Recommendation page generation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a recommendation page generation method and device, electronic equipment and a storage medium, and belongs to the technical field of data recommendation. The method comprises the following steps: acquiring sample user information, sample page information, a sample click rate and a sample conversion rate; training a preset click conversion rate prediction model according to the sample user information, the sample page information, the sample click rate and the sample conversion rate to obtain a target click conversion prediction model; generating a calibration model through the target click conversion prediction model, the sample user information, the sample page information, the sample click rate and the sample conversion rate to obtain a target click conversion calibration model; and obtaining a target user portrait and original page text information, and performing page adjustment on the candidate page through the target click conversion calibration model, the target click conversion prediction model, the target user portrait and the original page text information to obtain a target recommendation page. According to the embodiment of the invention, the click rate and conversion rate of the recommended page can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of data recommendation, and in particular to a method and device for generating a recommendation page, an electronic device, and a storage medium. Background Art

[0002] A recommendation algorithm is an algorithm that recommends products to customers. It can be used for product recommendations, song recommendations, article recommendations, etc. For example, the product recommendation function can be implemented through page recommendations. Page recommendations refer to the interface used to recommend products on the application interface, such as the product recommendation page on the mobile banking application. The existing recommendation page generation algorithm lacks an in-depth understanding of user personalized needs, and the recommended content does not match user interests, resulting in low click-through rate and conversion rate of the recommendation interface. Therefore, how to improve the click-through rate and conversion rate of the recommendation page has become an urgent problem to be solved. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a recommendation page generation method and device, an electronic device and a storage medium, aiming to improve the click rate and conversion rate of the recommendation page.

[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a method for generating a recommendation page, the method comprising:

[0005] Obtain sample user information of the sample object, obtain sample page information of the sample recommendation page, and obtain the sample click rate and sample conversion rate of the sample object on the sample recommendation page;

[0006] Generate a sample user profile based on the sample user information to obtain a sample user profile;

[0007] Extracting content according to the sample page information to obtain sample page text information;

[0008] The preset click-to-conversion rate prediction model is trained according to the sample user portrait, the sample page text information, the sample click-through rate and the sample conversion rate to obtain a target click-to-conversion prediction model;

[0009] A calibration model is generated by using the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate to obtain a target click-to-conversion calibration model;

[0010] The target user portrait of the target object is obtained, the original page text information of the candidate page is obtained, and the candidate page is adjusted through the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait and the original page text information to obtain the target recommended page.

[0011] In some embodiments, the training of a preset click-to-conversion rate prediction model according to the sample user portrait, the sample page text information, the sample click rate, and the sample conversion rate to obtain a target click-to-conversion prediction model includes:

[0012] The click-through conversion rate is predicted by using the preset click-through conversion rate prediction model, the sample user portrait and the sample page text information to obtain a sample predicted click-through rate and a sample predicted conversion rate;

[0013] Calculate the loss value according to the sample predicted click rate and the sample click rate to obtain the click rate loss value;

[0014] Calculate the loss value according to the sample predicted conversion rate and the sample conversion rate to obtain a conversion rate loss value;

[0015] Perform loss value aggregation according to the click rate loss value and the conversion rate loss value to obtain a training loss value;

[0016] The preset click-to-conversion rate prediction model is optimized based on the training loss value to obtain the target click-to-conversion prediction model.

[0017] In some embodiments, the preset click-to-conversion rate prediction model includes a click-to-conversion rate prediction network, a conversion rate prediction network, and a shared prediction network; the click-to-conversion rate prediction is performed by using the preset click-to-conversion rate prediction model, the sample user portrait, and the sample page text information to obtain a sample predicted click rate and a sample predicted conversion rate, including:

[0018] Perform click rate prediction on the sample user portrait and the sample page text information through the click rate prediction network to obtain original click rate prediction data;

[0019] Performing conversion rate prediction on the sample user portrait and the sample page text information through the conversion rate prediction network to obtain original conversion rate prediction data;

[0020] Performing shared prediction on the sample user portrait and the sample page text information through the shared prediction network to obtain original shared prediction data;

[0021] Performing aggregation calculation based on the original click rate prediction data and the original shared prediction data to obtain the sample predicted click rate;

[0022] Aggregate calculation is performed based on the original conversion rate prediction data and the original shared prediction data to obtain the sample predicted conversion rate.

[0023] In some embodiments, generating a calibration model by using the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate, and the sample conversion rate to obtain a target click-to-conversion calibration model includes:

[0024] The click-through conversion rate of the sample user portrait and the sample page text information is predicted by the target click-through conversion prediction model to obtain an initial click-through rate and an initial conversion rate;

[0025] Performing parameter fitting on a preset click rate calibration model according to the initial click rate and the sample click rate to obtain a target click rate calibration model;

[0026] Performing parameter fitting on a preset conversion rate calibration model according to the initial conversion rate and the sample conversion rate to obtain a target conversion rate calibration model;

[0027] Model merging is performed based on the target click rate calibration model and the target conversion rate calibration model to obtain the target click-to-conversion calibration model.

