Intelligent recommendation method and device for advertisement point position delivery category

Abstract

The application discloses an intelligent recommendation method and device for advertisement point position category delivery, and the method comprises the following steps: determining the point position characteristics of a target advertisement delivery device, inputting the point position characteristics of the target advertisement delivery device into a pre-trained category analysis model for analysis, and obtaining the analysis result output by the category analysis model as the result of recommending the advertisement category for delivery on the target advertisement delivery device. It can be seen that, by inputting the point position characteristics of the actual geographical position of the advertisement delivery device into the category analysis model for analysis, the advertisement delivery device suitable for advertisement category delivery and the corresponding delivery probability can be accurately and efficiently analyzed, and then targeted advertisement delivery of the advertisement category is carried out according to the actual situation of the position of the advertisement delivery device, so that the accuracy of advertisement delivery is improved, and the resource utilization rate of the advertisement delivery device is improved.

Description

Technical Field

[0001] This invention relates to the field of advertising placement technology, and in particular to an intelligent recommendation method and apparatus for advertising placement categories. Background Technology

[0002] With the rapid development of outdoor advertising technology, more and more businesses prefer to promote their products and / or services by placing advertisements outdoors in order to increase their visibility.

[0003] Currently, when advertising is needed, the ads are often directly placed through outdoor (such as shopping mall) advertising devices. However, practice has shown that directly placing ads through these devices often results in ads that are not relevant to the current context, leading to low accuracy and failing to achieve the desired promotional effect. Therefore, proposing a technical solution to improve the accuracy of ad placement is particularly important. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent recommendation method and device for advertising placement categories, which can improve the accuracy of advertising placement and thus improve the publicity effect.

[0005] To address the aforementioned technical problems, the first aspect of this invention discloses an intelligent recommendation method for advertising placement categories, the method comprising:

[0006] When there is a target advertising device that needs to be advertised, the location characteristics of the target advertising device are determined. The location characteristics of the target advertising device are determined based on the latitude and longitude of the actual geographical location of the target advertising device.

[0007] The location features of the target advertising placement device are input into a pre-trained category analysis model for analysis, and the analysis results output by the category analysis model are obtained as the advertising category results recommended for placement on the target advertising placement device. The advertising category results include all target advertising categories and the probability of each target advertising category.

[0008] As an optional implementation, in the first aspect of the present invention, the method further includes:

[0009] The characteristics of each of the multiple advertising placement samples are analyzed. The characteristics of each advertising placement sample include the location characteristics of the advertising placement device corresponding to the advertising placement sample, the category of the advertisement placed on the advertising placement device, the industry to which the advertisement placed on the advertising placement device belongs, and the matching parameters between the category of the advertisement placed on the advertising placement device and the advertising placement device.

[0010] Based on the features of each of the advertising samples, a pre-determined basic category analysis model is trained, and the trained basic category analysis model is determined as the pre-trained category analysis model.

[0011] As an optional implementation, in the first aspect of the present invention, when each of the advertising delivery samples includes the location characteristics of the advertising delivery device of the advertising delivery sample, the analysis of the characteristics of each of the multiple advertising delivery samples includes:

[0012] For any of the aforementioned advertising placement samples, obtain the latitude and longitude of the advertising placement device corresponding to the advertising placement sample, and perform a preset level of geographic hash analysis on the latitude and longitude of the advertising placement device to obtain the location identifier of the advertising placement device in the geographic hash coordinates.

[0013] Based on the location identifier corresponding to the advertising device, the network of the advertising device in the geographic hash coordinates is determined, and one or more of the traffic conditions, scene conditions and user conditions of the grid where the advertising device is located are extracted as the location features of the advertising device. The feature dimensions of the location features of each advertising device are equal.

[0014] As an optional implementation, in the first aspect of the present invention, all the advertising samples include positive advertising samples and negative advertising samples;

[0015] In the positive example advertising sample, each of the advertising devices is an advertising device that has already placed an advertisement, and the negative example advertising sample includes negative examples of all the advertising devices in the positive example advertising sample and / or negative examples of the advertising categories placed by all the advertising devices in the positive example advertising sample.

[0016] As an optional implementation, in the first aspect of the invention, when the negative example ad delivery sample includes negative example samples of all the ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner:

[0017] For any of the advertising delivery devices in the positive example advertising delivery sample, based on the first advertising category that has been placed on the advertising delivery device, the second advertising category that has not been placed on the advertising delivery device is determined, and the second advertising category corresponding to all the advertising delivery devices is determined as the negative example advertising delivery sample.

