Active opportunity intelligent prediction method, device, equipment and medium

Abstract

The present invention discloses an active timing intelligent prediction method, device, equipment and medium. The method includes: according to the received user historical behavior data and user real-time status data, statistical feature information is obtained and then encoded to obtain an active feature list, according to the model configuration information, a potential active prediction model and an active timing prediction model are respectively constructed to obtain the statistical feature information to be predicted and the active feature list to be predicted, and the received user information to be predicted is processed to obtain the statistical feature information to be predicted and the active feature list to be predicted, and whether it is active is predicted by the active prediction model, and if it is active, the active timing prediction is performed by the active timing prediction model to obtain the active timing prediction result. The present invention belongs to the field of intelligent prediction technology, and predicts whether the user information to be predicted is active by the potential active prediction model, and further predicts the active timing prediction result by the active timing prediction model, and can push messages based on the active timing prediction result, which greatly improves the accuracy of the message push timing prediction.

Description

Technical Field

[0001] The present invention relates to the field of intelligent prediction technology, and in particular to an active opportunity intelligent prediction method, device, equipment and medium. Background Art

[0002] Enterprises can promote their business by pushing messages. In order to increase the click rate of users on the pushed messages and thus improve the user conversion rate, the current conventional technical methods are to preset different dimensional screening conditions, based on the real-time status behavior of users such as user activity, user status updates, and user targeted behaviors, so as to push messages when the trigger conditions are met. However, in some scenarios, such as after the user opens the application in full screen, the message push triggered by judging the real-time status behavior is likely to be directly ignored by the user, thus missing the best time to push messages to the user. Therefore, there is a problem in the existing technical methods that the timing of message push cannot be accurately obtained when pushing messages. Summary of the invention

[0003] The embodiments of the present invention provide a method, device and equipment for intelligently predicting active timing, aiming to solve the problem that the timing of message push cannot be accurately obtained when message push is performed in the prior art methods.

[0004] In a first aspect, an embodiment of the present invention provides an active timing intelligent prediction method, which is applied to a management server, wherein a network connection is established between the management server and a client to realize data information transmission, and the method includes:

[0005] Receiving the user's historical behavior data and the user's real-time status data from the client, and performing statistics on the user's historical behavior data and the user's real-time status data according to a preset periodic statistical rule to obtain corresponding statistical feature information;

[0006] Encoding the statistical feature information according to a preset encoding conversion rule to obtain a corresponding active feature list;

[0007] Constructing a potential activity prediction model and an activity opportunity prediction model according to the preset model configuration information, the statistical feature information and the active feature list;

[0008] If the user information to be predicted is received, the user information to be predicted is processed according to the periodic statistical rule and the encoding conversion rule to obtain corresponding statistical feature information to be predicted and a list of active features to be predicted;

[0009] Predicting the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the potential active prediction model to obtain a prediction result of whether the activity is active;

[0010] If the prediction result is active, active timing prediction is performed on the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the active timing prediction model to obtain an active timing prediction result.

[0011] In a second aspect, an embodiment of the present invention provides an active opportunity intelligent prediction device, which includes:

[0012] A statistical feature information acquisition unit, used to receive the user's historical behavior data and the user's real-time status data from the client, and to perform statistics on the user's historical behavior data and the user's real-time status data according to a preset periodic statistical rule to obtain corresponding statistical feature information;

[0013] An active feature list acquisition unit, configured to encode the statistical feature information according to a preset encoding conversion rule to obtain a corresponding active feature list;

[0014] A model building unit, used to build a potential activity prediction model and an activity opportunity prediction model according to preset model configuration information, the statistical feature information and the active feature list;

[0015] A user information processing unit to be predicted is used to, upon receiving the user information to be predicted, process the user information to be predicted according to the periodic statistical rule and the encoding conversion rule to obtain corresponding statistical feature information to be predicted and a list of active features to be predicted;

[0016] an activity prediction result acquisition unit, configured to predict the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the potential activity prediction model, so as to obtain a prediction result of whether the activity is present;

[0017] The active timing prediction result acquisition unit is used to, if the prediction result is active, perform active timing prediction on the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the active timing prediction model to obtain an active timing prediction result.

[0018] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the active opportunity intelligent prediction method described in the first aspect is implemented.

[0019] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the active opportunity intelligent prediction method described in the first aspect above.

