D15, determination is It is no to generate the binary group generated in D1 again.It, can be in D2 to D15 for the binary group that target account and target chatroom form It interior each consecutive days, determines whether target account was again introduced into target chatroom, looks over so as to check if target account is again introduced into Chatroom is marked, then is denoted as observation period binary group.

Find, if account sends out virtual present in any chatroom, there is 95 percent account to exist in analysis The resident duration for sending out the chatroom of virtual present all can be more than 1 minute.If will be using business revenue as target, in multiple observation periods In binary group, the resident duration in chatroom is more than 1 minute these target observations phase binary groups can be more valuable.It therefore can be with Calculate the accounting ratio_tuple of effective time binary group.Ratio_tuple is equal to target observations phase binary group in all sights Examine the accounting in phase binary group.Since in a consecutive days, target account can form binary group with different chatrooms, can pass through Aforesaid way determines the corresponding observation period binary group of binary group of target account and different chatrooms composition, and calculates ratio_ tuple.Finally, the maximum value and minimum value in these ratio_tuple can be determined, it is denoted as max_ratio_tuple respectively And min_ratio_tuple.Referring to max_ratio_tuple and min_ratio_tuple, ratio_tuple equal proportion is put Greatly target, target=(ratio_tuple-min_ratio_tuple)/(max_ are exported for the sample of model training Ratio_tuple-min_ratio_tuple) value of × 100, target is located at section [0,100].

To sum up, the behavioural characteristic data feature_action1 to feature_ of each binary group can be obtained Action7, feature_action1 are discrete numerical value to feature_action7, and each behavioural characteristic data are corresponding The value of 7 behavioural characteristic data can be determined as a value vector by 1 value.The behavioural characteristic number of each binary group According to may map to a value vector.Server can count the value vector of a large amount of binary group, a large amount of binary group It can be used as a binary group collection.Server can count the number of the corresponding observation period binary group of binary group, determine target, Filter out value vector sum of the number of observation period binary group more than 1000 (can be adjusted according to actual needs) target.Using 1000 (can be adjusted according to actual needs) or more value vector sum target as input sample Originally and sample is exported, linear regression model (LRM) is trained, trained linear regression model (LRM) is obtained.During training, Linear regression model (LRM) can be initially set up, passes through intersection using target as output sample using value vector as input sample Verifying, assesses the fitting effect of linear regression model (LRM).

After obtaining trained linear regression model (LRM), it can determine weight therein, obtain following preference-score Formula:

Y=w1 × feature_action1+ ...+wi × feature_actioni+ ...+w7 × feature_ Action7 (formula 1)

Wherein, y is preference-score of the target account to target chatroom, and w is the corresponding weight of different reference behavior types, I is the integer between 1 to 7.

As shown in figure 4, in the embodiments of the present disclosure, account can be acquired and correspond to a variety of with reference to behavior type of chatroom Behavioral data extracts the behavioural characteristic data of behavioral data.In this course, it can acquire whether target account has paid close attention to mesh The number for the chatroom that house-owner's account, the target account of mark chatroom enter within a consecutive days, the mesh within a consecutive days Mark account the resident total duration of target chatroom, speech of the target account in target chatroom is total within a consecutive days when Length, target account sends total item of word message and default expression image message in target chatroom within a consecutive days Number, the number that target account sends out virtual present in target chatroom within a consecutive days, the target within a consecutive days Account, which is invited, enters the behavioral datas such as number of target chatroom action1 extremely with the associated good friend's account of target account action7.Pretreatment and characterization can be carried out to behavioral data action1 to action7, obtain behavioural characteristic data Feature_action1 to feature_action7.Then, the training objective that can select model training, establishes machine learning Model.In this course, the training objective of selected model training can be analyzed by data.It can establish linear regression mould Type.Linear regression model (LRM) effect can be assessed.Finally, can use the corresponding scoring of trained linear regression model (LRM) Formula calculates account to the preference-score of chatroom.In this course, it can be extracted from trained linear regression model (LRM) Score formula.Account can be calculated according to scoring formula to the preference-score of chatroom.