[0018] As an optional implementation, in the first aspect of the invention, when the negative example ad delivery sample includes negative example samples of all ad categories delivered by the ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner:

[0019] For any of the advertising categories that have been advertised in the positive example advertising sample, select advertising devices that have not advertised that advertising category from all the advertising devices in the positive example advertising sample, and determine the advertising devices that have not advertised that advertising category as negative example advertising samples.

[0020] As an optional implementation, in the first aspect of the invention, before training a predetermined basic category analysis model based on the features of each of the advertising samples in all the advertising samples, the method further includes:

[0021] Determine the location features of the sampling advertising delivery devices, and calculate the similarity between the location features of the sampling advertising delivery devices and the location features of each advertising delivery device in each advertising delivery sample;

[0022] Based on the similarity of each of the advertising delivery devices, a predetermined number of target advertising delivery devices are selected from all the advertising delivery devices, sorted by similarity from largest to smallest. Then, the operation of training a predetermined basic category analysis model based on the features of each of the advertising delivery samples in all the advertising delivery samples is performed. The advertising categories corresponding to all the advertising delivery samples include the advertising categories corresponding to all the target advertising delivery devices.

[0023] As an optional implementation, in the first aspect of the invention, training a pre-determined basic category analysis model based on the features of each of the advertising samples in all the advertising samples includes:

[0024] Based on a pre-determined basic category analysis model, the interaction information between the location characteristics, advertising category, industry, and matching parameters of the advertising device in each of the advertising placement samples is learned, and the learning result of the basic category analysis model is obtained. The learning result is the probability value of placing the advertising category at the location of the advertising device.

[0025] Calculate the loss between the probability value of the learned advertising category and the determined true label of the advertising category. If the loss is less than or equal to the preset loss, the basic category analysis model is considered to have been trained successfully.

[0026] If the current loss is greater than the preset loss, the network parameters of the basic category analysis model trained in the current iteration are adjusted based on the gradient descent method. Then, the advertising samples that did not participate in the training are selected from all the advertising samples to participate in the next training operation of the basic category analysis model. This process continues until the current loss is less than or equal to the preset loss. At this point, the training of the basic category analysis model is considered complete, and the next basic category analysis model is the one obtained in the previous training.

[0027] A second aspect of this invention discloses an intelligent recommendation device for advertising placement categories, the device comprising:

[0028] The determination module is used to determine the location characteristics of the target advertising device when there is a target advertising device that needs to be advertised. The location characteristics of the target advertising device are determined based on the latitude and longitude of the actual geographical location of the target advertising device.

[0029] The analysis module is used to input the location characteristics of the target advertising device into a pre-trained category analysis model for analysis;

[0030] The acquisition module is used to acquire the analysis results output by the category analysis model, which are used as the advertising category results for recommending placement on the target advertising device. The advertising category results include all target advertising categories and the probability of each target advertising category.

[0031] As an optional implementation, in a second aspect of the invention, the apparatus further includes:

[0032] The data acquisition module is used to analyze the characteristics of each of the multiple advertising samples. The characteristics of each advertising sample include the location characteristics of the advertising device corresponding to the advertising sample, the category of the advertisement placed on the advertising device, the industry to which the advertisement placed on the advertising device belongs, and the matching parameters between the category of the advertisement placed on the advertising device and the advertising device.

[0033] The training module is used to train a pre-determined basic category analysis model based on the features of each of the advertising samples in all the advertising samples.

[0034] The determining module is further configured to determine the trained basic category analysis model as a pre-trained category analysis model.

[0035] As an optional implementation, in a second aspect of the present invention, when each of the advertising delivery samples includes the location characteristics of the advertising delivery device of that advertising delivery sample, the specific method by which the acquisition module analyzes the characteristics of each of the multiple advertising delivery samples includes:

[0036] For any of the aforementioned advertising placement samples, obtain the latitude and longitude of the advertising placement device corresponding to the advertising placement sample, and perform a preset level of geographic hash analysis on the latitude and longitude of the advertising placement device to obtain the location identifier of the advertising placement device in the geographic hash coordinates.

[0037] Based on the location identifier corresponding to the advertising device, the network of the advertising device in the geographic hash coordinates is determined, and one or more of the traffic conditions, scene conditions and user conditions of the grid where the advertising device is located are extracted as the location features of the advertising device. The feature dimensions of the location features of each advertising device are equal.

[0038] As an optional implementation, in the second aspect of the present invention, all the advertising samples include positive advertising samples and negative advertising samples;

[0039] In the positive example advertising sample, each of the advertising devices is an advertising device that has already placed an advertisement, and the negative example advertising sample includes negative examples of all the advertising devices in the positive example advertising sample and / or negative examples of the advertising categories placed by all the advertising devices in the positive example advertising sample.

