A model training and map-based hierarchical recommendation method and device

Abstract

This specification discloses a model training and map-based hierarchical recommendation method and apparatus. Training samples are determined based on acquired user data, map-level data, and data on each recommended object. Labels for the training samples are determined based on the user's historical data at different map levels. Each training sample is input into a click-through rate prediction model to be trained, and the click-through rate of each training sample is determined. The click-through rate prediction model is then trained based on the click-through rate and labels of each training sample. This method determines training samples corresponding to different map levels based on each map-level data, enabling the trained click-through rate prediction model to recommend different content to different users at different scales, providing greater flexibility and improved recommendation efficiency.

Description

Technical Field

[0001] This specification relates to the field of computer technology, and in particular to a model training and map-based hierarchical recommendation method and device. Background Art

[0002] With the advancement of computer technology, selecting the most valuable information from vast amounts of data to recommend suitable products to users has become a challenge for service providers. Map-based recommendation methods are widely used in service providers because they can recommend different content at different scales.

[0003] In the prior art, a commonly used map-based recommendation method determines recommended content for users at different scales. When the user zooms in or out of the map, the recommended content corresponding to the current scale is displayed to the user. The recommended content includes landmarks, shopping malls, and food.

[0004] However, when determining the content recommended to the user at each scale, the prior art recommends fixed content for each scale, which makes the prior art inflexible and inefficient in making recommendations. Summary of the Invention

[0005] This specification provides a model training and map-based hierarchical recommendation method and device to partially solve the above-mentioned problems existing in the prior art.

[0006] This manual adopts the following technical solutions:

[0007] This specification provides a training method for a click-through rate prediction model, including:

[0008] Determine each training sample based on the acquired user data, each map level data, and each recommended object data;

[0009] Determine the labeling of each training sample based on each user's historical data at different map levels;

[0010] Taking each training sample as input, the click-through rate prediction model to be trained is used to determine the click-through rate corresponding to each training sample;

[0011] The click-through rate prediction model is trained based on the click-through rate and the annotation of each training sample. The click-through rate prediction model is used to determine the click-through rate of each user on each recommended object at each level of the map.

[0012] Optionally, obtain the map-level data for each region, including:

[0013] For each map level, determine the regional popularity of the area to which each recommended object in the map level belongs, and determine the map level data of the map level based on the regional popularity;

[0014] Combine the user data, the local map level data, and the recommended object data to determine the training samples.

[0015] Optionally, obtaining the data of each recommended object includes:

[0016] For each recommended object, determine the click distance corresponding to the recommended object and the popularity corresponding to the recommended object, and determine the recommended object data corresponding to the recommended object based on the click distance and the popularity, and the popularity is determined at least based on the historical click volume of the recommended object.

[0017] Optionally, the training samples are determined based on the acquired user data, the map-level data, and the recommended object data, specifically including:

[0018] For each map level, determine the regional popularity of the area corresponding to each recommended object, and for each recommended object, determine the map level data corresponding to the recommended object based on the regional popularity corresponding to the recommended object;

[0019] The training samples are determined according to the user data, the recommended object data and the map level data corresponding to the recommended objects.

[0020] Optionally, the labeling of each training sample is determined based on the historical data of each user at different map levels, specifically including:

[0021] For each user, based on the user's historical data at different map levels, determine the recommended objects that the user has clicked on at different map levels;

[0022] A training sample including the user data of the user, the recommended object data of the recommended object on which the user has clicked, and the map-level data of the map level when the user clicks the recommended object is marked as a positive sample.

[0023] This specification provides a map-based hierarchical recommendation method, including:

[0024] receiving a recommendation request carrying user data and map-level data, and determining the user data, the map-level data, and each recommended object data according to the recommendation request;

[0025] For each piece of recommended object data, the recommended object data, the user data, and the map-level data are input into a pre-trained click-through rate prediction model, and the click-through rate of the user on the recommended object data at the map level, as output by the click-through rate prediction model, is determined;

[0026] Sort the data of each recommended object according to the determined click-through rates;

[0027] According to the ranking, recommended content is determined.

