Method and apparatus for information prediction

Abstract

Embodiments of the present invention disclose a method and device for information prediction, including obtaining historical data of users to whom advertisements are delivered as sample data, and labeling tags for the sample data according to different user characteristics; after splitting the entire link into multiple service nodes, determining funnel samples; including: for each service node, taking the samples converted at this node as positive samples and the samples not converted at this node as negative samples, where the number of positive samples of the previous node in the entire link is used as the total amount of samples of its next node; using the funnel samples as the input of the prediction model and the actual conversion rate as the output of the prediction model to train the prediction model, where, during the training process, each node is used as an optimization sub-goal; performing multi-objective optimization on the model based on XGB. By training the model in a multi-objective optimization manner, the accuracy of user positioning is improved, and thus the advertising delivery accuracy can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information prediction, and particularly to a method and device for information prediction. Background Art

[0002] In the business of implementing advertising delivery based on RTA, usually a single target analysis is performed for a certain node, and the advertising delivery is optimized based on the analysis result. For example, the conversion rate on a single node is analyzed, and each node is optimized.

[0003] The above analysis method causes inaccurate value positioning of users, resulting in low accuracy of advertising delivery, as well as problems of high advertising delivery cost but low conversion rate of advertisements. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a method and device for information prediction.

[0005] To achieve the above purpose, the embodiments of the present invention provide a method for information prediction. The method is used to predict the conversion rate of a service in the entire link. The service includes the service of implementing advertising delivery based on RTA. The method includes: obtaining the user historical data of the advertised user as sample data, and labeling the sample data according to different user characteristics; after dividing the entire link into multiple service nodes, determining the funnel samples, including: for each service node, taking the samples converted at this node as positive samples, and taking the samples not converted at this node as negative samples, where the number of positive samples of the previous node in the entire link is used as the total amount of samples of its next node; using the funnel samples as the input of the prediction model and the actual conversion rate as the output of the prediction model to train the prediction model, where during the training process, each node is used as an optimization sub-goal; performing multi-objective optimization on the model based on XGB.

[0006] Optionally, the method further includes: when the trained prediction model receives the traffic of user data, predicting the conversion rate for this traffic; and / or, comparing the predicted conversion rate with a preset threshold to determine whether to perform an advertising bid.

[0007] Optionally, the method further includes: further dividing the funnel samples into a test sample set and a validation sample set; after the prediction model is trained, verifying the conversion rate of the prediction model based on the test sample data; when the verification passes, using the validation sample set to continue verifying the conversion rate of the prediction model.

[0008] Optionally, if the conversion rate verification fails, perform hyperparameter adjustment and residual penalty on the prediction model.

[0009] Optionally, the method further includes: when the advertisement is delivered to multiple channels, adjusting the structure of the funnel sample to adapt to each channel.

[0010] According to a second aspect of the present disclosure, there is provided an apparatus for information prediction, which is configured to predict the conversion rate of a service in the entire link. The service includes a service for advertising delivery based on RTA. The apparatus includes: a labeling unit configured to obtain historical user data of the advertised user as sample data and label the sample data for different user characteristics; a sample determination unit configured to determine a funnel sample after dividing the entire link into multiple service nodes, including: for each service node, using the sample converted at this node as a positive sample and the sample not converted at this node as a negative sample, where the number of positive samples of the previous node in the entire link is used as the total amount of samples of its next node; a model training unit configured to use the funnel sample as the input of the prediction model and the actual conversion rate as the output of the prediction model to train the prediction model. During the training process, each node is used as an optimization sub-goal, and multi-objective optimization of the model is performed based on XGB.

[0011] Optionally, the apparatus further includes: a prediction unit configured to perform conversion rate prediction for the traffic when the trained prediction model receives the traffic of user data; and / or a bidding unit configured to compare the predicted conversion rate with a preset threshold to determine whether to perform an advertisement bid.

[0012] Optionally, it further includes: dividing the funnel sample into a test sample set and a validation sample set; after the prediction model is trained, performing conversion rate verification on the prediction model based on the test sample data; and when the verification passes, continuing to perform conversion rate verification on the prediction model using the validation sample set.