[0020] The embodiments of the present invention provide an active timing intelligent prediction method, device, equipment and medium. According to the received user historical behavior data and user real-time status data, statistical feature information is obtained and then encoded to obtain an active feature list. According to the model configuration information, a potential active prediction model and an active timing prediction model are respectively constructed. The received user information to be predicted is processed to obtain the statistical feature information to be predicted and the active feature list to be predicted. The active prediction model is used to predict whether it is active. If it is active, the active timing prediction model is used to predict the active timing to obtain an active timing prediction result. Through the above method, a potential active prediction model and an active timing prediction model can be constructed based on historical behavior data and real-time status data, respectively. Whether the user information to be predicted is active is predicted by the potential active prediction model, and the active timing prediction result is further predicted by the active timing prediction model. Messages can be pushed based on the active timing prediction result, which greatly improves the accuracy of message push timing prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0022] Figure 1 A schematic diagram of a process flow of an active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of an application scenario of the active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0024] Figure 3 A schematic diagram of a sub-process of an active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0025] Figure 4 A schematic diagram of another sub-process of the active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0026] Figure 5 A schematic diagram of another sub-process of the active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0027] Figure 6 A schematic diagram of another sub-process of the active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0028] Figure 7 A schematic diagram of the latter sub-process of the active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0029] Figure 8 A schematic diagram of a subsequent sub-process of the active opportunity intelligent prediction method provided by an embodiment of the present invention;

[0030] Fig. 9 A schematic block diagram of an active opportunity intelligent prediction device provided by an embodiment of the present invention;

[0031] Fig.10 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0034] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0035] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0036] See also Figure 1 and Figure 2 , Figure 1 A schematic diagram of a process flow of an active opportunity intelligent prediction method provided by an embodiment of the present invention; Figure 2Schematic diagram of application scenario of the active timing intelligent prediction method provided in an embodiment of the present invention; the active timing intelligent prediction method is applied in a management server 10, and the active timing intelligent prediction method is executed by an application software installed in the management server 10, and a network connection is established between the management server 10 and the client 20 to realize the transmission of data information; the management server 10 is a server end for executing the active timing intelligent prediction method to realize the prediction of the active timing of the user, and the client 20 is a terminal device for communicating with the management server 10, such as a desktop computer, a laptop computer, a tablet computer or a mobile phone. Figure 2 The figure only illustrates the establishment of a network connection between one client 20 and the management server 10. In actual application, the management server 10 can simultaneously establish network connections with multiple clients 20 to achieve data information transmission.

[0037] like Figure 1 As shown, the method includes steps S110 to S160.

[0038] S110, receiving the user's historical behavior data and the user's real-time status data from the client, and performing statistics on the user's historical behavior data and the user's real-time status data according to a preset periodic statistical rule to obtain corresponding statistical feature information.

[0039] Receive the user's historical behavior data and the user's real-time status data from the client, and perform statistics on the user's historical behavior data and the user's real-time status data according to the preset periodic statistical rules to obtain corresponding statistical feature information. Among them, the statistical feature information includes list statistical information, basic statistical information and increase statistical information. Specifically, the client can report the historical behavior data of each user obtained to the management server through SDK (Software Development Kit). In the process of using the application software, the user will generate interactive events with the background server. The user's historical behavior data may include the corresponding behavior data of multiple users in the past period of time. The user's behavior data is the data generated by various usage behaviors during the user's use of the application. By obtaining the behavior data of each user, the user's usage habits of the application can be understood. The user's historical behavior data includes data such as the user's active information, push message click information, etc. The active information contains information such as the user's online time and offline time each time, and the push message click information contains information such as the user's click time for each push message.

[0040] The user's real-time status data is the real-time active status obtained by judging the user's current behavior data within a period of time. For example, if the user is active more than 2 times a day and the active time on the current day is greater than 30 minutes, the user's real-time active status is determined to be active.

[0041] The user's historical behavior data and the user's real-time status data can be counted according to periodic statistical rules to obtain statistical information including list statistical information, basic statistical information and increase statistical information.

[0042] In one embodiment, if Figure 3 As shown, step S110 includes sub-steps S111, S112 and S113.

[0043] S111. According to the statistical items in the periodic statistical rule, the behavior data of each user in the user historical behavior data and the real-time active status of each user in the user real-time status data are respectively tabulated and statistics are obtained to obtain the list statistical information corresponding to each of the users.

[0044] Specifically, the periodic statistical rules contain multiple statistical items, and the behavior data of each user and the real-time active status of each user can be listed and counted according to the statistical items. For example, the statistical items can be the number of active times, the total number of active durations, the type of active status, the active duration in the first time period, the active duration in the second time period, the active duration in the third time period, etc. Then, the behavior data and real-time active status of a user can be listed and counted in units of days, and the corresponding statistical values ​​of the user in the statistical items such as the number of active times, the number of message clicks, the total number of active durations, the type of active status, the active duration in the first time period, the active duration in the second time period, and the active duration in the third time period can be obtained every day, thereby forming list statistical information. Then, the statistics of each user's information can form corresponding list statistical information.

[0045] S112: Perform periodic statistics on the list statistical information of each user according to the statistical period in the periodic statistical rule to obtain corresponding basic statistical information.