[0040] As an optional implementation, in a second aspect of the invention, when the negative example ad delivery sample includes negative example samples of all the ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner:

[0041] For any of the advertising delivery devices in the positive example advertising delivery sample, based on the first advertising category that has been placed on the advertising delivery device, the second advertising category that has not been placed on the advertising delivery device is determined, and the second advertising category corresponding to all the advertising delivery devices is determined as the negative example advertising delivery sample.

[0042] As an optional implementation, in a second aspect of the invention, when the negative example ad delivery sample includes negative example samples of all ad categories delivered by the ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner:

[0043] For any of the advertising categories that have been advertised in the positive example advertising sample, select advertising devices that have not advertised that advertising category from all the advertising devices in the positive example advertising sample, and determine the advertising devices that have not advertised that advertising category as negative example advertising samples.

[0044] As an optional implementation, in a second aspect of the invention, the apparatus further includes:

[0045] The determining module is further configured to determine the location characteristics of the sampling advertising placement device before the training module trains the pre-determined basic category analysis model based on the characteristics of each of the advertising placement samples in all the advertising placement samples;

[0046] The calculation module is used to calculate the similarity between the location features of the sampled advertising delivery device and the location features of each advertising delivery device in each advertising delivery sample;

[0047] The filtering module is used to filter a preset number of target advertising devices from all the advertising devices based on the similarity of each advertising device, sorted from largest to smallest similarity, and to trigger the training module to perform the operation of training a pre-determined basic category analysis model based on the features of each advertising sample in all the advertising samples. The advertising categories corresponding to all the advertising samples include the advertising categories corresponding to all the target advertising devices.

[0048] As an optional implementation, in the second aspect of the invention, the training module trains a pre-determined basic category analysis model based on the features of each of the advertising samples in all the advertising samples, specifically including:

[0049] Based on a pre-determined basic category analysis model, the interaction information between the location characteristics, advertising category, industry, and matching parameters of the advertising device in each of the advertising placement samples is learned, and the learning result of the basic category analysis model is obtained. The learning result is the probability value of placing the advertising category at the location of the advertising device.

[0050] Calculate the loss between the probability value of the learned advertising category and the determined true label of the advertising category. If the loss is less than or equal to the preset loss, the basic category analysis model is considered to have been trained successfully.

[0051] If the current loss is greater than the preset loss, the network parameters of the basic category analysis model trained in the current iteration are adjusted based on the gradient descent method. Then, the advertising samples that did not participate in the training are selected from all the advertising samples to participate in the next training operation of the basic category analysis model. This process continues until the current loss is less than or equal to the preset loss. At this point, the training of the basic category analysis model is considered complete, and the next basic category analysis model is the one obtained in the previous training.

[0052] A third aspect of the present invention discloses another intelligent recommendation device for advertising placement categories, the device comprising:

[0053] Memory containing executable program code;

[0054] A processor coupled to the memory;

[0055] The processor calls the executable program code stored in the memory to execute some or all of the steps in the intelligent recommendation method for advertising placement categories disclosed in the first aspect of the present invention.

[0056] The fourth aspect of the present invention discloses a portable terminal for sorting goods, including a graphic code scanning device and a data processing device, wherein the data processing device is used to perform some or all of the steps in the intelligent recommendation method for advertising placement categories disclosed in the first aspect of the present invention.

[0057] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0058] This invention discloses an intelligent recommendation method and apparatus for advertising placement categories. The method includes: when a target advertising placement device needs to place an advertisement, determining the location characteristics of the target advertising placement device, wherein the location characteristics are determined based on the latitude and longitude of the actual geographical location of the target advertising placement device; inputting the location characteristics of the target advertising placement device into a pre-trained category analysis model for analysis, and obtaining the analysis results output by the category analysis model as the result for recommending advertising categories to be placed on the target advertising placement device, wherein the advertising category result includes all target advertising categories and the probability of each target advertising category. It can be seen that by inputting the location characteristics of the actual geographical location of the advertising placement device into the category analysis model for analysis, this invention can accurately and efficiently analyze the suitable advertising categories and corresponding placement probabilities for the advertising placement device, thereby enabling targeted advertising placement based on the actual location of the advertising placement device, improving the accuracy of advertising placement and the resource utilization rate of the advertising placement device. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart illustrating an intelligent recommendation method for advertising placement categories disclosed in an embodiment of the present invention.

[0061] Figure 2 This is a schematic diagram of the structure of an intelligent recommendation device for advertising placement categories disclosed in an embodiment of the present invention.