[0028] In some embodiments, the sample user information includes user identity information and user browsing history information; generating a portrait based on the sample user information to obtain a sample user portrait includes:

[0029] Performing clustering according to the user identity information and the preset clustering area to obtain user clustering information;

[0030] Perform preference identification based on the user's historical browsing information to obtain user preference information;

[0031] A portrait is generated according to the user clustering information and the user preference information to obtain the sample user portrait.

[0032] In some embodiments, the sample page information includes recommended content information; and extracting content according to the sample page information to obtain sample page text information includes:

[0033] Extracting a content format according to the recommended content information to obtain at least one modal content data;

[0034] Performing text topic extraction on each of the modal content data to obtain at least one content text data;

[0035] Text content aggregation is performed on each of the content text data to obtain the sample page text information.

[0036] In some embodiments, the step of adjusting the candidate page by using the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait, and the original page text information to obtain a target recommended page includes:

[0037] Performing click-through conversion prediction on the target user portrait and the original page text information through the target click-through conversion prediction model to obtain an initial predicted click-through rate and an initial predicted conversion rate of the candidate page;

[0038] Calibrate the initial predicted click rate by using the target click-through conversion calibration model to obtain a target predicted click rate;

[0039] Calibrate the initial predicted conversion rate using the target click-through conversion calibration model to obtain a target predicted conversion rate;

[0040] The candidate pages are optimized according to the target predicted click-through rate and the target predicted conversion rate to obtain the target recommended page.

[0041] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a recommendation page generation device, the device comprising:

[0042] An information acquisition module is used to acquire sample user information of a sample object, acquire sample page information of a sample recommendation page, and acquire a sample click rate and a sample conversion rate of the sample object on the sample recommendation page;

[0043] A portrait generation module, used to generate a portrait based on the sample user information to obtain a sample user portrait;

[0044] A content extraction module, used to extract content according to the sample page information to obtain sample page text information;

[0045] A model training module is used to train a preset click-to-conversion rate prediction model according to the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate to obtain a target click-to-conversion prediction model;

[0046] A model generation module, used to generate a calibration model through the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate, to obtain a target click-to-conversion calibration model;

[0047] The page adjustment module is used to obtain the target user portrait of the target object, obtain the original page text information of the candidate page, and adjust the candidate page through the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait and the original page text information to obtain the target recommended page.

[0048] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0049] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes 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 method described in the first aspect is implemented.

[0050] The present application proposes a method and device for generating a recommendation page, an electronic device, and a storage medium. The method obtains sample user information of a sample object, obtains sample page information of a sample recommendation page, and obtains a sample click rate and a sample conversion rate of the sample object for the sample recommendation page. Then, a portrait is generated according to the sample user information to obtain a sample user portrait, and content is extracted according to the sample page information to obtain sample page text information. Then, a preset click-to-conversion rate prediction model is trained according to the sample user portrait, the sample page text information, the sample click rate, and the sample conversion rate, so as to obtain a target click-to-conversion prediction model that can accurately predict the click rate and conversion rate based on the sample user portrait and the sample page text; further, a calibration model is generated through the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate, and the sample conversion rate, so as to obtain a target click-to-conversion calibration model that can further accurately calibrate the click rate and conversion rate; finally, a target user portrait of the target object is obtained, the original page text information of the candidate page is obtained, and the candidate page is adjusted through the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait, and the original page text information to obtain a target recommendation page, so as to improve the click rate and conversion rate of the page. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flowchart of a method for generating a recommendation page provided in an embodiment of the present application;

[0052] Figure 2 yes Figure 1 Flow chart of step S102 in FIG.

[0053] Figure 3 yes Figure 1 Flow chart of step S103 in FIG.

[0054] Figure 4 yes Figure 1 Flow chart of step S104 in FIG.

[0055] Figure 5 yes Figure 4 Flow chart of step S401 in FIG.

[0056] Figure 6 yes Figure 1 Flow chart of step S105 in FIG.

[0057] Figure 7 yes Figure 1 Flow chart of step S106 in FIG.

[0058] Figure 8 is a schematic diagram of the structure of a recommendation page generating device provided in an embodiment of the present application;

[0059] Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0061] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0063] First, some nouns involved in this application are analyzed:

[0064] Collaborative Gating Control Network (CGC network): is a deep learning architecture for multi-task learning, which aims to optimize the performance of multiple related tasks simultaneously through collaborative sharing and dynamic control mechanisms. The CGC network introduces core components such as shared experts, task-specific experts, and gating mechanisms to achieve efficient information sharing and orderly control between different tasks, thereby improving the performance and generalization capabilities of the overall model. This network architecture is widely used in recommendation systems, natural language processing, and computer vision. It can effectively balance the mutual influence between tasks and improve the accuracy and efficiency of the model. The design of the CGC network not only enhances the adaptability of the model to multiple tasks, but also dynamically adjusts the information flow through the gating mechanism, avoids the negative transfer phenomenon between tasks, and ensures that each task can play its specific advantages on the basis of shared information.

[0065] Shared Experts: In the deep learning architecture of multi-task learning, it is a neural network module used to serve multiple learning tasks at the same time. Shared experts achieve information sharing and collaborative learning by extracting common features from the input data that are useful for all tasks. Shared experts are an integral part of the field of deep learning in computer science, which aims to understand how to handle multi-task problems more effectively and produce intelligent models that can optimize multiple learning tasks simultaneously. Research in this field includes multi-task learning, feature extraction, information sharing, and neural network design.