[0013] According to a third aspect of the present disclosure, there is provided a computer-readable storage medium storing computer instructions for causing the computer to execute the method for information prediction according to any one of the implementation manners of the first aspect.

[0014] According to a fourth aspect of the present disclosure, there is provided an electronic device including: at least one processor; and a memory communicatively connected to the at least one processor, where the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to cause the at least one processor to execute the method of the apparatus for information prediction according to any one of the implementation manners of the first aspect.

[0015] In the method and apparatus for information prediction according to the embodiments of the present disclosure, historical data of users to whom advertisements are delivered is obtained as sample data, and tags are labeled for the sample data according to different user characteristics; after the full link is segmented into multiple service nodes, funnel samples are determined, including: for each service node, the samples converted at this node are used as positive samples, and the samples not converted at this node are used as negative samples, wherein the number of positive samples of the previous node in the full link is used as the total amount of samples of its next node; the funnel samples are used as the input of the prediction model, and the actual conversion rate is used as the output of the prediction model to train the prediction model. During the training process, each node is used as an optimization sub-goal; the model is optimized with multiple goals based on XGB. By training the model in a multi-objective optimization manner, the accuracy of user positioning is improved, and thus the advertising delivery accuracy can be improved, overcoming the problems of inability to globally optimize, low accuracy of user positioning, and high customer acquisition cost caused by optimizing through each single node in marketing customer acquisition. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, other implementation drawings can be obtained according to the provided drawings without creative efforts.

[0017] The structures, ratios, sizes, etc. depicted in this specification are only used to cooperate with the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the limiting conditions under which the present invention can be implemented. Therefore, they do not have a substantial technical meaning. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that can be covered by the technical content disclosed in the present invention.

[0018] Figure 1 It is a flowchart of the method for information prediction provided by the embodiments of the present invention.

[0019] Figure 2 It is an application diagram of the method for information prediction provided by the embodiments of the present invention;

[0020] Figure 3 It is another application diagram of the method for information prediction provided by the embodiments of the present invention;

[0021] Figure 4 It is a schematic diagram of an electronic device according to the embodiments of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The following specific embodiments illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] It should be noted that, without conflict, the embodiments in this disclosure and the features in the embodiments can be combined with each other. The following will detail this disclosure with reference to the accompanying drawings and in combination with the embodiments.

[0024] RTA, REAL TIME API, is a real-time advertising participation process from the user side. Users, through modeling, find out the user groups they need, and then participate in the bidding according to the user groups they need when the advertising request reaches the media. The advantages of this method are that on the one hand, it can get rid of the algorithm logic of the media to a certain extent, and on the other hand, it can use its own data or algorithm advantages to more freely find user groups, expanding the traffic pool.

[0025] According to an embodiment of the present disclosure, a method for information prediction is provided. The method is used to predict the conversion rate of a service in the entire link, and the service includes a service for realizing advertising placement based on RTA.

[0026] In the existing customer acquisition services (services for determining user groups for the advertised advertisements), most of them develop and optimize a single target at different conversion nodes, and their overall efficiency and accuracy are relatively low, resulting in low customer acquisition quality (that is, the determined users to be advertised are inaccurate) and high advertising placement costs. The method of this embodiment predicts the conversion rate of the entire link in the advertising placement service and optimizes the target based on the entire link, improving the accuracy of user positioning in the advertising placement service based on RTA. As Figure 1 shown, the method includes the following steps 101 to step 103:

[0027] Step 101: Obtain the user historical data of the advertised advertisement as sample data, and label the sample data according to different user characteristics.

[0028] In this embodiment, for a specific advertised advertisement, the user data within a preset historical period can be obtained. The user data can be labeled with preset labels, and the content of the labels can include the device information of the user (for example, the model of the device, etc.), the attribute information of the user (for example, the age of the user, etc.), the historical behavior information of the user, the information of the advertising channels, and variables such as the frequency, maximum, minimum, and variance derived from processing.