[0046] The periodic statistical rules also include statistical cycles, and the list statistical information of each user can be periodically counted according to the statistical cycle. Since the list statistical information is counted in days to obtain the statistical value of each statistical item, in order to reflect the behavioral characteristics of the user over a period of time, the statistical cycle can be set for periodic statistics. If the statistical cycle is set to a week, a month, or a longer period, if the statistical cycle is set to a week, the number of active days, active times, total active time, message clicks, and other information of the user in the current week can be counted, and the number of active days, active times, total active time, message clicks, and other information of the user in the previous week can also be counted. The statistical information corresponding to multiple cycles can be counted to obtain the basic statistical information corresponding to the user.

[0047] S113: Perform comparative statistics based on the statistical information of adjacent statistical periods in the basic statistical information to obtain corresponding increase statistical information.

[0048] On the basis of basic statistical information, comparative statistics can also be performed on statistical information of adjacent statistical periods to obtain the increase / decrease of each indicator item in the basic statistical information based on the previous statistical period, thereby obtaining the increase statistical information corresponding to each indicator item.

[0049] S120: Encode the statistical feature information according to a preset encoding conversion rule to obtain a corresponding active feature list.

[0050] The statistical feature information is encoded according to a preset encoding conversion rule to obtain a corresponding active feature list. The active feature list includes a frequency feature list and a duration feature list. The list statistical information in the statistical feature information can be encoded according to the encoding conversion rule to obtain a corresponding active feature list, that is, the numerical value in the list statistical information is converted into an encoding form associated with the feature, thereby realizing structured storage of information.

[0051] In one embodiment, if Figure 4 As shown, step S120 includes sub-steps S121 and S122.

[0052] S121. Convert the frequency value in the statistical feature information according to the conversion formula in the encoding conversion rule to obtain a corresponding frequency feature list.

[0053] Specifically, the statistical feature information includes multiple frequency values, and each value can be converted separately through the conversion formula in the encoding conversion rule, so as to convert the integer type frequency value into a decimal value to obtain a corresponding frequency feature list.

[0054] For example, the conversion formula can be expressed using formula (1):

[0055]

[0056] Where x is the number of conversions required, and C is the calculated decimal value.

[0057] S122: Perform feature mapping processing on the active duration in the statistical feature information according to the mapping information table in the encoding conversion rule to obtain a corresponding duration feature list.

[0058] The encoding conversion rules also include a mapping information table. The active duration of each time period in the list statistics is counted in minutes, and the values ​​are all integer values. The total active duration in the statistical feature information is also counted in groups, and the values ​​are integer values. The active duration of each time period and the total active duration can be feature mapped according to the mapping information table, such as mapping the values ​​of active duration between 30 and 60 to "0.1", and mapping the values ​​between 60 and 180 to "0.3". Obtaining the mapping value obtained by the mapping process corresponding to each active duration can obtain the corresponding duration feature list.

[0059] S130: construct a potential activity prediction model and an activity opportunity prediction model according to the preset model configuration information, the statistical feature information and the active feature list.

[0060] The potential active prediction model and the active timing prediction model are constructed according to the preset model configuration information, the statistical feature information and the active feature list. The management server pre-stores the model configuration information, and the model can be constructed according to the model configuration information and the statistical feature information and the active feature list obtained in the above steps, thereby respectively constructing the potential active prediction model and the active timing prediction model.

[0061] In one embodiment, if Figure 5 As shown, step S130 includes sub-steps S131 and S132.

[0062] S131. Construct a corresponding potential activity prediction model according to the model configuration information, the statistical features in the statistical feature information, and the list features in the active feature list.

[0063] Specifically, a potential active prediction model can be constructed based on the model configuration information, the statistical features in the statistical feature information, and the list features in the active feature list. Specifically, the model configuration information includes the configuration parameters of the fully connected hidden layer, the configuration parameters of the weight layer, etc. The active prediction model template can be configured based on the model configuration information, the statistical feature information, and the active feature list. At this time, the active prediction model template is only a template framework. The active prediction training of the active prediction model template is performed through the specific content contained in the statistical feature information and the active feature list, and finally the potential active prediction model is constructed.

[0064] In one embodiment, if Figure 6 As shown, step S131 includes sub-steps S1311, S1312 and S1313.

[0065] S1311. Obtain an active prediction model template according to the model configuration information, the number of the statistical features, and the number of the list features.