Optionally, according to formula 1, the preference-score in section [0,100] can be calculated.It is a large amount of in statistics After account is to the preference-score of chatroom, the distribution situation of preference-score can be determined, may further divide preference level. For example, preference-score less than or equal to 30 binary group accounting be 40%, preference-score be located at section (30,50] binary group Accounting is 30%, preference-score be located at section (50,75] binary group accounting be 20%, scoring be greater than 75% binary group accounting It is 10%.4 preference levels can be divided into, if preference-score is low inclined less than or equal to the preference level of 30 chatroom Good, corresponding preference level is 1 grade.Positioned at section (30,50] the corresponding chatroom of preference-score preference level be it is lower partially Good, corresponding preference level is 2 grades.Positioned at section (50,75] the corresponding chatroom of preference-score preference level be it is medium partially Good, corresponding preference level is 3 grades.The preference level of chatroom of the preference-score greater than 75 is higher preference, corresponding preference level It is 4 grades.According to ready-portioned preference level, it is subsequent the preference-score for calculating the corresponding target chatroom of target account it Afterwards, it can determine the corresponding preference level in target chatroom, be embodied as the operation that chatroom is classified.

Optionally, the method that the embodiment of the present disclosure provides further include: determine that target account obtains the preference of multiple chatrooms Point；It determines the highest chatroom of preference-score, obtains the corresponding target chat room-type information in the highest chatroom of preference-score； In preset multiple chatrooms to be recommended, chat room-type information and target chat room-type information match first is determined Preset number target chatroom to be recommended；The terminal logged in target account sends the chatroom comprising target chatroom to be recommended The recommendation message of information.

In an implementation, target account can be determined to the preference-score of multiple chatrooms, for example, target account is to chatroom The preference-score of A is 10 points, and the preference-score to chatroom B is 8 points, and the preference-score to chatroom C is 3 points.It can determine The highest chatroom of preference-score, for example, determining that the highest chatroom of preference-score is chatroom A.It can be stored in advance and respectively chat The corresponding chat room-type information in its room, the highest chatroom of preference-score (such as chatroom A) available in this way is corresponding to be chatted Its room type information (such as singing type).Then, can be in preset multiple chatrooms to be recommended, determination is equally to sing class First preset number chatroom of type, as target chatroom to be recommended.Type is sung since target account is interested Chatroom, therefore target account can be recommended into target chatroom to be recommended, it can it will include target chatroom to be recommended Chatroom information recommendation message be sent to target account login terminal.

Optionally, other than the above-mentioned way of recommendation, the method that the embodiment of the present disclosure provides can also comprise determining that multiple Account is respectively to the preference-score of each chatroom；For each chatroom, determine multiple accounts to the preference-score of chatroom Preference-score summation；The second highest chatroom of preset number preference-score summation is determined, as chatroom to be recommended.

In an implementation, different accounts can be counted to the preference-score of different chatrooms, it is corresponding to calculate these chatrooms Preference-score summation chooses the second highest chatroom of preset number preference-score summation, and the chatroom selected is that hot topic is chatted Its room is the chatroom that numerous users compare hobby.The chatroom selected can be recommended to other and do not entered this merely The account of its room, or the chatroom selected can also be recommended and enter the chatroom but do not paid close attention to the chatroom Account.

Disclosure another exemplary embodiment provides a kind of device of determining preference-score, as shown in figure 5, the device packet It includes:

Module 510 is obtained, a variety of behavior numbers with reference to behavior type of target chatroom are corresponded to for obtaining target account According to；

Computing module 520, for calculating separately every kind and referring to behavior class based on every kind of behavioral data with reference to behavior type The corresponding participation score of type；

Grading module 530, for corresponding with reference to behavior type with reference to the corresponding weight of behavior type and every kind based on every kind Participation score, the corresponding participation score of goal behavior type is calculated, as the target account to the target chatroom Preference-score, wherein the weight is the weight obtained from linear regression model (LRM) trained in advance, the line trained in advance Property regression model be using every kind with reference to the corresponding sample preference-score of behavior type be sample input, with goal behavior type correspondence Sample preference-score be that sample output is trained, described every kind with reference to the corresponding sample preference-score of behavior type It is a variety of behavioral datas determinations with reference to behavior type that sample chatroom is corresponded to based on sample account, the goal behavior class The corresponding sample preference-score of type is that the behavioral data for the goal behavior type that sample chatroom is corresponded to based on sample account is determined 's.