[0029] Step 102: After splitting the full link into multiple service nodes, determine the funnel samples, including: for each service node, taking the samples converted at this node as positive samples and the samples not converted at this node as negative samples, where the number of positive samples of the previous node in the full link is used as the total number of samples of its next node.

[0030] In this embodiment, the service of realizing advertisement delivery based on RTA can be split into multiple nodes, including but not limited to click node, registration node, first login node, service application node (for example, loan service application node), node where the applied service is implemented (for example, node where loan operation is performed after applying for the loan service), evaluation node (for example, credit limit evaluation node), usage node (for example, usage rate of the quota), charging node (for example, CPS, CPA, CPCL, CPC, or CPM, etc.) and other nodes.

[0031] At each node, it can be determined whether conversion has occurred at this node based on sample data. For example, at the click node, if a click occurs, it indicates that the sample is a converted sample, and this sample is taken as a positive sample; if no click occurs, it indicates non-conversion, and the non-converted sample is taken as a negative sample.

[0032] Since the sequentiality of each node constitutes the full link of the service, the number of positive samples of the previous node in the full link is used as the total number of samples of its next node. In this way, funnel samples can be obtained. For example, taking the clicked samples in the click node as the total samples of the next node, i.e., the registration node, and counting the number of samples that complete registration and the number of samples that do not complete registration in this total sample; then taking the samples that complete registration as the total samples of the first login node, and counting the number of samples that complete login and the number of samples that do not complete login at the first login node; then taking the number of samples that complete login as the total number of samples of the service application node, and counting the samples that complete the service application and the samples that do not complete the service application; taking the samples that complete the service application as the total amount of the node where the applied service is implemented, and counting the nodes that complete implementation and the nodes that do not complete implementation; taking the samples of the nodes that complete implementation as the total number of samples of the evaluation node, and counting the number of samples that complete evaluation and the number of samples that do not complete evaluation; taking the number of samples that complete evaluation as the total number of samples of the usage node, and counting the number of samples that complete usage and the number of samples that do not complete usage; taking the number of samples that complete usage as the total number of samples of the charging node, and counting the number of samples that complete charging and the number of samples that do not complete charging. This is only exemplary. In this way, samples of different layers (each node can be regarded as a layer) can be formed, and the samples of each layer have different sample concentrations. The calculation method of this sample concentration can be the ratio of the number of converted samples to the total number of samples.

[0033] Step 103: Use the funnel sample as the input of the prediction model and the actual conversion rate as the output of the prediction model to train the prediction model. During the training process, each node is regarded as an optimization sub-goal; multi-objective optimization is performed on the model based on XGB.

[0034] In this embodiment, a prediction model can be established in advance. The prediction model includes a multi-objective learning model, and each objective learning model can correspond to an optimization objective. When training the prediction model, the funnel sample can be used as the input and the actual conversion rate as the output for training. The actual conversion rate can be determined as needed, including the final conversion rate for all nodes or the conversion rate of some nodes. During training, model training is performed based on the training sample set (the samples can be divided into a training sample set, a test sample set, and a validation sample set). Refer to Figure 2 , XGB can be used for multi-objective residual optimization regression to improve the fitting effect of the gradient boosting model.

[0035] By training the model through multi-objective optimization, it overcomes the problems of inability to globally optimize, low user positioning accuracy, and high customer acquisition cost caused by optimizing through each single node in marketing customer acquisition.

[0036] As an optional implementation manner of this embodiment, the method further includes: when the trained prediction model receives the traffic of user data, predicting the conversion rate for this traffic; and / or comparing the predicted conversion rate with a preset threshold to determine whether to perform an advertising bid.

[0037] In this optional implementation manner, after the model training is completed, the model can be used to predict the conversion rate of new user traffic (which can be the conversion rate including all nodes). If the predicted conversion rate is greater than the preset threshold, an advertising bid can be performed; if the conversion rate is lower than the preset threshold, no bid is made and no advertisement is placed for this user traffic.