[0028] This specification provides a training device for a click-through rate prediction model, comprising:

[0029] The sample determination module is used to determine each training sample based on the acquired user data, each region level data and each recommended object data;

[0030] The label determination module is used to determine the label of each training sample based on the historical data of each user at different map levels;

[0031] The click-through rate prediction module is used to take each training sample as input and determine the click-through rate corresponding to each training sample through the click-through rate prediction model to be trained;

[0032] The training module is used to train the click-through rate prediction model based on the click-through rate and annotation of each training sample. The click-through rate prediction model is used to determine the click-through rate of each user on each recommended object at each level of the map.

[0033] This specification provides a training device for a click-through rate prediction model, comprising:

[0034] a first determining module, configured to receive a recommendation request carrying user data and map-level data, and determine the user data, the map-level data, and each recommended object data according to the recommendation request;

[0035] a second determination module, configured to, for each piece of recommended object data, input the recommended object data, the user data, and the map-level data into a pre-trained click-through rate prediction model, and determine a click-through rate of the user on the recommended object data at the map level, as output by the click-through rate prediction model;

[0036] A sorting module is used to sort the recommended object data according to the determined click-through rates;

[0037] The recommendation module is used to determine the recommended content according to the ranking.

[0038] The computer-readable storage medium provided in this specification stores a computer program, which, when executed by a processor, implements any of the above-mentioned training methods for the click-through rate prediction model or the map-based hierarchical recommendation method.

[0039] The electronic device provided in this specification includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements any of the above-mentioned methods for training a click-through rate prediction model or a map-based hierarchical recommendation method.

[0040] At least one of the above technical solutions adopted in this specification can achieve the following beneficial effects:

[0041] In the training method of the click-through rate prediction model provided in this specification, training samples are determined based on the acquired user data, map level data and each recommended object data, and the annotations of the training samples are determined based on the user's historical data in different map levels. Each training sample is input into the click-through rate prediction model to be trained, the click-through rate of each training sample is determined, and the click-through rate prediction model is trained based on the click-through rate and annotations of each training sample.

[0042] As can be seen from the above method, this method determines the training samples corresponding to different map levels based on the map level data of each region, so that the click-through rate prediction model obtained by training can recommend different content to users based on different scales, which is more flexible and has higher recommendation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The exemplary embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation of this specification. In the drawings:

[0044] Figure 1 A flowchart of the training method for the click-through rate prediction model provided in this manual;

[0045] Figure 2 A flowchart of the map-based hierarchical recommendation method provided in this manual;

[0046] Figure 3 A schematic diagram of the training device for the click-through rate prediction model provided in this manual;

[0047] Figure 4 Schematic diagram of the map-based hierarchical recommendation device provided in this manual;

[0048] Figure 5 The corresponding Figure 1 or Figure 2 Schematic diagram of electronic equipment. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of this specification more clear, the following will clearly and completely describe the technical solutions of this specification in conjunction with the specific embodiments of this specification and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this specification.

[0050] The technical solutions provided by the embodiments of this specification are described in detail below with reference to the accompanying drawings.

[0051] Figure 1 This is a flowchart of the training method for the click-through rate prediction model provided in this specification, which specifically includes the following steps:

[0052] S100: Determine each training sample based on the acquired user data, each map level data and each recommended object data.

[0053] Typically, a click-through rate prediction model is pre-trained based on training samples by a server used for training the model. This specification provides a method for training a click-through rate prediction model. Similarly, the process of training the click-through rate prediction model can be performed by a server used for training the model.

[0054] The training model can be divided into a sample generation phase and a model training phase. In the sample generation phase, samples for training the model can be determined based on model requirements and training needs. In this specification, the server can first determine training samples for training the click-through rate prediction model. Since the click-through rate prediction model generally determines the click-through rate of the recommended object based on the user data of each user and the object data of each recommended object, the server can first determine the user data, the local level data, and the data of each recommended object to determine the training samples.