Optionally, described to refer to behavior type including first with reference to behavior type, the computing module 520 is used for:

Optionally, described to refer to behavior type including second with reference to behavior type, the reference behavior type is the mesh The number for the chatroom that mark account enters in preset duration, the computing module 520 are used for:

Optionally, the behavior type that refers to includes third with reference to behavior type, and the reference behavior type is described In the resident total duration of the target chatroom, the computing module 520 is used for the target account in preset duration:

Optionally, described to refer to behavior type including the 4th with reference to behavior type, the reference behavior type is described Speech total duration of the target account in the target chatroom in preset duration, the computing module 520 are used for:

Optionally, described to refer to behavior type including the 5th with reference to behavior type, the reference behavior type is described The target account sends the total of word message and default expression image message in the target chatroom in preset duration Item number, the computing module 520, is used for:

Optionally, described to refer to behavior type including the 6th with reference to behavior type, the reference behavior type is described The target account sends out the number of virtual present in the target chatroom in preset duration, and the computing module 520 is used In:

Optionally, described to refer to behavior type including the 7th with reference to behavior type, the reference behavior type is described The target account invites time for entering the target chatroom with the associated good friend's account of the target account in preset duration Number, the computing module 520 are used for:

Optionally, described device further include:

About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method Embodiment in be described in detail, no detailed explanation will be given here.

In the embodiments of the present disclosure, a variety of behavioral datas with reference to behavior type of chatroom can be corresponded to according to account, The high linear regression model (LRM) of degree of fitting is established, using the weight in the high linear regression model (LRM) of degree of fitting, refers to behavior to every kind The preference-score or behavioural characteristic data of type are weighted summation process.And then account can be conveniently calculated to merely The preference-score of its room, avoid by virtue of experience provide weight cause occur calculate target account the preference of target chatroom is obtained Divide the problem of inaccuracy.The device that the embodiment of the present disclosure provides can make calculated account to the preference-score of chatroom more Stick on that conjunction is practical, relationship that can more between actual response account and chatroom.

It should be understood that the device of determining preference-score provided by the above embodiment is when determining preference-score, only with The division progress of above-mentioned each functional module can according to need and for example, in practical application by above-mentioned function distribution by not Same functional module is completed, i.e., the internal structure of server is divided into different functional modules, described above complete to complete Portion or partial function.In addition, the method for the device of determining preference-score provided by the above embodiment and determining preference-score is real It applies example and belongs to same design, specific implementation process is detailed in embodiment of the method, and which is not described herein again.

Fig. 6 shows the structural schematic diagram of the server 1900 of one exemplary embodiment of disclosure offer.The server 1900 can generate bigger difference because configuration or performance are different, may include one or more processors (central Processing units, CPU) 1910 and one or more memory 1920.Wherein, it is deposited in the memory 1920 At least one instruction is contained, at least one instruction is loaded by the processor 1910 and executed to realize above-described embodiment institute The method for the determination preference-score stated.

Those skilled in the art will readily occur to its of the disclosure after considering specification and practicing disclosure disclosed herein Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or Person's adaptive change follows the general principles of this disclosure and including the undocumented common knowledge in the art of the disclosure Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are wanted by right It asks and points out.

It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the accompanying claims.

Claims

1. a kind of method of determining preference-score, which is characterized in that the described method includes:

Behavioral data based on every kind with reference to behavior type calculates separately every kind with reference to the corresponding participation score of behavior type；

Based on every kind with reference to the corresponding weight of behavior type and the corresponding participation score of every kind of reference behavior type, target line is calculated For the corresponding participation score of type, as the target account to the preference-score of the target chatroom, wherein the weight It is the weight obtained from linear regression model (LRM) trained in advance, the linear regression model (LRM) trained in advance is referred to every kind The corresponding sample preference-score of behavior type is that sample inputs, is defeated as sample using the corresponding sample preference-score of goal behavior type It is trained out, described every kind is to correspond to sample based on sample account with reference to the corresponding sample preference-score of behavior type What a variety of behavioral datas with reference to behavior type of chatroom determined, the corresponding sample preference-score of the goal behavior type is It is determined based on the behavioral data that sample account corresponds to the goal behavior type of sample chatroom.