[0066] Specifically, the model configuration information includes configuration parameters of the fully connected hidden layer, configuration parameters of the weight layer, etc. For example, the model configuration information can be configured with a fully connected hidden layer consisting of three fully connected layers, and the configuration parameters of the three fully connected layers are 1024, 512, and 256, respectively. Then, according to the configuration parameters, three fully connected layers with dimensions of 1024, 512, and 256 can be configured respectively, and the fully connected layers are concatenated into a fully connected hidden layer. At the same time, the corresponding first weight layer can be configured according to the configuration parameters of the weight layer. The number of statistical features can be obtained, and the first input node corresponding to the number of statistical features can be generated, that is, the number of first input nodes is equal to the number of statistical features, and the first input node is directly connected to the fully connected layer of the highest dimension. The number of features in the list features can be obtained, and the first input matrix is ​​configured according to the number of features in the list features. The horizontal dimension of the first input matrix is ​​equal to the number of features in the list features, and its vertical dimension is equal to the number of data values ​​contained in each feature. The first input matrix is ​​concatenated with the configured first weight layer and then connected to the fully connected layer of the highest dimension. In the actual data processing process, the values ​​of each column in the first input matrix are added and averaged and then weightedly calculated with the configured first weight layer, and then the calculated values ​​obtained by the weighted calculation are input into the corresponding fully connected layer.

[0067] S1312: Calculate the statistical features and list features of each user according to the activity prediction model template to obtain activity prediction template output information of each user.

[0068] The statistical features and list features of each user can be input into the active prediction model template respectively. The statistical features of the user are input through the first input node in the active prediction model template, and the list features of the user are input through the first input matrix. Finally, the active prediction template input information of the user is input from the fully connected layer with the lowest dimension in the active prediction model template. For example, in this embodiment, the fully connected layer with the lowest dimension is located at the top layer of the fully connected hidden layer, and its dimension is 256, so the corresponding 256-dimensional active prediction template output information can be input. Each user's information input active prediction model template can input a corresponding active prediction template output information.

[0069] S1313: construct a potential activity prediction model based on binary classification according to the activity prediction model template, the activity prediction template output information of each user, and the real-time activity status of each user in the activity feature list.

[0070] The active feature list contains the real-time active status of each user. In the specific processing process, the real-time active status can be recorded in the form of a numerical label. For example, if the real-time active status is active, the numerical label is recorded as "1", and if the real-time active status is inactive, the numerical label is recorded as "0". The corresponding numerical label is stored in the active feature list as a label feature. The label feature is not used when constructing the active prediction model template. The numerical label corresponding to the real-time active status of each user is the classification target in the binary tree model construction process. A binary tree model based on the binary classification problem can be constructed according to the output information of the active prediction template of each user and the real-time active status of each user in the active feature list, that is, the 256-dimensional active prediction template output information is converted into a two-dimensional output result of "active or not". After combining the binary tree model with the active prediction model template, a potential active prediction model based on binary classification can be obtained. In the actual calculation process, the probability of active and inactive can be output by using an activation function (such as a sogmoid function), that is, the two-dimensional output result of the binary tree model contains the probability value of active and the probability value of inactive.

[0071] S132: construct a corresponding active timing prediction model according to the model configuration information, the active duration of each time period in the statistical feature information, and the list features in the active feature list.

[0072] Specifically, the active timing prediction model can be constructed based on the model configuration information, the active duration of each time period in the statistical feature information, and the list features in the active feature list. Specifically, the model configuration information includes the configuration parameters of the fully connected hidden layer, the configuration parameters of the weight layer, etc. The timing prediction model template can be configured based on the model configuration information, the statistical feature information, and the active feature list. At this time, the timing prediction model template is only a template framework. The timing prediction training of the timing prediction model template is performed through the specific content contained in the statistical feature information and the active feature list, and finally the active timing prediction model is constructed.

[0073] In one embodiment, if Figure 7 As shown, step S132 includes sub-steps S1321, S1322 and S1323.

[0074] S1321. Obtain a timing prediction model template according to the model configuration information, the number of time periods in the statistical feature information, and the number of list features.

[0075] Specifically, the model configuration information includes configuration parameters of the fully connected hidden layer, configuration parameters of the weight layer, etc. For example, the model configuration information can be configured with a fully connected hidden layer consisting of three fully connected layers, and the configuration parameters of the three fully connected layers are 1024, 512, and 256, respectively. Then, three fully connected layers with dimensions of 1024, 512, and 256 can be configured according to the configuration parameters, and the fully connected layers are concatenated into a fully connected hidden layer. At the same time, the corresponding second weight layer can be configured according to the configuration parameters of the weight layer. The number of statistical time periods in the statistical feature information can be obtained, and a second input node corresponding to the number of time periods is generated, that is, the number of the second input nodes is equal to the number of statistical time periods in the statistical feature information, and the second input node is directly connected to the fully connected layer of the highest dimension, so the number of features in the duration feature list in the list feature can be obtained, and the second input matrix is ​​configured according to the number of features in the list feature, the horizontal dimension of the second input matrix is ​​equal to the number of features in the duration feature list, and the vertical dimension is equal to the number of data values ​​contained in each feature, and the second input matrix is ​​spliced ​​with the configured second weight layer and connected to the fully connected layer of the highest dimension. In the actual data processing process, the values ​​of each column in the second input matrix are added and averaged, and then weighted calculated with the configured second weight layer, and then the calculated values ​​obtained by the weighted calculation are input into the corresponding fully connected layer.