[0066] Click-Through Rate (CTR): It is a key indicator to measure the effectiveness of the recommendation system, indicating the proportion of users clicking on the advertisement or recommended content after it is displayed. Click-through rate is an important parameter in the field of online marketing and advertising, and is often used to evaluate the attractiveness of advertisements and the relevance of content. A high click-through rate usually means that the advertisement or recommended content is highly relevant to the target audience and can effectively attract the user's attention, thereby increasing conversion rate and business revenue. Click-through rate is widely used in many fields such as search engine advertising, social media advertising, email marketing, mobile application recommendations, etc., to optimize advertising delivery strategies and recommendation algorithms.

[0067] Conversion Rate (CVR): It is a key indicator to measure the effectiveness of online marketing or recommendation systems, indicating the ratio between the number of users completing the expected behavior and the total number of interactions. A high conversion rate means that the marketing strategy or recommended content can effectively prompt users to take the target behavior, thereby improving business revenue and user satisfaction.

[0068] A recommendation algorithm is an algorithm that recommends products to customers. It can be used for product recommendations, song recommendations, article recommendations, etc. For example, the product recommendation function can be implemented through page recommendations. Page recommendations refer to the interface used to recommend products on the application interface, such as the product recommendation page on the mobile banking application. The existing recommendation page generation algorithm lacks an in-depth understanding of user personalized needs, and the recommended content does not match user interests, resulting in low click-through rate and conversion rate of the recommendation interface. Therefore, how to improve the click-through rate and conversion rate of the recommendation page has become an urgent problem to be solved.

[0069] Based on this, the embodiments of the present application provide a method and device for generating a recommendation page, an electronic device and a storage medium, aiming to improve the click rate and conversion rate of the recommendation page.

[0070] A recommendation page generation method and device, electronic device and storage medium provided in the embodiments of the present application are specifically described through the following embodiments. First, the recommendation page generation method in the embodiments of the present application is described.

[0071] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0072] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0073] The recommendation page generation method provided in the embodiment of the present application relates to the field of data recommendation technology. The recommendation page generation method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the recommendation page generation method, etc., but is not limited to the above forms.

[0074] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0075] Figure 1 is an optional flowchart of the method for generating a recommendation page provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S106.

[0076] Step S101, obtaining sample user information of sample objects, obtaining sample page information of sample recommendation pages, and obtaining sample click rates and sample conversion rates of sample objects on sample recommendation pages;

[0077] Step S102, generating a portrait according to the sample user information to obtain a sample user portrait;

[0078] Step S103, extracting content according to the sample page information to obtain sample page text information;

[0079] Step S104, training a preset click-to-conversion rate prediction model according to sample user portraits, sample page text information, sample click-through rates, and sample conversion rates to obtain a target click-to-conversion prediction model;

[0080] Step S105, generating a calibration model through the target click-through conversion prediction model, sample user portraits, sample page text information, sample click-through rates, and sample conversion rates, to obtain a target click-through conversion calibration model;

[0081] Step S106, obtain the target user portrait of the target object, obtain the original page text information of the candidate page, and adjust the candidate page through the target click-through conversion calibration model, the target click-through conversion prediction model, the target user portrait and the original page text information to obtain the target recommended page.

[0082] In the steps S101 to S106 shown in the embodiment of the present application, sample user information of the sample object is obtained, sample page information of the sample recommendation page is obtained, and the sample click rate and sample conversion rate of the sample object for the sample recommendation page are obtained, and then a portrait is generated according to the sample user information to obtain a sample user portrait, and content is extracted according to the sample page information to obtain sample page text information, and then a preset click-to-conversion rate prediction model is trained according to the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate, so as to obtain a target click-to-conversion prediction model that can accurately predict the click rate and conversion rate based on the sample user portrait and the sample page text; further, a calibration model is generated through the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate, so as to obtain a target click-to-conversion calibration model that can further accurately calibrate the click rate and conversion rate; finally, a target user portrait of the target object is obtained, the original page text information of the candidate page is obtained, and the candidate page is adjusted through the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait and the original page text information to obtain a target recommended page, so as to improve the click rate and conversion rate of the page.

[0083] In step S101 of some embodiments, the sample object is an object recommended to the sample recommendation page in the historical data, the sample user information is the information of the sample object, including but not limited to browsing history, click record, purchase record and identity information, the sample page information is the content of the sample recommendation page, and the sample page information includes the information of multiple recommended items in the sample recommendation page. The recommended items can be video recommendations, graphic recommendations, text recommendations and recommendations in combination thereof, and this application does not make specific restrictions. The sample click rate is the probability of the sample object clicking on the recommended items in the sample recommendation page after being recommended to the sample recommendation page. The click rate is calculated as the ratio between the number of clicks and the number of exposures. The number of exposures is the total number of times a recommended item appears on the sample recommendation page. The conversion rate is the ratio of the sample object to purchase, subscribe and other operations after clicking on a recommended item. The conversion rate is the ratio of the number of clicks to the number of purchase and subscription operations.