[0062] Figure 3 This is a schematic diagram of the structure of another intelligent recommendation device for advertising placement categories disclosed in an embodiment of the present invention.

[0063] Figure 4 This is a schematic diagram of the structure of another intelligent recommendation device for advertising placement categories disclosed in an embodiment of the present invention. Detailed Implementation

[0064] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] The terms "second," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0066] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0067] This invention discloses an intelligent recommendation method and device for advertising placement categories. By inputting the location characteristics of the actual geographical location of the advertising placement device into a category analysis model, it can accurately and efficiently analyze the suitable advertising categories and corresponding placement probabilities for the advertising placement device. Furthermore, it can perform targeted advertising placement based on the actual location of the advertising placement device, improving the accuracy of advertising placement and the resource utilization rate of the advertising placement device. Detailed descriptions follow.

[0068] Example 1

[0069] Please see Figure 1 , Figure 1 This is a flowchart illustrating an intelligent recommendation method for advertising placement categories disclosed in an embodiment of the present invention. Figure 1 The described intelligent recommendation method for ad placement categories is applied to scenarios where advertising is required and advertising placement equipment is installed. These scenarios include one or more of the following: shopping malls, schools, factories, and commercial areas. Figure 1 As shown, the intelligent recommendation method for the product category of this ad placement can include the following operations:

[0070] 101. When there are target advertising placement devices that need to place advertisements, determine the location characteristics of the target advertising placement devices.

[0071] In this embodiment of the invention, optionally, the location characteristics of the target advertising delivery device are determined based on the latitude and longitude of the actual geographical location of the target advertising delivery device.

[0072] 102. Input the location characteristics of the target advertising devices into the pre-trained category analysis model for analysis.

[0073] 103. Obtain the analysis results output by the category analysis model and use them as the result of recommending advertising categories for the target advertising devices.

[0074] In this embodiment of the invention, optionally, the advertising category result includes all target advertising categories and the probability of each target advertising category. Optionally, encapsulating this technical solution using an HTTP service and employing asynchronous and multi-threaded processing methods can improve processing efficiency.

[0075] As can be seen, by inputting the location characteristics of the actual geographical location of the advertising placement device into the category analysis model for analysis, the implementation of the embodiments of the present invention can accurately and efficiently analyze the advertising categories suitable for the advertising placement device and the corresponding placement probabilities. Then, based on the actual location of the advertising placement device, targeted advertising categories can be placed, thereby improving the accuracy of advertising placement and the resource utilization rate of the advertising placement device.

[0076] In an optional embodiment, the method may further include the following steps:

[0077] The characteristics of each advertising sample in multiple advertising samples are analyzed. The characteristics of each advertising sample include the location characteristics of the advertising device corresponding to the advertising sample, the advertising category of the advertisement placed on the advertising device, the industry to which the advertisement placed on the advertising device belongs, and the matching parameters between the advertising category of the advertisement placed on the advertising device and the advertising device.

[0078] Based on the characteristics of each advertising sample in all advertising samples, a pre-determined basic category analysis model is trained, and the trained basic category analysis model is identified as the pre-trained category analysis model.

[0079] In this optional embodiment, the features of each advertising sample may further include one or more of the following: the identity identifier of the advertising device corresponding to the advertising sample, the advertising duration of each advertising category, and the advertising effect of each advertising category. The more content included in the features of the advertising sample, the better it is for improving the training accuracy and reliability of the category analysis model. Further optionally, corresponding training parameters are set for each feature of the corresponding advertising sample based on the advertising duration and advertising effect of each advertising category, and a correlation is established between the training parameters and the corresponding features. Specifically, a pre-determined basic category analysis model is trained based on the features of each advertising sample in all advertising samples, the corresponding training parameters for that feature, and the corresponding correlation. By setting training parameters for each feature of the corresponding advertising sample based on the advertising duration and advertising effect, and by having the features and the correlation between the features and their training parameters participate in the training of the category analysis model, the training accuracy and reliability of the category analysis model can be further improved.

[0080] In this optional embodiment, the data collection involved in the features may be provided by the advertising placement device provider, or the data collection device may travel to the geographical location of the advertising placement device to collect the data, or a person may travel to the geographical location of the advertising placement device to collect the data.

[0081] In this optional embodiment, the basic category analysis model can be any model capable of performing category analysis, such as the NeuralCF model.

[0082] In this optional embodiment, the higher the value (0-1) of the matching parameter between the industry and product category of the advertisement placed on the advertising placement device and the advertising placement device, the higher the matching degree between the two, that is, the more suitable the placed advertisement is for the corresponding advertising placement device. Alternatively, 0 and 1 can be used directly, where 0 indicates unsuitable placement and 1 indicates suitable placement.