[0076] S1322. Calculate the active duration and list features of each user in each time period according to the timing prediction model template to obtain timing prediction template output information of each user.

[0077] The specific calculation process is the same as the process of obtaining the active prediction template output information, which will not be described in detail here. Each user's information input timing prediction model template can input a corresponding timing prediction template output information.

[0078] S1323: construct an active timing prediction model based on multiple classifications according to the timing prediction model template, the timing prediction template output information of each user, and the active time period of each user in the active feature list.

[0079] The active time period of each user can be obtained based on the active feature list, and the active time period of each user is the classification target in the model construction process. A binary tree model based on multi-classification problems can be constructed based on the output information of the timing prediction template of each user and the active time period of each user, that is, the 256-dimensional active prediction template output information is converted into a multi-dimensional output result. The dimension of the multi-dimensional output result is equal to the number of categories of the time period. After combining the binary tree model with the timing prediction model template, an active timing prediction model based on multi-classification can be obtained. In the actual calculation process, the probability corresponding to each time period can be output by using an activation function (such as a softmax function), that is, the multi-dimensional output result of the binary tree model contains the probability value corresponding to each time period.

[0080] S140: If the user information to be predicted is received, the user information to be predicted is processed according to the periodic statistical rule and the encoding conversion rule to obtain corresponding statistical feature information to be predicted and a list of active features to be predicted.

[0081] The management server may receive user information to be predicted, which is information for predicting the possible active time period of the user. The user information to be predicted includes the behavior data of one or more users in a period of time. In this embodiment, the user information to be predicted includes only one user. The behavior data of the user in the user information to be predicted may be periodically counted according to the periodic statistical rules to obtain the statistical feature information to be predicted, and the statistical feature information to be predicted may be converted according to the encoding conversion rules to obtain the active feature information to be predicted. The specific processing method is the same as the processing in step S110 and step S120.

[0082] S150: predicting the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the potential activity prediction model to obtain an activity prediction result.

[0083] The statistical information to be predicted and the list of active features to be predicted can be predicted according to the potential activity prediction model, that is, the corresponding information of the user to be predicted is input through the first input node and the first input matrix of the activity prediction model, and the two-dimensional output result is calculated, and the classification with the largest probability value is obtained from the two-dimensional output result as the prediction result of whether it is active. If the two-dimensional output result of a user to be predicted is: active 75%, inactive 33%, then the prediction result of the user to be predicted is active.

[0084] S160: If the prediction result is active, perform active timing prediction on the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the active timing prediction model to obtain an active timing prediction result.

[0085] If the prediction result of the user to be predicted is active, the active timing prediction can be performed on the statistical feature information to be predicted and the active feature list to be predicted according to the active timing prediction model to obtain the active timing prediction result corresponding to the user. That is, only if the user is predicted to be active can the active timing prediction be continued. If the user's prediction result is inactive, there is no need to further perform active timing prediction.

[0086] In one embodiment, if Figure 8 As shown, step S160 includes sub-steps S161 and S162.

[0087] S161: Input the to-be-predicted statistical feature information and the to-be-predicted active feature list into the active opportunity prediction model for calculation to obtain a model output result.

[0088] Specifically, the statistical feature information to be predicted and the active feature list to be predicted can be input through the second input node and the second input matrix of the active timing prediction model, and the model output result is calculated by the active timing prediction model, and the model output result is a multi-dimensional output result.

[0089] S162. According to a preset segmentation threshold, a time period in the model output result whose prediction probability is higher than the segmentation threshold is obtained as an active opportunity prediction result.

[0090] The management server is also pre-configured with a segmentation threshold, and the model output result can be segmented according to the segmentation threshold to obtain the time period in the model output result with a prediction probability higher than the segmentation threshold as the active timing prediction result, and the active timing prediction result can include zero, one or more time periods. If the active timing prediction result includes at least one time period, it indicates that the user's active timing is successfully predicted, and the active timing prediction result can be fed back; if the active timing prediction result does not include a time period, it indicates that the user's active timing is not successfully predicted. At this time, the time period with the highest active frequency (online frequency) can be selected as the user's active time period according to the user's behavior data.