[0084] See also Figure 2 In some embodiments, the sample user information includes user identity information and user historical browsing information. Step S102 may include but is not limited to steps S201 to S203:

[0085] Step S201, clustering is performed according to user identity information and preset clustering areas to obtain user clustering information;

[0086] Step S202, performing preference identification based on the user's historical browsing information to obtain user preference information;

[0087] Step S203, generating a portrait based on the user clustering information and the user preference information to obtain a sample user portrait.

[0088] Steps S201 to S203 shown in the embodiment of the present application perform clustering based on user identity information and preset clustering areas to obtain user clustering information, then perform preference identification based on user historical browsing information to obtain user preference information, and finally perform portrait generation based on user clustering information and user preference information to obtain a sample user portrait, thereby accurately determining the user portrait of the sample object, thereby providing an accurate data basis for subsequent training of the target click-through conversion prediction model, so that the target click-through conversion prediction model can make accurate predictions of click-through rate and conversion rate.

[0089] In step S201 of some embodiments, the preset clustering area is a pre-set clustering area, and the clustering area is a specific area for dividing users into different areas or groups according to user identity information and preset conditions, and users in the same clustering area have similar characteristics and behavior patterns. Clustering is performed according to user identity information and preset clustering areas, and the clustering method is KNN clustering. First, the user identity information is converted into a vector representation, and then the Euclidean cosine value is calculated with the cluster center of the clustering area to obtain the cosine distance value, and the cluster center corresponding to the minimum cosine distance value is selected as the cluster center of the user identity information, and the user identity information is included in the cluster area corresponding to the cluster center corresponding to the minimum cosine distance value, and clustering is completed to obtain user clustering information. For example, the user identity information includes age information, and according to the age and the preset age group as the clustering area, it can be obtained that 40% of the users are in the age group of 40-50 years old. For example, the user identity information includes geographic location information, and according to the geographic location information and the preset province information, the geographical distribution of the user can be obtained.

[0090] In step S202 of some embodiments, preference identification is performed based on the user's historical browsing information, and the preference identification is association rule mining. For example, the user is accustomed to browsing the second product information after browsing the first product information, for example, the user is accustomed to browsing the first product information purchase interface after browsing the first product information recommendation interface. In one embodiment, the association rule mining algorithm is the Apriori algorithm, and the Apriori association rule mining is performed based on the user's historical browsing information to obtain the user preference information.

[0091] In step S203 of some embodiments, a user portrait is generated based on user clustering information and user preference information. The portrait is generated by merging information. For example, 40% of users are between 40 and 50 years old and are accustomed to browsing the second product information after browsing the first product information. After the portrait is generated, a sample user portrait is obtained.

[0092] See also Figure 3 In some embodiments, the sample page information includes recommended content information, and step S103 may include but is not limited to steps S301 to S303:

[0093] Step S301, extracting the content format according to the recommended content information to obtain at least one modal content data;

[0094] Step S302, extracting text topics from each modal content data to obtain at least one content text data;

[0095] Step S303: perform text content aggregation on each content text data to obtain sample page text information.

[0096] Steps S301 to S303 shown in the embodiment of the present application are to extract the content format according to the recommended content information to obtain at least one modal content data, then perform text topic extraction on each modal content data to obtain at least one content text data, and finally perform text content aggregation on each content text data to obtain sample page text information, thereby extracting the text information of the recommended content from the sample page information, thereby converting the data of multiple modal contents into text, so that the subsequent target click-through conversion prediction model only needs to process the text data without processing the data of multiple modalities. The exclusive processing of text data improves the prediction accuracy of click-through rate and conversion rate.

[0097] In step S301 of some embodiments, content format extraction is performed based on the recommended content information. For example, if the recommended information content includes video and text, text data and video data are obtained after format extraction, and the context positions of the video data and text data in the recommended information content are marked to obtain modal content data. In one embodiment, the recommended content information includes images, text and video, and content format extraction is performed on the recommended content information to obtain the content and position of the image, and the content and position of the image are used as a modal content data. The content format extraction principle of text and video is consistent with that of image, and will not be repeated here.

[0098] In step S302 of some embodiments, text topic extraction is performed for each modal content data. For example, if the modal content data is a video, topic extraction is performed on the video to obtain the topic of the video, and the topic of the video and the position of the video in the recommended content information are combined to obtain content text data.

[0099] In step S303 of some embodiments, all content text data are aggregated by using a pre-trained Transformer model. The content text data includes two parts of data. The first part of the data is the content text, and the second part of the data is the position information of the modal content data corresponding to the content text in the recommended content information. The content text and the position information are combined and input into the pre-trained Transformer model to obtain the sample page text information.

[0100] See also Figure 4 In some embodiments, step S104 may include but is not limited to steps S401 to S405:

[0101] Step S401, predicting the click-through conversion rate by using a preset click-through conversion rate prediction model, sample user portraits, and sample page text information to obtain a sample predicted click-through rate and a sample predicted conversion rate;

[0102] Step S402, calculating the loss value according to the sample predicted click rate and the sample click rate to obtain the click rate loss value;

[0103] Step S403, calculating the loss value according to the sample predicted conversion rate and the sample conversion rate to obtain the conversion rate loss value;

[0104] Step S404, performing loss value aggregation according to the click rate loss value and the conversion rate loss value to obtain a training loss value;

[0105] Step S405: Optimize the parameters of the preset click-through conversion rate prediction model based on the training loss value to obtain a target click-through conversion prediction model.