[0055] Based on this, the server can obtain data for each user, data for each map level, and data for each recommended object. The user data can include profile data for each user, or data such as user reviews of each product. The recommended objects are merchants, and the recommended object data includes merchant reviews, likes, popularity, and geographic location. Each map level represents the different scales of the map, with different levels corresponding to different scales, and generally, the higher the level, the smaller the scale.

[0056] Then, the server may combine the user data, each map level data, and each recommended object data to determine each training sample. In other words, a training sample includes the user data of a user, the map level data, and the recommended object data.

[0057] In addition, the user data of each user may be determined by the server based on the user profile of the user. The map level information of each region may be determined by the server for each map level, and the regional popularity of the region to which each recommended object belongs in the map level is determined based on the regional popularity. The data of each recommended object may be determined by the server for each recommended object, and the distance between the recommended object and each user, as well as the recommendation popularity corresponding to the recommended object, and determined based on the distance and the recommendation popularity. The recommendation popularity is determined based on at least the historical click volume of the recommended object, and the recommendation popularity is positively correlated with the historical click volume.

[0058] It should be noted that for each training sample, the training sample can be a feature determined by fusing user data, map-level data, and recommended object data. The fusion method can be splicing, addition, etc. Fusion of multiple features to determine a single feature is a relatively mature technology and is not limited in this specification.

[0059] S102: Determine the labeling of each training sample based on the historical data of each user at different map levels.

[0060] In one or more embodiments provided in this specification, model training can be performed based on training samples and their annotations. Therefore, after determining the training samples, the server can determine the annotations of each training sample based on the historical data of each user in different map levels.

[0061] Specifically, the server may determine, for each map level, the historical data of each user in the map level.

[0062] Afterwards, the server may determine the historical click operations of each user on each recommended object in the map layer based on the determined historical data of each user in the map layer.

[0063] Finally, based on the determined historical click operations of each user on each recommended object at the map level, the server can use the training samples corresponding to the map level, consisting of each user and the recommended object they clicked, as positive samples. The remaining training samples are used as negative samples. Positive samples are marked as clicked, and negative samples are marked as not clicked.

[0064] Of course, the training samples can be labeled as clicked, not clicked, or represented by numbers, such as 1 for clicked and 0 for not clicked. The loss can then be determined based on the numerical labels.

[0065] In addition, when determining the labels for training samples, the server can also match the determined hierarchical data, each user and the recommended objects they have clicked, with each training sample, and determine the training sample containing the recommended object clicked by each user in the hierarchy as a positive sample. In other words, it determines that it is labeled as clicked.

[0066] S104: Taking each training sample as input, and determining the click rate corresponding to each training sample through the click rate prediction model to be trained.

[0067] In one or more embodiments provided herein, the click-through rate prediction model may be a deep learning model, and the server may train the click-through rate prediction model based on the click-through rates corresponding to each training sample and the labels of each training sample. Based on this, after determining each training sample and its labels, the server may determine the click-through rate corresponding to each training sample, thereby training the model based on the click-through rates of each training sample.

[0068] Specifically, in this specification, for each training sample, the server may use the training sample as input into a click-through rate prediction model to be trained to determine the click-through rate of the training sample.

[0069] Among them, the model structure of the click-through rate prediction model may include multiple network layers such as convolutional neural networks and recurrent neural networks. Of course, the specific model structure and model parameters can be set as needed, and this manual does not limit this.

[0070] S106: Training the click-through rate prediction model based on the click-through rate and the annotation of each training sample. The click-through rate prediction model is used to determine the click-through rate of each user for each recommended object at each level of each map.

[0071] In one or more embodiments provided in this specification, after determining the click-through rate of each training sample, the server may train the click-through rate prediction model based on the click-through rate of each training sample and its annotation.

[0072] Specifically, the server can determine the click-through rate and label of each training sample, and adjust the model parameters to train the click-through rate prediction model with the minimum difference between the label and the click-through rate as the optimization goal. Of course, the specific loss function used can be set as needed.

[0073] Additionally, the CTR prediction model can be a linear model or a tree model. If the CTR prediction model is a linear model, the server can construct the variables in the linear regression model based on user data, map-level data, and recommended object data. If the CTR prediction model is a tree model, the server can construct a decision tree based on user data, map-level data, and recommended object data. Of course, the specific structure and training method of the CTR prediction model can be customized as needed and are not limited in this specification.