[0083] In this optional embodiment, optionally, based on the characteristics of each advertising sample in all advertising samples, a pre-determined basic category analysis model is trained, including:

[0084] Based on a pre-determined basic category analysis model, the interaction information between the location characteristics of the advertising device, the advertising category, the industry, and the matching parameters in each advertising sample is learned in all advertising samples. The learning result of the basic category analysis model is obtained, which is the probability value of advertising that category at the location of the advertising device.

[0085] Calculate the loss between the probability value of the learned advertising category and the true label of the determined advertising category. If the loss is less than or equal to the preset loss, the basic category analysis model training is complete.

[0086] If the current loss is greater than the preset loss, the network parameters of the basic category analysis model trained in this iteration are adjusted based on the gradient descent method. Then, advertising samples that did not participate in the training are selected from all advertising samples to participate in the next training operation of the basic category analysis model. This process continues until the current loss is less than or equal to the preset loss. At this point, the training of the basic category analysis model is considered complete, and the next basic category analysis model will be the basic category analysis model obtained in the previous training.

[0087] In this optional embodiment, the network parameters may include, but are not limited to, word vector parameters and hidden unit parameters in the neural network.

[0088] In this optional embodiment, after the category analysis model is trained, an advertising placement prediction sample is obtained, and the trained category analysis model is used to make a prediction to obtain the placement probability of the advertising category corresponding to the advertising placement prediction sample.

[0089] As can be seen, this optional embodiment trains the category analysis model by considering the location characteristics of the advertising devices that have already placed ads, the category of the ads, the industry to which they belong, and the matching situation between the ad category and the advertising devices. This improves the training accuracy of the category analysis model, resulting in a more accurate advertising category analysis model. This is beneficial for subsequent direct use in recommending advertising categories, thus improving the accuracy of advertising category recommendations. Furthermore, by learning the interaction information between the location characteristics of the advertising devices, the ad category, the industry, and the matching parameters through the basic category analysis model, the probability of the advertising devices for the placed ads is obtained, and the loss between the probability and the actual label is calculated. This determines whether training is complete or requires adjustment and continued training until the corresponding loss is small, thus further improving the training accuracy of the category analysis model.

[0090] As an optional implementation, when each ad delivery sample includes the location characteristics of the ad delivery device for that ad delivery sample, the characteristics of each ad delivery sample among multiple ad delivery samples are analyzed, including:

[0091] For any given ad placement sample, obtain the latitude and longitude of the ad placement device corresponding to the ad placement sample, and perform a geographic hash analysis of the latitude and longitude of the ad placement device at a preset level (such as level 7 or level 8) to obtain the location identifier of the ad placement device in the corresponding grid in the geographic hash coordinates.

[0092] Extract one or more of the following from the grid where the advertising device is located: traffic conditions, scene conditions, and user conditions, as the location features of the advertising device. The feature dimensions of the location features of each advertising device are equal.

[0093] In this optional embodiment, the latitude and longitude of the advertising delivery device are the latitude and longitude of the actual geographical location of the advertising delivery device.

[0094] In this optional embodiment, a geospatial analysis of the latitude and longitude of the advertising device is performed at a preset level (such as level 7 or 8) to obtain the location identifier of the advertising device in the corresponding grid in the geospatial coordinates. Specifically, the open-source pygeohash library is used. The latitude and longitude of the actual geographical location of the advertising device are input. pygeohash performs geospatial encoding on the latitude and longitude and outputs a string of a preset level as the location identifier of the grid. For example, after the latitude and longitude (121.604595, 31.179705) are encoded by pygeohash, the location identifier "wtw3rhj" is obtained, indicating that the actual geographical location of the advertising device falls within the grid.

[0095] In this optional embodiment, the traffic situation of each grid (e.g., 150m*150m) includes, but is not limited to, bus stop locations, subway station locations, train station locations, the number of bus stops, subway stations, and train stations, and the proportion of each mode of transportation. The proportion of each mode of transportation can be understood as the percentage of people choosing that mode of transportation within the same time period, or as the percentage of the number of people using that mode of transportation within the area corresponding to the grid. The scene situation of each grid includes, but is not limited to, all scene types and the number of each scene type. All scene types include, but are not limited to, one or more of the following: school, subway, shopping mall, factory, residential area, etc. The user situation of each grid includes, but is not limited to, one or more of the following: gender proportion, age distribution proportion (e.g., the number and proportion of residents under 19 years old, 19-25 years old, 26-35 years old, 36-45 years old, and over 45 years old), income proportion (e.g., the number and proportion of residents with high, medium, and low incomes), and occupation proportion. The more detailed and comprehensive the extracted location features, the more accurate and reliable the category analysis model will be, and consequently, the more precise the analysis of the suitable advertising categories for actual advertising placement devices.