[0106] In steps S401 to S405 shown in the embodiment of the present application, click-through conversion rate is predicted by using a preset click-through conversion rate prediction model, sample user portraits, and sample page text information to obtain sample predicted click-through rate and sample predicted conversion rate, and then loss values ​​are calculated based on the sample predicted click-through rate and the sample click-through rate to obtain a click-through rate loss value, and then loss values ​​are calculated based on the sample predicted conversion rate and the sample conversion rate to obtain a conversion rate loss value, and then loss values ​​are aggregated based on the click-through rate loss value and the conversion rate loss value to obtain a training loss value, and finally, based on the training loss value, parameters of the preset click-through conversion rate prediction model are optimized to obtain a target click-through conversion prediction model, so that the target click-through conversion prediction model can well perceive the potential correlation between the sample user portrait and the sample page text information and the click-through rate and conversion rate, thereby being able to make accurate click-through rate and conversion rate predictions.

[0107] See also Figure 5In some embodiments, the preset click-to-conversion rate prediction model includes a click-through rate prediction network, a conversion rate prediction network, and a shared prediction network. Step S401 includes but is not limited to steps S501 to S505:

[0108] Step S501, performing click rate prediction on sample user portraits and sample page text information through a click rate prediction network to obtain original click rate prediction data;

[0109] Step S502, performing conversion rate prediction on the sample user portrait and the sample page text information through a conversion rate prediction network to obtain original conversion rate prediction data;

[0110] Step S503, performing shared prediction on the sample user portrait and the sample page text information through a shared prediction network to obtain original shared prediction data;

[0111] Step S504, performing aggregation calculation based on the original click rate prediction data and the original shared prediction data to obtain the sample predicted click rate;

[0112] Step S505 , performing aggregation calculation based on the original conversion rate prediction data and the original shared prediction data to obtain the sample predicted conversion rate.

[0113] In steps S501 to S505 shown in the embodiment of the present application, click rate prediction is performed on the sample user portrait and sample page text information through a click rate prediction network to obtain original click rate prediction data, and then conversion rate prediction is performed on the sample user portrait and sample page text information through a conversion rate prediction network to obtain original conversion rate prediction data, and then shared prediction is performed on the sample user portrait and sample page text information through a shared prediction network to obtain original shared prediction data, and then aggregation calculation is performed based on the original click rate prediction data and the original shared prediction data to obtain a sample predicted click rate, and finally aggregation calculation is performed based on the original conversion rate prediction data and the original shared prediction data to obtain a sample predicted conversion rate, thereby achieving accurate prediction of click rate and conversion rate.

[0114] In step S501 of some embodiments, the preset click-to-conversion rate prediction model is a collaborative gating control network (CGC network). In the CGC network, shared experts (Shared Experts), task-specific experts (Task-specific Experts) and a gating mechanism (Gating Mechanism) are included. The gating mechanism includes a task tower (Task Tower), which is particularly suitable for the prediction of joint information. Through the CGC network, the potential information in the click-through rate and conversion rate can be effectively combined to make predictions of the click-through rate and conversion rate. In the preset click-to-conversion rate prediction model, a click-through rate prediction network, i.e., a click-through rate task-specific expert, is included. By inputting sample user portraits and sample page text information into the click-through rate prediction network, the original click-through rate prediction data is obtained.

[0115] In step S502 of some embodiments, the preset click-through conversion rate prediction model includes a conversion rate prediction network, i.e., a dedicated expert for conversion rate tasks. The original conversion rate prediction data is obtained by inputting sample user portraits and sample page text information into the conversion rate prediction network.

[0116] In step S503 of some embodiments, a preset click-through conversion rate prediction model includes a shared prediction network, namely a shared expert, and original shared prediction data is obtained by inputting sample user portraits and sample page text information into the shared prediction network.

[0117] In step S504 of some embodiments, the preset click-through conversion rate prediction model includes a click-through rate task tower, which includes click-through rate prediction data weights and click-through rate shared prediction data weights. A weighted calculation is performed based on the click-through rate prediction data weights, the original click-through rate prediction data, the click-through rate shared prediction data weights and the original shared prediction data to obtain a sample predicted click-through rate.

[0118] In step S505 of some embodiments, the preset click-to-conversion rate prediction model includes a conversion rate task tower, which includes conversion rate prediction data weights and conversion rate shared prediction data weights. A weighted calculation is performed based on the conversion rate prediction data weights, the original conversion rate prediction data, the conversion rate shared prediction data weights, and the original shared prediction data to obtain a sample predicted conversion rate.

[0119] In step S402 of some embodiments, a loss value is calculated based on the sample predicted click rate and the sample click rate. The loss value may be calculated by cross entropy loss value calculation or mean square error loss value calculation, and this application does not impose any specific restrictions.