[0074] based on Figure 1 A method for training a click-through rate prediction model determines training samples based on acquired user data, map-level data, and data on each recommended object. The training samples are labeled based on the user's historical data at different map levels. Each training sample is then fed into the trained click-through rate prediction model to determine the click-through rate of each training sample. The model is then trained based on the click-through rate and labels of each training sample. This method determines training samples corresponding to different map levels based on each map-level data, enabling the trained click-through rate prediction model to recommend different content to users at different scales, providing greater flexibility and improved recommendation efficiency.

[0075] Furthermore, the click distance in the user profile information usually has a significant impact on the click-through rate of each recommended object. Therefore, in order to train an accurate click-through rate prediction model, in step S100, when determining the user characteristics of each user, the user's historical click data can also be determined for each user, and the user's average historical click distance can be determined based on the historical click data, and the user data of the user can be determined based on the click distance.

[0076] Furthermore, the popularity of different recommended objects usually affects the click-through rate of each recommended object. Therefore, in step S100, for each recommended object, the server can determine the historical click count, historical click-through rate, number of comments, number of likes, exposure, etc. of the recommended object, determine the popularity corresponding to the recommended object, and determine the recommended object data based on the popularity.

[0077] In addition, the click distance of the recommended object has a greater impact on the click rate of the recommended object. Therefore, when determining the data of each recommended object, the server can also determine the data of each recommended object based on the click distance of each recommended object.

[0078] The click distance of a recommended object can be determined based on the distance between the user who clicked the recommended object and the recommended object when the user clicked the recommended object. For example, if the distance between user A and recommended object B is 2 km when user A clicks on recommended object B, the click distance of the recommended object is 2 km. Of course, in the case where multiple users have clicked on the recommended object, the server can determine the click distances of the recommended object separately and use the maximum value, minimum value, median, average, etc. of the determined click distances as the click rate of the recommended object. The specific method for determining the click distance can be set as needed, and this manual does not impose any restrictions on this.

[0079] Furthermore, there is a correspondence between the popularity and click distance of the recommended object. Therefore, when determining the data of the recommended object of the recommended object, the server can first determine the popularity and click distance of each recommended object respectively, and establish a correspondence between the popularity and click distance corresponding to each recommended object. Then, the server can count each corresponding relationship and determine the click distance corresponding to each popularity. For each recommended object, the server can determine the data of each recommended object based on the popularity corresponding to the recommended object and the click distance corresponding to the popularity. Among them, the form of popularity can include a numerical value or a range to which the popularity belongs. Of course, the specific form of popularity can be set as needed, and this manual does not limit this.

[0080] Furthermore, there is also a correspondence between the type of the recommended object and its corresponding click distance. Therefore, the server can first determine the type and click distance of each recommended object respectively, and establish a correspondence between the type and click distance corresponding to each recommended object. Then, the server can count each corresponding relationship and determine the click distance corresponding to each type. For each recommended object, the server can determine the data of each recommended object based on the type of the recommended object and the click distance corresponding to the type. Among them, the types of recommended objects may include various types such as attractions, transportation facilities, hotels, food, pets, fitness, etc. Of course, the specific type can be set as needed, and this manual does not limit this.

[0081] In addition, the click-through rate of recommended objects is often affected by the surrounding area. Even at the same map level, the click-through rates of recommended objects of the same type in different areas are often different. For example, the click-through rates of convenience stores around scenic spots and convenience stores around residential communities are also different. Therefore, in step S100, the server may divide each map level into different areas and, when determining training samples, determine the training samples based on the map level and the regional popularity and regional merchant click-through rates of the areas to which the recommended objects contained in the training samples belong.

[0082] Among them, the regional heat is the sum or average heat of each recommended object contained in the area, and the regional merchant click-through rate is the sum or average click-through rate of each recommended object contained in the area. Of course, the method of determining the regional heat and regional merchant click-through rate can be set as needed, and the area division method can be based on preset shapes, or according to "county, district, street" and other methods. The specific division method can be set as needed, and this manual does not limit this.