[0091] In the active timing intelligent prediction method provided in the embodiment of the present invention, statistical feature information is obtained based on the received user historical behavior data and user real-time status data, and then encoded to obtain an active feature list. A potential active prediction model and an active timing prediction model are respectively constructed according to the model configuration information. The received user information to be predicted is processed to obtain the statistical feature information to be predicted and the active feature list to be predicted. The active prediction model is used to predict whether it is active. If it is active, the active timing prediction model is used to predict the active timing to obtain the active timing prediction result. Through the above method, a potential active prediction model and an active timing prediction model can be constructed based on historical behavior data and real-time status data, and the potential active prediction model can be used to predict whether the user information to be predicted is active. The active timing prediction result is further predicted by the active timing prediction model. Messages can be pushed based on the active timing prediction result, which greatly improves the accuracy of the message push timing prediction.

[0092] The embodiment of the present invention further provides an active timing intelligent prediction device, which can be configured in the management server 10, and is used to execute any embodiment of the active timing intelligent prediction method described above. Fig. 9 , Fig. 9 A schematic block diagram of an active opportunity intelligent prediction device provided by an embodiment of the present invention.

[0093] like Fig. 9 As shown, the active timing intelligent prediction device 100 includes a statistical feature information acquisition unit 110, an active feature list acquisition unit 120, a model construction unit 130, a to-be-predicted user information processing unit 140, an active prediction result acquisition unit 150 and an active timing prediction result acquisition unit 160.

[0094] The statistical feature information acquisition unit 110 is used to receive the user's historical behavior data and the user's real-time status data from the client, and to obtain corresponding statistical feature information by performing statistics on the user's historical behavior data and the user's real-time status data according to a preset periodic statistical rule.

[0095] In a specific embodiment, the statistical feature information includes list statistical information, basic statistical information and increase statistical information, and the statistical feature information acquisition unit 110 includes sub-units: a list statistical information acquisition unit, which is used to perform list statistics on the behavior data of each user in the user historical behavior data and the real-time active status of each user in the user real-time status data according to the statistical items in the periodic statistical rules, so as to obtain list statistical information corresponding to each of the users; a basic statistical information acquisition unit, which is used to perform periodic statistics on the list statistical information of each user according to the statistical period in the periodic statistical rules to obtain corresponding basic statistical information; and an increase statistical information acquisition unit, which is used to perform comparative statistics on the statistical information of adjacent statistical periods in the basic statistical information to obtain corresponding increase statistical information.

[0096] The active feature list acquisition unit 120 is used to encode the statistical feature information according to a preset encoding conversion rule to obtain a corresponding active feature list.

[0097] In a specific embodiment, the active feature list acquisition unit 120 includes sub-units: a frequency feature list acquisition unit, which is used to convert the frequency value in the statistical feature information according to the conversion formula in the encoding conversion rule to obtain a corresponding frequency feature list; a duration feature list acquisition unit, which is used to perform feature mapping processing on the active duration in the statistical feature information according to the mapping information table in the encoding conversion rule to obtain a corresponding duration feature list.

[0098] The model building unit 130 is used to build a potential activity prediction model and an activity opportunity prediction model according to preset model configuration information, the statistical feature information and the active feature list.

[0099] In a specific embodiment, the model building unit 130 includes sub-units: a potential activity prediction model building unit, which is used to build a corresponding potential activity prediction model based on the model configuration information, the statistical features in the statistical feature information, and the list features in the active feature list; an active timing prediction model building unit, which is used to build a corresponding active timing prediction model based on the model configuration information, the active duration of each time period in the statistical feature information, and the list features in the active feature list.

[0100] In a specific embodiment, the potential activity prediction model construction unit includes subunits: an activity prediction model template configuration unit, which is used to configure the activity prediction model template according to the model configuration information, the number of statistical features and the number of list features; an activity prediction template output information acquisition unit, which is used to calculate the statistical features and list features of each user according to the activity prediction model template to obtain the activity prediction template output information of each user; a model construction unit, which is used to construct a potential activity prediction model based on binary classification according to the activity prediction model template, the activity prediction template output information of each user and the real-time activity status of each user in the active feature list.

[0101] In a specific embodiment, the active timing prediction model construction unit includes subunits: a timing prediction model template configuration unit, which is used to obtain a timing prediction model template according to the model configuration information, the number of time periods in the statistical feature information, and the number of list features; a timing prediction template output information acquisition unit, which is used to calculate the active duration of each time period of each user and the list features according to the timing prediction model template to obtain the timing prediction template output information of each user; a construction unit, which is used to construct an active timing prediction model based on multiple classifications according to the timing prediction model template, the timing prediction template output information of each user, and the active time period of each user in the active feature list.

[0102] The user information to be predicted processing unit 140 is used to, if receiving the user information to be predicted, process the user information to be predicted according to the periodic statistical rule and the encoding conversion rule to obtain corresponding statistical feature information to be predicted and a list of active features to be predicted.

[0103] The activity prediction result acquisition unit 150 is used to predict the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the potential activity prediction model to obtain an activity prediction result.