[0120] In step S403 of some embodiments, a loss value is calculated based on the sample prediction conversion rate and the sample conversion rate. The loss value may be calculated by a cross entropy loss value or a mean square error loss value, and this application does not impose any specific restrictions.

[0121] In step S404 of some embodiments, a loss value aggregation calculation is performed based on the click rate loss value and the conversion rate loss value to obtain a training loss value. The loss value aggregation calculation is shown in formula (1):

[0122] Loss total =α·loss CTR +β·loss CVR (1)

[0123] Among them, Loss total is the training loss value, α is the preset click rate loss value weight, β is the preset conversion rate loss value weight, loss CTR is the click rate loss value, loss CVR is the conversion rate loss value. It should be noted that α and β will be optimized together during the optimization process of the preset click-through conversion rate prediction model.

[0124] In step S405 of some embodiments, the parameters of the preset click-to-conversion rate prediction model are optimized according to the training loss value, that is, the preset click-to-conversion rate prediction model is back-propagated and gradient optimized based on the training loss value to obtain the target click-to-conversion rate prediction model.

[0125] See also Figure 6 In some embodiments, step S105 includes but is not limited to steps S601 to S604:

[0126] Step S601, predicting the click-through conversion rate of sample user portraits and sample page text information through a target click-through conversion prediction model to obtain an initial click-through rate and an initial conversion rate;

[0127] Step S602, performing parameter fitting on a preset click rate calibration model according to the initial click rate and the sample click rate to obtain a target click rate calibration model;

[0128] Step S603, performing parameter fitting on a preset conversion rate calibration model according to the initial conversion rate and the sample conversion rate to obtain a target conversion rate calibration model;

[0129] Step S604: merging the target click rate calibration model and the target conversion rate calibration model to obtain a target click-to-conversion calibration model.

[0130] In steps S601 to S604 shown in the embodiment of the present application, the click-through conversion rate of the sample user portrait and the sample page text information is predicted by the target click-through conversion prediction model to obtain the initial click-through rate and the initial conversion rate, and then the parameters of the preset click-through rate calibration model are fitted according to the initial click-through rate and the sample click-through rate to obtain the target click-through rate calibration model, and then the parameters of the preset conversion rate calibration model are fitted according to the initial conversion rate and the sample conversion rate to obtain the target conversion rate calibration model, and finally the models are merged based on the target click-through rate calibration model and the target conversion rate calibration model to obtain the target click-through conversion calibration model.

[0131] In step S601 of some embodiments, the sample user portrait and sample page text information are input into a target click-to-conversion prediction model, and the target click-to-conversion prediction model performs click-through rate prediction and conversion rate prediction to obtain an initial click-through rate and an initial conversion rate.

[0132] In step S602 of some embodiments, a preset fitting function is fitted by the initial click rate and the sample click rate to obtain a target click rate calibration model. The preset fitting function may be a linear fit, a polynomial fit, an exponential fit, a logarithmic fit or a power law fit, and this application does not impose any specific limitation.

[0133] In step S603 of some embodiments, a preset fitting function is fitted by the initial conversion rate and the sample conversion rate to obtain a target conversion rate calibration model. The preset fitting function may be a linear fit, a polynomial fit, an exponential fit, a logarithmic fit or a power law fit, and this application does not impose any specific limitation.

[0134] In step S604 of some embodiments, a model is merged based on the target click rate calibration model and the target conversion rate calibration model to obtain a target click-to-conversion calibration model. The model is merged to combine the two models together. When the initial click rate is input, the target click rate calibration model in the target click-to-conversion calibration model is calibrated. When the initial conversion rate is input, the target conversion rate calibration model in the target click-to-conversion calibration model is calibrated. It should be noted that the target click-to-conversion calibration model is trained to avoid the phenomenon of overfitting of the target click-to-conversion prediction model. The click rate prediction value of the target click-to-conversion prediction model does not need to be completely consistent with the sample click rate, and the conversion rate prediction value of the target click-to-conversion prediction model does not need to be completely consistent with the sample click rate. Instead, it is further calibrated by the target click-to-conversion calibration model to avoid the phenomenon of overfitting of the target click-to-conversion prediction model, so that the target click-to-conversion prediction model has higher versatility.

[0135] In step S106 of some embodiments, the principle of obtaining the target user portrait of the target object is consistent with the principle of generating a portrait based on sample user information to obtain a sample user portrait, which is not repeated here. The principle of obtaining the original page text information of the candidate page is consistent with the principle of extracting content based on sample page information to obtain sample page text information, which is not repeated here.

[0136] See also Figure 7 In some embodiments, step S106 may include but is not limited to steps S701 to S704:

[0137] Step S701, using a target click-through conversion prediction model to perform click-through conversion prediction on the target user portrait and the original page text information, and obtain an initial predicted click-through rate and an initial predicted conversion rate of the candidate page;

[0138] Step S702, calibrating the initial predicted click rate through the target click-through conversion calibration model to obtain a target predicted click rate;

[0139] Step S703, calibrating the initial predicted conversion rate through the target click-through conversion calibration model to obtain a target predicted conversion rate;

[0140] Step S704, optimizing the candidate pages by using the target predicted click rate and the target predicted conversion rate to obtain the target recommended page.