[0083] Of course, when determining training samples, we can also directly determine the regional popularity of the areas to which each recommended object belongs at each map level, and determine the map-level data for that map level based on the regional popularity. We then combine the user data, the regional map-level data, and the recommended object data to determine the training samples. In other words, for each map level, the map-level data is determined based on the areas contained in that map level.

[0084] In addition, when recommending highly popular objects to users at a smaller scale, the corresponding recommendation efficiency may be higher. Therefore, in order to make accurate recommendations, the server can also determine the regional popularity of the area corresponding to each recommended object at each map level.

[0085] Then, the server can determine the map level data corresponding to each recommended object based on the regional popularity of the recommended object and the predetermined correspondence between regional popularity and click distance. Click distance and map level are positively correlated, that is, the farther the click distance, the higher the map level and the smaller the scale.

[0086] Finally, the server may determine each training sample according to each user, each recommended object data, and the map level data corresponding to each recommended object.

[0087] based on Figure 1 The present specification also provides a map-based hierarchical recommendation method, such as Figure 2 shown.

[0088] Figure 2 The flowchart of the map-based hierarchical recommendation method provided for this specification includes:

[0089] S200: receiving a recommendation request carrying user data and map-level data, and determining the user data, the map-level data, and each recommended object data according to the recommendation request.

[0090] Generally, in the field of content recommendation, a recommendation method can be used to select, based on user data in a received recommendation request, a recommended object that is likely to attract a user's click from among various recommended objects and present it to the user, thereby recommending the recommended object to the user.

[0091] Upon receiving a user's recommendation request, the server can determine the user data, map-level data, and recommended object data contained in the recommendation request. The recommendation request is sent by the terminal to the server when the user zooms in or out of the map. For a description of the user data, map-level data, and recommended object data, refer to the aforementioned training method for the click-through rate prediction model, and will not be further elaborated in this specification.

[0092] S202: For each recommended object data, the recommended object data, the user data and the map level data are input into a pre-trained click rate prediction model to determine the click rate of the user on the recommended object data at the map level output by the click rate prediction model.

[0093] S204: Sort the recommended objects according to the determined click rates.

[0094] S206: Determine recommended content based on the ranking.

[0095] In one or more embodiments provided in this specification, the step of determining the click rate of the user on the recommended object at the map level can refer to the description of the training method of the above-mentioned click rate prediction model.

[0096] After determining the click rate of each recommended object, the server can sort the recommended objects and determine the recommended content based on the determined sorting, and recommend it to the user. The number of recommended content can be set as needed, and this manual does not impose any restrictions on this.

[0097] The click-through rate prediction model in this specification estimates the probability of users clicking on recommended objects at different map levels. However, due to the limited field of view of the terminal, the click-through rate prediction model actually estimates the recommendation probability of each recommended object around the terminal user at different map levels determined by the preset size and shape with the terminal user's position as the center. For example, user A holds a terminal, and there is recommended object 1 100 meters away from user A, and recommended object 2 1000 meters away from user A. Assuming that the probability of each recommended object within 500 meters of user A is determined at the current map level, it is only necessary to determine the probability of recommended object 1. Therefore, the server can also determine the relationship between the map level and the distance between the recommended object and the user.

[0098] When determining the recommended object data, the recommended object data around the user may be determined based on the map hierarchy, and the recommended object data may be re-determined when the user moves or zooms the map.

[0099] The above is a training method for a click-through rate prediction model and a map-based hierarchical recommendation method provided in one or more embodiments of this specification. Based on the same idea, this specification also provides a corresponding training device for a click-through rate prediction model and a map-based hierarchical recommendation device, such as Figure 3 and Figure 4 shown.

[0100] Figure 3 A schematic diagram of a training device for a click-through rate prediction model provided in this specification, specifically including:

[0101] The sample determination module 300 is used to determine each training sample according to the acquired user data, each region level data and each recommended object data.

[0102] The label determination module 302 is used to determine the label of each training sample based on the historical data of each user at different map levels.