[0038] As an optional implementation manner of this embodiment, the method further includes: dividing the funnel sample into a test sample set and a validation sample set; after the prediction model training is completed, verifying the conversion rate of the prediction model based on the test sample data; when the verification passes, continuing to verify the conversion rate of the prediction model using the validation sample set.

[0039] As an optional implementation manner of this embodiment, if the conversion rate verification fails, hyperparameter adjustment and residual penalty are performed on the prediction model.

[0040] Refer to Figure 3, the trained model is tested on the test set, and the model training process is adjusted according to the test set results until the test set effect is satisfactory (for example, including being close to the actual conversion rate). This tuning process includes adjusting hyperparameter coefficients, penalty residual optimization, etc. Finally, the OOT effect is verified on the validation sample, and the model effects on the training, test, and validation sub-samples are compared and analyzed until the model effects are all good.

[0041] As an optional implementation manner of this embodiment, the method further includes: when the advertisement is delivered to multiple objects, adjusting the structure of the funnel sample to adapt to each object.

[0042] In this optional implementation manner, each platform may correspond to multiple channels. For example, platform A includes channels A1, A2, A3... etc. The prediction model may be a model for predicting the conversion rate of the entire platform A. When predicting any specific channel, the sample concentration of each layer of the funnel sample can be adjusted to achieve the prediction of the specific channel.

[0043] From the above description, it can be seen that the present disclosure achieves the following technical effects:

[0044] This embodiment trains the model based on a multi-objective optimization method, improves the prediction accuracy and prediction efficiency of the model, and thus can improve the advertising delivery accuracy.

[0045] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0046] According to an embodiment of the present disclosure, there is also provided an apparatus for information prediction, which is used to predict the conversion rate of a service in the entire link. The service includes a service for realizing advertisement delivery based on RTA. The apparatus includes:

[0047] A labeling unit, configured to obtain the user historical data of the advertisement to be delivered as sample data, and label the sample data according to different user characteristics;

[0048] A sample determination unit, configured to determine a funnel sample after dividing the entire link into multiple service nodes; including: for each service node, taking the sample converted at this node as a positive sample, and taking the sample not converted at this node as a negative sample, where the number of positive samples of the previous node in the entire link is used as the total amount of samples of its next node;

[0049] A model training unit is configured to use the funnel samples as the input of a prediction model and the actual conversion rate as the output of the prediction model to train the prediction model. During the training process, each node is regarded as an optimization sub-goal, and multi-objective optimization of the model is performed based on XGB.

[0050] As an optional implementation manner of this embodiment, the device further includes: a prediction unit configured to perform conversion rate prediction for the traffic when the trained prediction model receives the traffic of user data; and / or a bidding unit configured to compare the predicted conversion rate with a preset threshold to determine whether to perform an advertising bid.

[0051] As an optional implementation manner of this embodiment, the device further includes: dividing the funnel samples into a test sample set and a validation sample set; after the prediction model is trained, performing conversion rate validation on the prediction model based on the test sample data; and when the validation passes, continuing to perform conversion rate validation on the prediction model using the validation sample set.

[0052] As an optional implementation manner of this embodiment, if the conversion rate validation fails, hyperparameter adjustment and residual penalty are performed on the prediction model.

[0053] As an optional implementation manner of this embodiment, the device further includes: when the advertisement is delivered to multiple channels, adjusting the structure of the funnel samples to adapt to each channel.

[0054] An embodiment of the present disclosure provides an electronic device, as Figure 4 shown, the electronic device includes one or more processors 41 and a memory 42, Figure 4 Taking one processor 41 as an example in

[0055] The controller may further include: an input device 43 and an output device 44.

[0056] The processor 41, the memory 42, the input device 43, and the output device 44 may be connected through a bus or other means, Figure 4 Taking connection through a bus as an example in

[0057] The processor 41 may be a Central Processing Unit (CPU). The processor 41 may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. chips, or combinations of the above types of chips. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0058] As a non-transitory computer-readable storage medium, the memory 42 can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the control method in the embodiments of the present disclosure. The processor 41 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 42, that is, implements the method of the above method embodiments.