[0141] In steps S701 to S704 shown in the embodiment of the present application, click conversion prediction is performed on the target user portrait and the original page text information through the target click-through conversion prediction model to obtain the initial predicted click-through rate and the initial predicted conversion rate of the candidate page, and then the initial predicted click-through rate is calibrated through the target click-through conversion calibration model to obtain the target predicted click-through rate, and then the initial predicted conversion rate is calibrated through the target click-through conversion calibration model to obtain the target predicted conversion rate, and finally the candidate page is optimized according to the target predicted click-through rate and the target predicted conversion rate to obtain the target recommended page, thereby achieving page optimization of the candidate page, thereby improving the click-through rate and conversion rate of the recommended page.

[0142] In step S701 of some embodiments, the target user portrait and the original page text information are input into a target click-to-conversion prediction model to obtain an initial predicted click rate and an initial predicted conversion rate output by the target click-to-conversion prediction model.

[0143] In step S702 of some embodiments, the initial predicted click rate is calibrated using a target click rate calibration model in a target click-through conversion calibration model to obtain a target predicted click rate.

[0144] In step S703 of some embodiments, the initial predicted conversion rate is calibrated using a target conversion rate calibration model in a target click-to-conversion calibration model to obtain a target predicted conversion rate.

[0145] In step S704 of some embodiments, the target predicted click rate is filtered by a preset click rate threshold, and the target predicted conversion rate is filtered by a preset conversion rate threshold. When the target predicted click rate is greater than the preset click rate threshold and the target predicted conversion rate is greater than the preset conversion rate threshold, the candidate page passes the screening and obtains the target recommended page.

[0146] See also Figure 8 The embodiment of the present application also provides a recommendation page generation device, which can implement the above-mentioned recommendation page generation method, and the device includes:

[0147] The information acquisition module 801 is used to acquire sample user information of the sample object, acquire sample page information of the sample recommendation page, and acquire the sample click rate and sample conversion rate of the sample object on the sample recommendation page;

[0148] A portrait generation module 802 is used to generate a portrait according to the sample user information to obtain a sample user portrait;

[0149] The content extraction module 803 is used to extract content according to the sample page information to obtain the sample page text information;

[0150] Model training module 804, used to train a preset click-to-conversion rate prediction model according to sample user portraits, sample page text information, sample click-through rates, and sample conversion rates, to obtain a target click-to-conversion prediction model;

[0151] Model generation module 805, used to generate a calibration model through the target click-through conversion prediction model, sample user portraits, sample page text information, sample click-through rate and sample conversion rate, to obtain a target click-through conversion calibration model;

[0152] The page adjustment module 806 is used to obtain the target user portrait of the target object, obtain the original page text information of the candidate page, and adjust the candidate page through the target click-through conversion calibration model, the target click-through conversion prediction model, the target user portrait and the original page text information to obtain the target recommended page.

[0153] The specific implementation of the recommendation page generation device is substantially the same as the specific implementation of the recommendation page generation method described above, and will not be described in detail herein.

[0154] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned recommendation page generation method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0155] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0156] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0157] The memory 902 may be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 may store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the recommendation page generation method of the embodiment of this application;

[0158] Input / output interface 903, used to implement information input and output;

[0159] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0160] A bus 905 that transmits information between the various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0161] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0162] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned recommendation page generation method is implemented.

[0163] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0164] The recommendation page generation method, recommendation page generation device, electronic device and storage medium provided in the embodiments of the present application obtain sample user information of a sample object, obtain sample page information of a sample recommendation page, and obtain sample click rate and sample conversion rate of the sample object for the sample recommendation page, then generate a portrait according to the sample user information to obtain a sample user portrait, extract content according to the sample page information to obtain sample page text information, and then train a preset click-to-conversion rate prediction model according to the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate, thereby obtaining a target click-to-conversion prediction model that can accurately predict the click rate and conversion rate based on the sample user portrait and the sample page text; further, generate a calibration model through the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate, and obtain a target click-to-conversion calibration model that can further accurately calibrate the click rate and conversion rate; finally, obtain a target user portrait of the target object, obtain the original page text information of the candidate page, and adjust the candidate page through the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait and the original page text information to obtain a target recommendation page, thereby improving the click rate and conversion rate of the page.

[0165] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0166] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0167] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0168] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0169] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0170] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0171] In the several embodiments provided in the present application, it should be understood that the disclosed 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 above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0172] The units described above 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 solution of this embodiment.

[0173] In addition, each functional unit in each embodiment of the present application 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.

[0174] 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 application, 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 storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0175] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for generating a recommendation page, characterized in that: The method comprises: Obtain sample user information of the sample object, obtain sample page information of the sample recommendation page, and obtain the sample click rate and sample conversion rate of the sample object on the sample recommendation page; Generate a sample user profile based on the sample user information to obtain a sample user profile; Extracting content according to the sample page information to obtain sample page text information; The preset click-to-conversion rate prediction model is trained according to the sample user portrait, the sample page text information, the sample click-through rate and the sample conversion rate to obtain a target click-to-conversion prediction model; A calibration model is generated by using the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate to obtain a target click-to-conversion calibration model; The target user portrait of the target object is obtained, the original page text information of the candidate page is obtained, and the candidate page is adjusted through the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait and the original page text information to obtain the target recommended page.