[0096] In this optional embodiment, after the point features of each grid are extracted, they are entered into a feature library, such as a grid feature library.

[0097] In this optional embodiment, all advertising samples may include positive advertising samples and negative advertising samples; wherein, each advertising device in the positive advertising samples is an advertising device that has already placed an advertisement, and the negative advertising samples include negative samples of all advertising devices in the positive advertising samples and / or negative samples of the advertising categories placed by all advertising devices in the positive advertising samples.

[0098] In this optional embodiment, when the negative example ad serving sample includes negative example samples from all ad serving devices of the positive example ad serving sample, the negative example ad serving sample is constructed in the following manner:

[0099] For any advertising device in the positive example advertising sample, based on the first advertising category already displayed on that device, determine the second advertising category not displayed on that device, and determine the corresponding second advertising categories for all advertising devices, which will be used as negative example advertising samples. Specifically, for any given advertising device, the second advertising category not displayed on that device can be any advertising category other than those already displayed on that device, or it can be any advertising category in the advertising category database other than those already displayed on that device.

[0100] In this optional embodiment, when the negative example ad delivery sample includes negative examples of ad categories delivered by all ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner:

[0101] For any of the aforementioned advertising categories that have been placed in the positive example advertising placement sample, select advertising placement devices that have not placed advertising placement devices in the positive example advertising placement sample, and determine the advertising placement devices that have not placed advertising placement devices in the positive example advertising placement sample as negative example advertising placement samples.

[0102] As can be seen, this optional embodiment improves the accuracy and efficiency of point feature extraction by performing geographic hash analysis on the latitude and longitude of the actual geographical location of the advertising device to obtain the corresponding location identifier and then extracting the point features of the grid. Furthermore, by obtaining the advertising categories already placed on the advertising devices as positive advertising samples, and then further determining the corresponding negative advertising samples based on the obtained advertising devices and advertising categories of the placed ads, the comprehensiveness and accuracy of the advertising sample determination can be improved. This is beneficial to improving the training accuracy and reliability of the category analysis model, making the category analysis model more adaptable. Finally, by determining the same dimension of point features for each advertising device, the comparability of each grid, i.e., each point feature of the advertising device, can be guaranteed during training. This facilitates the category analysis model's analysis and learning of point features, improving the learning efficiency and success rate of the category analysis model, thereby increasing the probability of successfully training an accurate category analysis model.

[0103] In another alternative embodiment, before training a pre-determined basic category analysis model based on the features of each ad delivery sample across all ad delivery samples, the method may further include the following steps:

[0104] Determine the location characteristics of the sampling advertising devices, and calculate the similarity between the location characteristics of the sampling advertising devices and the location characteristics of each advertising device in each advertising sample;

[0105] Based on the similarity of each advertising device, select a preset number (e.g., 5 or 8) of target advertising devices from all advertising devices, sorted by similarity from largest to smallest. Then, perform the above-mentioned operation of training a pre-determined basic category analysis model based on the features of each advertising sample in all advertising samples. The advertising categories corresponding to all advertising samples include the advertising categories corresponding to all target advertising devices.

[0106] In this optional embodiment, the sampling advertising delivery device can be the aforementioned target advertising delivery device, or it can be a device other than the aforementioned advertising delivery device.

[0107] As can be seen, this optional embodiment performs category recall analysis by sampling the location characteristics of advertising devices before training the category analysis model. This reduces the number of advertising categories and the amount of data, thereby enabling the training of the category analysis model. While ensuring accurate training, it can also improve training efficiency, which is conducive to improving the application efficiency of the category analysis model.

[0108] Example 2

[0109] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of an intelligent recommendation device for advertising placement categories disclosed in an embodiment of the present invention. Wherein, Figure 2 The described intelligent recommendation device for advertising placement categories is applied in scenarios where advertising is required and advertising placement equipment is installed. These scenarios include one or more of the following: shopping malls, schools, factories, commercial areas, etc. The device may include a data processing chip, a processing terminal, or a processing server (wherein the processing server can be a local server or a cloud server). Figure 2 As shown, the intelligent recommendation device for the product category of this advertising location may include:

[0110] The determination module 201 is used to determine the location characteristics of the target advertising delivery device when there is a target advertising delivery device that needs to deliver advertising.