[0103] The click rate prediction module 304 is configured to take each training sample as input and determine the click rate corresponding to each training sample through the click rate prediction model to be trained.

[0104] The training module 306 is used to train the click-through rate prediction model based on the click-through rate and the labels of each training sample. The click-through rate prediction model is used to determine the click-through rate of each user for each recommended object at each level.

[0105] Optionally, the sample determination module 300 is used to determine the regional popularity of the area to which each recommended object in each map layer belongs, and determine the map level data of the map level based on the regional popularity, and combine the user data, the regional map level data, and the recommended object data to determine each training sample.

[0106] Optionally, the sample determination module 300 is used to determine, for each recommended object, the click distance corresponding to the recommended object and the popularity corresponding to the recommended object, and determine the recommended object data corresponding to the recommended object based on the click distance and the popularity, wherein the popularity is determined at least based on the historical click volume of the recommended object.

[0107] Optionally, the sample determination module 300 is used to determine the regional popularity of the area corresponding to each recommended object for each map level, and for each recommended object, determine the map level data corresponding to the recommended object based on the regional popularity corresponding to the recommended object, and determine each training sample based on each user data, each recommended object data and each recommended object's corresponding map level data.

[0108] Optionally, the labeling determination module 302 is used to determine, for each user, the recommended objects on which the user has clicked in different map levels based on the user's historical data in different map levels, and label as positive samples the training samples that include the user data of the user, the recommended object data on which the user has clicked, and the map level data of the map level when the user clicked the recommended object.

[0109] Figure 4 The schematic diagram of the map-based hierarchical recommendation device provided for this specification includes:

[0110] The first determining module 400 is configured to receive a recommendation request carrying user data and map-level data, and determine the user data, the map-level data, and each recommended object data according to the recommendation request.

[0111] The second determination module 402 is used to input, for each recommended object data, the recommended object data, the user data and the map level data into a pre-trained click-through rate prediction model, and determine the click-through rate of the user on the recommended object data at the map level output by the click-through rate prediction model.

[0112] The sorting module 404 is configured to sort the recommended object data according to the determined click rates.

[0113] The recommendation module 406 is configured to determine recommended content based on the ranking.

[0114] This specification also provides a computer-readable storage medium, which stores a computer program that can be used to execute the above Figure 1 The training method of the click rate prediction model provided is the same as the above Figure 2 At least one of the provided map-based hierarchical recommendation methods.

[0115] This manual also provides Figure 5 The schematic structure diagram of the electronic device shown in FIG. Figure 5As mentioned above, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 The training method of the click rate prediction model and the above Figure 2 A map-based hierarchical recommendation method is provided. Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0116] In the 1990s, technological improvements could be clearly distinguished as either hardware improvements (for example, improvements to circuit structures like diodes, transistors, and switches) or software improvements (improvements to process flows). However, with the advancement of technology, many process flow improvements today can now be considered direct improvements to hardware circuit structures. Designers almost always create the corresponding hardware circuit structure by programming the improved process flow into the hardware circuit. Therefore, it cannot be said that a process flow improvement cannot be implemented using hardware modules. For example, a programmable logic device (PLD), such as a field programmable gate array (FPGA), is an integrated circuit whose logical function is determined by user programming. Designers can "integrate" a digital system on a PLD through their own programming, without having to hire a chip manufacturer to design and manufacture a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly done using "logic compiler" software. This is similar to the software compiler used when developing programs. Before compilation, the original code must also be written in a specific programming language, called a hardware description language (HDL). There is not just one HDL, but many, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art will also understand that by simply programming the method flow in one of these hardware description languages ​​and then programming it into an integrated circuit, a hardware circuit that implements the logic method flow can be easily obtained.

[0117] The controller can be implemented in any suitable manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to implementing the controller in a purely computer-readable program code format, the controller can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules that implement the method and structures within the hardware component.

[0118] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0119] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0120] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0122] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0123] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0124] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0125] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0126] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0127] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0128] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Thus, this specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] This specification may be described in the general context of computer-executable instructions, such as program modules, executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including storage devices.