2. The method according to claim 1, characterized in that The step of training a preset click-to-conversion rate prediction model according to the sample user portrait, the sample page text information, the sample click-through rate, and the sample conversion rate to obtain a target click-to-conversion prediction model includes: The click-through conversion rate is predicted by using the preset click-through conversion rate prediction model, the sample user portrait and the sample page text information to obtain a sample predicted click-through rate and a sample predicted conversion rate; Calculate the loss value according to the sample predicted click rate and the sample click rate to obtain the click rate loss value; Calculate the loss value according to the sample predicted conversion rate and the sample conversion rate to obtain a conversion rate loss value; Perform loss value aggregation according to the click rate loss value and the conversion rate loss value to obtain a training loss value; The preset click-to-conversion rate prediction model is optimized based on the training loss value to obtain the target click-to-conversion prediction model.

3. The method according to claim 2, characterized in that The preset click-to-conversion rate prediction model includes a click-to-conversion rate prediction network, a conversion rate prediction network, and a shared prediction network; the click-to-conversion rate prediction is performed by using the preset click-to-conversion rate prediction model, the sample user portrait, and the sample page text information to obtain a sample predicted click rate and a sample predicted conversion rate, including: Perform click rate prediction on the sample user portrait and the sample page text information through the click rate prediction network to obtain original click rate prediction data; Performing conversion rate prediction on the sample user portrait and the sample page text information through the conversion rate prediction network to obtain original conversion rate prediction data; Performing shared prediction on the sample user portrait and the sample page text information through the shared prediction network to obtain original shared prediction data; Performing aggregation calculation based on the original click rate prediction data and the original shared prediction data to obtain the sample predicted click rate; Aggregate calculation is performed based on the original conversion rate prediction data and the original shared prediction data to obtain the sample predicted conversion rate.

4. The method according to any one of claims 1 to 3, characterized in that: The generating of the calibration model by using the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate to obtain the target click-to-conversion calibration model includes: The click-through conversion rate of the sample user portrait and the sample page text information is predicted by the target click-through conversion prediction model to obtain an initial click-through rate and an initial conversion rate; Performing parameter fitting on a preset click rate calibration model according to the initial click rate and the sample click rate to obtain a target click rate calibration model; Performing parameter fitting on a preset conversion rate calibration model according to the initial conversion rate and the sample conversion rate to obtain a target conversion rate calibration model; Model merging is performed based on the target click rate calibration model and the target conversion rate calibration model to obtain the target click-to-conversion calibration model.

5. The method according to any one of claims 1 to 3, characterized in that: The sample user information includes user identity information and user browsing history information; the profile generation based on the sample user information to obtain the sample user profile includes: Performing clustering according to the user identity information and the preset clustering area to obtain user clustering information; Perform preference identification based on the user's historical browsing information to obtain user preference information; A portrait is generated according to the user clustering information and the user preference information to obtain the sample user portrait.

6. The method according to any one of claims 1 to 3, characterized in that: The sample page information includes recommended content information; and the content extraction based on the sample page information to obtain the sample page text information includes: Extracting a content format according to the recommended content information to obtain at least one modal content data; Performing text topic extraction on each of the modal content data to obtain at least one content text data; Text content aggregation is performed on each of the content text data to obtain the sample page text information.

7. The method according to any one of claims 1 to 3, characterized in that: The step of adjusting the candidate page by using the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait, and the original page text information to obtain a target recommended page includes: Performing click-through conversion prediction on the target user portrait and the original page text information through the target click-through conversion prediction model to obtain an initial predicted click-through rate and an initial predicted conversion rate of the candidate page; Calibrate the initial predicted click rate by using the target click-through conversion calibration model to obtain a target predicted click rate; Calibrate the initial predicted conversion rate using the target click-through conversion calibration model to obtain a target predicted conversion rate; The candidate pages are optimized according to the target predicted click-through rate and the target predicted conversion rate to obtain the target recommended page.

8. A recommendation page generating device, characterized in that: The device comprises: An information acquisition module is used to acquire sample user information of a sample object, acquire sample page information of a sample recommendation page, and acquire a sample click rate and a sample conversion rate of the sample object on the sample recommendation page; A portrait generation module, used to generate a portrait based on the sample user information to obtain a sample user portrait; A content extraction module, used to extract content according to the sample page information to obtain sample page text information; A model training module is used to train a preset click-to-conversion rate prediction model according to the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate to obtain a target click-to-conversion prediction model; A model generation module, used to generate a calibration model through the target click-to-conversion prediction model, the sample user portrait, the sample page text information, the sample click rate and the sample conversion rate, to obtain a target click-to-conversion calibration model; The page adjustment module is used to obtain the target user portrait of the target object, obtain the original page text information of the candidate page, and adjust the candidate page through the target click-to-conversion calibration model, the target click-to-conversion prediction model, the target user portrait and the original page text information to obtain the target recommended page.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the recommendation page generation method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the recommendation page generating method according to any one of claims 1 to 7 is implemented.

Images