[0059] The memory 42 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the processing device of the server operation, etc. In addition, the memory 42 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory 42 may optionally include a memory remotely set relative to the processor 41, and these remote memories can be connected to the network connection device through a network. Examples of the above networks include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and combinations thereof.

[0060] The input device 43 can receive input digital or character information, and generate key signal inputs related to the user settings and function controls of the processing device of the server. The output device 44 may include a display device such as a display screen.

[0061] One or more modules are stored in the memory 42 and, when executed by one or more processors 41, execute as Figure 1 shown in the method.

[0062] Those skilled in the art can understand that to implement all or part of the processes in the above method embodiments, it can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above embodiments of each motor control method. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memories.

[0063] Although the present invention has been described in detail above with general descriptions and specific embodiments, based on the present invention, some modifications or improvements can be made, which are obvious to those skilled in the art. Therefore, these modifications or improvements made without departing from the spirit of the present invention all fall within the scope of protection required by the present invention.

Claims

1. A method for information prediction, characterized in that, The method is used to predict the conversion rate of a service in the entire link, where the service includes the service of realizing advertisement delivery based on RTA. The method includes: Obtain the historical user data of the advertised users as sample data, and label the sample data according to different user characteristics; After dividing the entire link into multiple service nodes, determine the funnel samples, including: for each service node, use the samples converted at this node as positive samples and the samples not converted at this node as negative samples, where the number of positive samples of the previous node in the entire link is used as the total amount of samples of its next node; Use the funnel samples as the input of the prediction model and the actual conversion rate as the output of the prediction model to train the prediction model. During the training process, each node is used as an optimization sub-goal, and multi-objective optimization is performed on the model based on XGB; When the trained prediction model receives the traffic of user data, perform conversion rate prediction for this traffic; And / or, compare the predicted conversion rate with a preset threshold to determine whether to perform an advertisement auction.

2. The method for information prediction according to claim 1, wherein The method further includes: Also divide the funnel samples into a test sample set and a validation sample set; After the prediction model is trained, perform conversion rate verification on the prediction model based on the test sample data; After passing the verification, continue to perform conversion rate verification on the prediction model using the validation sample set.

3. The method for information prediction according to claim 2, wherein If the conversion rate verification fails, perform hyperparameter adjustment and residual penalty on the prediction model.

4. The method for information prediction according to claim 1, wherein The method further includes: When the advertisement is delivered to multiple channels, adjust the structure of the funnel samples to adapt to each channel.

5. An apparatus for information prediction, characterized in that, The device is used to predict the conversion rate of a service in the entire link, where the service includes the service of realizing advertisement delivery based on RTA. The device includes: A labeling unit configured to obtain the historical user data of the advertised users as sample data and label the sample data according to different user characteristics; A sample determination unit configured to determine the funnel samples after dividing the entire link into multiple service nodes, including: for each service node, use the samples converted at this node as positive samples and the samples not converted at this node as negative samples, where the number of positive samples of the previous node in the entire link is used as the total amount of samples of its next node; A model training unit configured to use the funnel samples as the input of the prediction model and the actual conversion rate as the output of the prediction model to train the prediction model. During the training process, each node is used as an optimization sub-goal, and multi-objective optimization is performed on the model based on XGB; The device further includes: A prediction unit configured to perform conversion rate prediction for the traffic when the trained prediction model receives the traffic of user data; And / or, a bidding unit configured to compare the predicted conversion rate with a preset threshold to determine whether to perform an advertisement auction.

6. The apparatus for information prediction according to claim 5, wherein It further includes: Divide the funnel samples into a test sample set and a validation sample set; after the prediction model is trained, verify the conversion rate of the prediction model based on the test sample data; when the verification passes, continue to verify the conversion rate of the prediction model using the validation sample set.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the method for information prediction according to any one of claims 1-4.

8. An electronic device, characterized in that, Comprising: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to cause the at least one processor to execute the method for information prediction according to any one of claims 1-4.

Images