[0130] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0131] The foregoing is merely an example of the present invention and is not intended to limit the present invention. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

Claims

1. A training method for a click-through rate prediction model, characterized in that: include: Determine each training sample based on the acquired user data, each map level data, and each recommended object data; Determine the labeling of each training sample based on each user's historical data at different map levels; Taking each training sample as input, determining the click-through rate corresponding to each training sample through the click-through rate prediction model to be trained; The click-through rate prediction model is trained based on the click-through rate and the annotation of each training sample, and the click-through rate prediction model is used to determine the click-through rate of each recommended object by each user at each level of the map; Based on the acquired user data, local map data, and recommended object data, the training samples are determined, including: For each map level, determine the regional popularity of the area to which each recommended object in the map level belongs, and determine the map level data of the map level based on the regional popularity; Combine the user data, the local map level data, and the recommended object data to determine the training samples.

2. The method according to claim 1, wherein Get the data of each recommended object, including: For each recommended object, determine the click distance corresponding to the recommended object and the popularity corresponding to the recommended object, and determine the recommended object data corresponding to the recommended object based on the click distance and the popularity, and the popularity is determined at least based on the historical click volume of the recommended object.

3. The method according to claim 1, wherein Based on the acquired user data, local map data, and recommended object data, the training samples are determined, including: For each map level, determine the regional popularity of the area corresponding to each recommended object, and for each recommended object, determine the map level data corresponding to the recommended object based on the regional popularity corresponding to the recommended object; The training samples are determined according to the user data, the recommended object data and the map level data corresponding to the recommended objects.

4. The method according to claim 1, wherein Determine the labeling of each training sample based on each user's historical data at different map levels, including: For each user, based on the user's historical data at different map levels, determine the recommended objects that the user has clicked on at different map levels; A training sample including the user data of the user, the recommended object data of the recommended object on which the user has clicked, and the map-level data of the map level when the user clicks the recommended object is marked as a positive sample.

5. A map-based hierarchical recommendation method, characterized in that: include: receiving a recommendation request carrying user data and map-level data, and determining the user data, the map-level data, and each recommended object data according to the recommendation request; For each piece of recommended object data, the recommended object data, the user data, and the map-level data are input into a pre-trained click-through rate prediction model, and the click-through rate of the user on the recommended object data at the map level, as output by the click-through rate prediction model, is determined. The click-through rate prediction model is trained using the click-through rate prediction model training method according to claim 1. Sort the data of each recommended object according to the determined click-through rates; According to the ranking, recommended content is determined.

6. A training device for a click-through rate prediction model, characterized in that: The device comprises: The sample determination module is used to determine each training sample based on the acquired user data, each region-level data, and each recommended object data. Specifically, the sample determination module includes: For each map level, determine the regional popularity of the area to which each recommended object in the map level belongs, and determine the map level data of the map level based on the regional popularity; Combine the user data, the local map data, and the recommended object data to determine the training samples; The label determination module is used to determine the label of each training sample based on the historical data of each user at different map levels; The click-through rate prediction module is used to take each training sample as input and determine the click-through rate corresponding to each training sample through the click-through rate prediction model to be trained; The training module is used to train the click-through rate prediction model based on the click-through rate and annotation of each training sample. The click-through rate prediction model is used to determine the click-through rate of each user on each recommended object at each level of the map.

7. A training device for a click-through rate prediction model, characterized in that: The device comprises: a first determining module, configured to receive a recommendation request carrying user data and map-level data, and determine the user data, the map-level data, and each recommended object data according to the recommendation request; a second determination module for inputting, for each piece of recommended object data, the recommended object data, the user data, and the map-level data into a pre-trained click-through rate prediction model, and determining a click-through rate of the user on the recommended object data at the map level, as output by the click-through rate prediction model, wherein the click-through rate prediction model is trained using the click-through rate prediction model training method according to claim 1; A sorting module is used to sort the recommended object data according to the determined click-through rates; The recommendation module is used to determine the recommended content according to the ranking.

8. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 or claim 5 is implemented.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 4 or claim 5 is implemented.

Images