2. the method according to claim 1, wherein a variety of reference behavior types include the target account Whether the number of the house-owner account of the target chatroom, chatroom that the target account preset duration in enter has been paid close attention to Mesh, in the preset duration target account in the resident total duration of the target chatroom, in the preset duration Speech total duration of the target account in the target chatroom, in the preset duration target account described Word message and the total number of default expression image message, the target in the preset duration are sent in target chatroom Account sends out the number of virtual present in the target chatroom or the target account is invited in the preset duration Enter at least two in the number of the target chatroom with the associated good friend's account of the target account.

3. according to the method described in claim 2, it is characterized in that, described refer to behavior class including first with reference to behavior type Type, described first with reference to house-owner's account that behavior type is whether the target account has paid close attention to the target chatroom, described Behavioral data based on every kind with reference to behavior type calculates separately every kind with reference to the corresponding participation score of behavior type, comprising:

If the behavioral data is that the target account has paid close attention to house-owner's account of the target chatroom, it is determined that described the One is scored at the first default value with reference to the corresponding participation of behavior type；

If the behavioral data is that the target account does not pay close attention to house-owner's account of the target chatroom, it is determined that described the One is scored at the second default value with reference to the corresponding participation of behavior type.

4. according to the method described in claim 2, it is characterized in that, described refer to behavior class including second with reference to behavior type Type, the number of the chatroom entered in preset duration with reference to behavior type for the target account are described to be based on every kind With reference to the behavioral data of behavior type, every kind is calculated separately with reference to the corresponding participation score of behavior type, comprising:

If the number for the chatroom that the target account enters in preset duration is less than or equal to third default value, Determine that described second is scored at the chatroom that the target account enters in preset duration with reference to the corresponding participation of behavior type Number and the 4th default value difference；

If the number for the chatroom that the target account enters in preset duration is greater than the third default value and is less than Or it is equal to the 5th default value, it is determined that described second is scored at the 6th default value with reference to the corresponding participation of behavior type；

If the number for the chatroom that the target account enters in preset duration is greater than the 5th default value, it is determined that Described second is scored at the 7th default value with reference to the corresponding participation of behavior type.

5. according to the method described in claim 2, it is characterized in that, the behavior type that refers to includes third with reference to behavior class Type, it is described with reference to behavior type be in the preset duration target account the target chatroom it is resident total when It is long, it is described that the corresponding participation score of every kind of reference behavior type is calculated separately based on every kind of behavioral data with reference to behavior type, Include:

The target user described in the preset duration is determined in the resident total duration of other multiple chatrooms, when according to being resident total Long sequence from high to low exists to the target account in the resident total duration of the target chatroom and the target user The resident total duration of other multiple chatrooms is ranked up, and obtains the corresponding sequence serial number in different chatrooms；

Based on the corresponding sequence serial number in the difference chatroom, the corresponding sequence serial number in the target chatroom is normalized Processing, the sequence serial number after obtaining normalized；

Based on the sequence serial number after the normalized, the third is calculated with reference to the corresponding participation score of behavior type.

6. according to the method described in claim 2, it is characterized in that, described refer to behavior class including the 4th with reference to behavior type Type, it is described with reference to behavior type be speech of the target account in the target chatroom is total in the preset duration when It is long, it is described that the corresponding participation score of every kind of reference behavior type is calculated separately based on every kind of behavioral data with reference to behavior type, Include:

Determine speech total duration of the target user described in the preset duration in other multiple chatrooms, it is total according to speech Speech total duration and target use of the sequence of duration from high to low to the target account in the target chatroom Speech total duration of the family in other multiple chatrooms is ranked up, and obtains the corresponding sequence serial number in different chatrooms；

Based on the corresponding sequence serial number in the target chatroom, the described 4th is calculated with reference to the corresponding participation score of behavior type.