[0104] The active timing prediction result acquisition unit 160 is configured to, if the prediction result is active, perform active timing prediction on the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the active timing prediction model to obtain an active timing prediction result.

[0105] In a specific embodiment, the active timing prediction result acquisition unit 160 includes sub-units: a model output result acquisition unit, which is used to input the statistical feature information to be predicted and the active feature list to be predicted into the active timing prediction model for calculation to obtain the model output result; a prediction result acquisition unit, which is used to obtain the time period in the model output result with a prediction probability higher than the segmentation threshold as the active timing prediction result according to a preset segmentation threshold.

[0106] The active timing intelligent prediction device provided in the embodiment of the present invention applies the above-mentioned active timing intelligent prediction method, obtains statistical feature information based on the received user historical behavior data and user real-time status data, and then encodes it to obtain an active feature list, and constructs a potential active prediction model and an active timing prediction model according to the model configuration information, processes the received user information to be predicted to obtain the statistical feature information to be predicted and the active feature list to be predicted, and predicts whether it is active through the active prediction model, and if it is active, predicts the active timing through the active timing prediction model to obtain an active timing prediction result. Through the above method, a potential active prediction model and an active timing prediction model can be constructed based on historical behavior data and real-time status data, and whether the user information to be predicted is active can be predicted through the potential active prediction model, and the active timing prediction result can be further predicted through the active timing prediction model, and a message can be pushed based on the active timing prediction result, which greatly improves the accuracy of the message push timing prediction.

[0107] The above-mentioned active opportunity intelligent prediction device can be implemented in the form of a computer program. The computer program can be used in Fig.10 Runs on the computer device shown.

[0108] See also Fig.10 , Fig.10 is a schematic block diagram of a computer device provided by an embodiment of the present invention. The computer device may be a management server 10 for executing an active timing intelligent prediction method to predict the active timing of a user.

[0109] See also Fig.10 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a storage medium 503 and an internal memory 504 .

[0110] The storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 may execute the active opportunity intelligent prediction method, wherein the storage medium 503 may be a volatile storage medium or a non-volatile storage medium.

[0111] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0112] The internal memory 504 provides an environment for the operation of the computer program 5032 in the storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute the active opportunity intelligent prediction method.

[0113] The network interface 505 is used for network communication, such as providing data information transmission, etc. Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device 500 to which the solution of the present invention is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0114] The processor 502 is used to run the computer program 5032 stored in the memory to implement the corresponding functions in the above-mentioned active opportunity intelligent prediction method.

[0115] Those skilled in the art will understand that Fig.10 The embodiments of the computer device shown in the figure do not constitute a limitation on the specific composition of the computer device. In other embodiments, the computer device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, in some embodiments, the computer device may only include a memory and a processor. In such embodiments, the structure and function of the memory and the processor are the same as those of the embodiment of the present invention. Fig.10 The embodiments shown are consistent and will not be described again here.

[0116] It should be understood that in the embodiment of the present invention, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0117] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium may be a volatile or non-volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps included in the above-mentioned active opportunity intelligent prediction method are implemented.

[0118] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the above-described equipment, devices and units can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. Those of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0119] In the several embodiments provided by the present invention, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. Units with the same function may also be combined into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices or units, or may be an electrical, mechanical or other form of connection.

[0120] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0121] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0122] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a computer-readable storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a magnetic disk or an optical disk.

[0123] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An active opportunity intelligent prediction method, characterized in that: The method is applied to a management server, and a network connection is established between the management server and a client to realize the transmission of data information. The method includes: Receiving the user's historical behavior data and the user's real-time status data from the client, and performing statistics on the user's historical behavior data and the user's real-time status data according to a preset periodic statistical rule to obtain corresponding statistical feature information; Encoding the statistical feature information according to a preset encoding conversion rule to obtain a corresponding active feature list; Constructing a potential activity prediction model and an activity opportunity prediction model according to the preset model configuration information, the statistical feature information and the active feature list; If the user information to be predicted is received, the user information to be predicted is processed according to the periodic statistical rule and the encoding conversion rule to obtain corresponding statistical feature information to be predicted and a list of active features to be predicted; Predicting the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the potential active prediction model to obtain a prediction result of whether the activity is active; If the prediction result is active, performing active timing prediction on the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the active timing prediction model to obtain an active timing prediction result; The step of constructing a potential activity prediction model and an activity opportunity prediction model according to the preset model configuration information, the statistical feature information and the active feature list includes: A corresponding potential active prediction model is constructed according to the model configuration information, the statistical features in the statistical feature information, and the list features in the active feature list; the model configuration information includes configuration parameters of a fully connected hidden layer and configuration parameters of a weight layer; A corresponding active timing prediction model is constructed according to the model configuration information, the active duration of each time period in the statistical feature information, and the list features in the active feature list.