[0111] Analysis module 202 is used to input the location characteristics of the target advertising placement device into a pre-trained category analysis model for analysis;

[0112] The acquisition module 203 is used to acquire the analysis results output by the category analysis model, which are used as the advertising category results for recommending the advertising categories to be placed on the target advertising devices. The advertising category results include all target advertising categories and the probability of each target advertising category.

[0113] It is evident that implementation Figure 2 The described embodiments input the location characteristics of the actual geographical location of the advertising placement device into the category analysis model for analysis. This enables accurate and efficient analysis of the advertising categories suitable for the advertising placement device and the corresponding placement probability. Furthermore, based on the actual location of the advertising placement device, targeted advertising categories are placed, thereby improving the accuracy of advertising placement and the resource utilization rate of the advertising placement device.

[0114] In another alternative embodiment, such as Figure 4 As shown, the device also includes:

[0115] The data acquisition module 204 is used to analyze the characteristics of each advertising sample in multiple advertising samples. The characteristics of each advertising sample include the location characteristics of the advertising device corresponding to the advertising sample, the advertising category of the advertisement placed on the advertising device, the industry to which the advertisement placed on the advertising device belongs, and the matching parameters between the advertising category placed on the advertising device and the advertising device.

[0116] Training module 205 is used to train a pre-determined basic category analysis model based on the features of each advertising sample in all advertising samples;

[0117] The determination module 201 is also used to determine the trained basic category analysis model as the pre-trained category analysis model.

[0118] In this optional embodiment, the training module 205 may train a pre-determined basic category analysis model based on the features of each advertising sample in all advertising samples in the following specific ways:

[0119] Based on a pre-determined basic category analysis model, the interaction information between the location characteristics of the advertising device, the advertising category, the industry, and the matching parameters in each advertising sample is learned in all advertising samples. The learning result of the basic category analysis model is obtained, which is the probability value of advertising that category at the location of the advertising device.

[0120] Calculate the loss between the probability value of the learned advertising category and the true label of the determined advertising category. If the loss is less than or equal to the preset loss, the basic category analysis model training is complete.

[0121] If the current loss is greater than the preset loss, the network parameters of the basic category analysis model trained in this iteration are adjusted based on the gradient descent method. Then, advertising samples that did not participate in the training are selected from all advertising samples to participate in the next training operation of the basic category analysis model. This process continues until the current loss is less than or equal to the preset loss. At this point, the training of the basic category analysis model is considered complete, and the next basic category analysis model will be the basic category analysis model obtained in the previous training.

[0122] In this optional embodiment, the network parameters may include, but are not limited to, word vector parameters and hidden unit parameters in the neural network.

[0123] In this optional embodiment, after the category analysis model is trained, an advertising placement prediction sample is obtained, and the trained category analysis model is used to make a prediction to obtain the placement probability of the advertising category corresponding to the advertising placement prediction sample.

[0124] As can be seen, this optional embodiment trains the category analysis model by considering the location characteristics of the advertising devices that have already placed ads, the category of the ads, the industry to which they belong, and the matching situation between the ad category and the advertising devices. This improves the training accuracy of the category analysis model, resulting in a more accurate advertising category analysis model. This is beneficial for subsequent direct use in recommending advertising categories, thus improving the accuracy of advertising category recommendations. Furthermore, by learning the interaction information between the location characteristics of the advertising devices, the ad category, the industry, and the matching parameters through the basic category analysis model, the probability of the advertising devices for the placed ads is obtained, and the loss between the probability and the actual label is calculated. This determines whether training is complete or requires adjustment and continued training until the corresponding loss is small, thus further improving the training accuracy of the category analysis model.

[0125] In this optional embodiment, when each advertising delivery sample includes the location characteristics of the advertising delivery device for that advertising delivery sample, the specific method by which the acquisition module 204 analyzes the characteristics of each advertising delivery sample among multiple advertising delivery samples includes:

[0126] For any given ad placement sample, obtain the latitude and longitude of the ad placement device corresponding to the ad placement sample, and perform a geographic hash analysis of the latitude and longitude of the ad placement device at a preset level (such as level 7 or level 8) to obtain the location identifier of the ad placement device in the corresponding grid in the geographic hash coordinates.

[0127] Extract one or more of the following from the grid where the advertising device is located: traffic conditions, scene conditions, and user conditions, as the location features of the advertising device. The feature dimensions of the location features of each advertising device are equal.

[0128] In this optional embodiment, the latitude and longitude of the advertising delivery device are the latitude and longitude of the actual geographical location of the advertising delivery device.