7. according to the method described in claim 2, it is characterized in that, described refer to behavior class including the 5th with reference to behavior type Type, described with reference to behavior type is that the target account sends text in the target chatroom and disappears in the preset duration The total number of breath and default expression image message, it is described based on every kind of behavioral data with reference to behavior type, it calculates separately every Kind refers to the corresponding participation score of behavior type, comprising:

It obtains the target account in the preset duration and sends word message and preset table in other multiple chatrooms The total number of feelings image message sends out the target account according to the sequence of total number from high to low in the target chatroom The total number and the target account for sending word message and default expression image message are sent in other multiple chatrooms Word message and the total number of default expression image message are ranked up, and obtain the corresponding sequence serial number in different chatrooms；

Based on the corresponding sequence serial number in the target chatroom, the described 5th is calculated with reference to the corresponding participation score of behavior type.

8. according to the method described in claim 2, it is characterized in that, described refer to behavior class including the 6th with reference to behavior type Type, the reference behavior type are that the target account sends out virtual gift in the target chatroom in the preset duration The number of object, it is described based on every kind of behavioral data with reference to behavior type, every kind is calculated separately with reference to the corresponding ginseng of behavior type With score, comprising:

If the number that the target account sends out virtual present in the target chatroom in the preset duration is greater than Or it is equal to the 8th default value, it is determined that the described 6th is scored at the 9th default value with reference to the corresponding participation of behavior type；

If the number that the target account sends out virtual present in the target chatroom in the preset duration is less than 8th default value, it is determined that the described 6th is scored at the tenth default value with reference to the corresponding participation of behavior type.

9. according to the method described in claim 2, it is characterized in that, described refer to behavior class including the 7th with reference to behavior type Type, it is described to be invited and the associated good friend of the target account with reference to behavior type for the target account described in the preset duration Account enters the number of the target chatroom, described based on every kind of behavioral data with reference to behavior type, calculates separately every kind With reference to the corresponding participation score of behavior type, comprising:

If described in the target account invitation and the associated good friend's account of the target account enter in the preset duration The number of target chatroom is more than or equal to the 11st default value, it is determined that the described 7th with reference to the corresponding ginseng of behavior type Be scored at the 12nd default value；

If described in the target account invitation and the associated good friend's account of the target account enter in the preset duration The number of target chatroom is less than the 11st default value, it is determined that the described 7th participates in reference to behavior type is corresponding It is divided into the 13rd default value.

10. the method according to claim 1, wherein the method also includes:

Determine the target account to the preference-score of multiple chatrooms；

It determines the highest chatroom of preference-score, obtains the corresponding target chat room-type in the highest chatroom of the preference-score Information；

In preset multiple chatrooms to be recommended, chat room-type information and the target chat room-type information phase are determined The the first preset number target chatroom to be recommended matched；

The recommendation that the terminal logged in the target account sends the chatroom information comprising target chatroom to be recommended disappears Breath.

11. the method according to claim 1, wherein the method also includes:

12. a kind of device of determining preference-score, which is characterized in that described device includes:

Computing module, for calculating separately every kind and being corresponded to reference to behavior type based on every kind of behavioral data with reference to behavior type Participation score；

Grading module, for being referred to based on the corresponding weight of every kind of reference behavior type and every kind, behavior type is corresponding to be participated in Point, the corresponding participation score of goal behavior type is calculated, as the target account to the preference-score of the target chatroom, Wherein, the weight is the weight obtained from linear regression model (LRM) trained in advance, the linear regression mould trained in advance Type is to be inputted with reference to the corresponding sample preference-score of behavior type as sample using every kind, is inclined with the corresponding sample of goal behavior type It is scored at what sample output was trained well, described every kind is based on sample with reference to the corresponding sample preference-score of behavior type What a variety of behavioral datas with reference to behavior type of the corresponding sample chatroom of this account determined, the goal behavior type is corresponding Sample preference-score is that the behavioral data for the goal behavior type that sample chatroom is corresponded to based on sample account is determined.

13. a kind of server, which is characterized in that the server includes processor, communication interface, memory and communication bus, Wherein:

The memory, for storing computer program；

The processor, for executing the program stored on the memory, to realize that claim 1-11 is any described Method and step.

14. a kind of computer readable storage medium, which is characterized in that be stored with computer in the computer readable storage medium Program realizes claim 1-11 any method and step when the computer program is executed by processor.