2. The active timing intelligent prediction method according to claim 1, characterized in that: The statistical feature information includes list statistical information, basic statistical information and increment statistical information. The corresponding statistical feature information is obtained by performing statistics on the user historical behavior data and the user real-time status data according to the preset periodic statistical rules, including: According to the statistical items in the periodic statistical rule, the behavior data of each user in the user historical behavior data and the real-time active status of each user in the user real-time status data are respectively tabulated and counted to obtain the list statistical information corresponding to each of the users; Performing periodic statistics on the list statistical information of each user according to the statistical period in the periodic statistical rule to obtain corresponding basic statistical information; Corresponding increase statistical information is obtained by performing comparative statistics based on the statistical information of adjacent statistical periods in the basic statistical information.

3. The active timing intelligent prediction method according to claim 1, characterized in that: The active feature list includes a frequency feature list and a duration feature list, and the encoding process of the statistical feature information according to the preset encoding conversion rule to obtain the corresponding active feature list includes: Convert the frequency value in the statistical feature information according to the conversion formula in the encoding conversion rule to obtain a corresponding frequency feature list; According to the mapping information table in the encoding conversion rule, feature mapping processing is performed on the active duration in the statistical feature information to obtain a corresponding duration feature list.

4. The active timing intelligent prediction method according to claim 1, characterized in that: The constructing a corresponding potential activity prediction model according to the model configuration information, the statistical features in the statistical feature information, and the list features in the active feature list includes: According to the model configuration information, the number of the statistical features and the number of the list features, an active prediction model template is obtained; Calculating the statistical features and list features of each user according to the activity prediction model template to obtain activity prediction template output information of each user; A potential activity prediction model based on binary classification is constructed according to the activity prediction model template, the activity prediction template output information of each of the users, and the real-time activity status of each user in the active feature list.

5. The active timing intelligent prediction method according to claim 1, characterized in that: The constructing a corresponding active timing prediction model according to the model configuration information, the active duration of each time period in the statistical feature information, and the list features in the active feature list includes: Obtaining an opportunity prediction model template according to the model configuration information, the number of time periods in the statistical feature information, and the number of list features; Calculating the active duration and list features of each user in each time period according to the timing prediction model template to obtain timing prediction template output information of each user; An active timing prediction model based on multiple classifications is constructed according to the timing prediction model template, the timing prediction template output information of each of the users, and the active time period of each user in the active feature list.

6. The active timing intelligent prediction method according to claim 1, characterized in that: The step of performing active timing prediction on the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the active timing prediction model to obtain an active timing prediction result includes: Inputting the to-be-predicted statistical feature information and the to-be-predicted active feature list into the active opportunity prediction model for calculation to obtain a model output result; According to a preset segmentation threshold, a time period in the model output result whose prediction probability is higher than the segmentation threshold is obtained as an active opportunity prediction result.

7. An active opportunity intelligent prediction device, characterized in that: The device is configured in a management server, and a network connection is established between the management server and the client to realize the transmission of data information. The device includes: A statistical feature information acquisition unit, used to receive the user's historical behavior data and the user's real-time status data from the client, and to perform statistics on the user's historical behavior data and the user's real-time status data according to a preset periodic statistical rule to obtain corresponding statistical feature information; An active feature list acquisition unit, configured to encode the statistical feature information according to a preset encoding conversion rule to obtain a corresponding active feature list; A model building unit, used to build a potential activity prediction model and an activity opportunity prediction model according to preset model configuration information, the statistical feature information and the active feature list; A user information processing unit to be predicted is used to, upon receiving the user information to be predicted, process the user information to be predicted according to the periodic statistical rule and the encoding conversion rule to obtain corresponding statistical feature information to be predicted and a list of active features to be predicted; an activity prediction result acquisition unit, configured to predict the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the potential activity prediction model, so as to obtain a prediction result of whether the activity is present; an active timing prediction result acquisition unit, configured to, if the prediction result is active, perform active timing prediction on the to-be-predicted statistical feature information and the to-be-predicted active feature list according to the active timing prediction model to obtain an active timing prediction result; The model construction unit includes subunits: a potential activity prediction model construction unit, which is used to construct a corresponding potential activity prediction model according to the model configuration information, the statistical features in the statistical feature information, and the list features in the active feature list; the model configuration information contains configuration parameters of the fully connected hidden layer and configuration parameters of the weight layer; an active timing prediction model construction unit, which is used to construct a corresponding active timing prediction model according to the model configuration information, the active duration of each time period in the statistical feature information, and the list features in the active feature list.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the active opportunity intelligent prediction method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the active opportunity intelligent prediction method according to any one of claims 1 to 6 is implemented.

Images