[0129] In this optional embodiment, a geospatial analysis of the latitude and longitude of the advertising device is performed at a preset level (such as level 7 or 8) to obtain the location identifier of the advertising device in the corresponding grid in the geospatial coordinates. Specifically, the open-source pygeohash library is used. The latitude and longitude of the actual geographical location of the advertising device are input. pygeohash performs geospatial encoding on the latitude and longitude and outputs a string of a preset level as the location identifier of the grid. For example, after the latitude and longitude (121.604595, 31.179705) are encoded by pygeohash, the location identifier "wtw3rhj" is obtained, indicating that the actual geographical location of the advertising device falls within the grid.

[0130] In this optional embodiment, the traffic situation of each grid (e.g., 150m*150m) includes, but is not limited to, bus stop locations, subway station locations, train station locations, the number of bus stops, subway stations, and train stations, and the proportion of each mode of transportation. The proportion of each mode of transportation can be understood as the percentage of people choosing that mode of transportation within the same time period, or as the percentage of the number of people using that mode of transportation within the area corresponding to the grid. The scene situation of each grid includes, but is not limited to, all scene types and the number of each scene type. All scene types include, but are not limited to, one or more of the following: school, subway, shopping mall, factory, residential area, etc. The user situation of each grid includes, but is not limited to, one or more of the following: gender proportion, age distribution proportion (e.g., the number and proportion of residents under 19 years old, 19-25 years old, 26-35 years old, 36-45 years old, and over 45 years old), income proportion (e.g., the number and proportion of residents with high, medium, and low incomes), and occupation proportion. The more detailed and comprehensive the extracted location features, the more accurate and reliable the category analysis model will be, and consequently, the more precise the analysis of the suitable advertising categories for actual advertising placement devices.

[0131] In this optional embodiment, after the point features of each grid are extracted, they are entered into a feature library, such as a grid feature library.

[0132] In this optional embodiment, all advertising delivery samples include positive advertising delivery samples and negative advertising delivery samples; wherein, each advertising delivery device in the positive advertising delivery samples is an advertising delivery device that has already delivered an advertisement, and the negative advertising delivery samples include negative samples of all advertising delivery devices in the positive advertising delivery samples and / or negative samples of the advertising categories delivered by all advertising delivery devices in the positive advertising delivery samples.

[0133] In this optional embodiment, when the negative example ad serving sample includes negative example samples from all ad serving devices of the positive example ad serving sample, the negative example ad serving sample is constructed in the following manner:

[0134] For any advertising device in the positive example advertising sample, based on the first advertising category already displayed on that device, determine the second advertising category not displayed on that device, and determine the corresponding second advertising categories for all advertising devices, which will be used as negative example advertising samples. Specifically, for any given advertising device, the second advertising category not displayed on that device can be any advertising category other than those already displayed on that device, or it can be any advertising category in the advertising category database other than those already displayed on that device.

[0135] In this optional embodiment, when the negative example ad delivery sample includes negative examples of ad categories delivered by all ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner:

[0136] For any of the aforementioned advertising categories that have been placed in the positive example advertising placement sample, select advertising placement devices that have not placed advertising placement devices in the positive example advertising placement sample, and determine the advertising placement devices that have not placed advertising placement devices in the positive example advertising placement sample as negative example advertising placement samples.

[0137] It is evident that implementation Figure 3 The described embodiments can improve the accuracy and efficiency of point feature extraction by performing geospatial analysis on the latitude and longitude of the actual geographical location of the advertising device to obtain the corresponding location identifier, and then extracting the point features of the grid. Furthermore, by obtaining the advertising categories already placed on the advertising device as positive advertising samples, and then further determining the corresponding negative advertising samples based on the obtained advertising devices and advertising categories, the comprehensiveness and accuracy of the advertising sample determination can be improved. This is beneficial to improving the training accuracy and reliability of the category analysis model, making the trained category analysis model more adaptable. Finally, by determining the same dimension of point features for each advertising device, the comparability of each grid, i.e., each point feature of the advertising device, can be guaranteed during training. This facilitates the category analysis model's analysis and learning of point features, improving the learning efficiency and success rate of the category analysis model, thereby increasing the probability of successfully training an accurate category analysis model.

[0138] In yet another alternative embodiment, such as Figure 4 As shown, the determination module 201 is also used to determine the location characteristics of the sampling advertising placement device before the training module 205 trains the pre-determined basic category analysis model based on the characteristics of each advertising placement sample in all advertising placement samples.

[0139] Calculation module 206 is used to calculate the similarity between the location features of the sampled advertising delivery device and the location features of each advertising delivery device in each advertising delivery sample;

[0140] The filtering module 207 is used to filter all target advertising devices from all advertising devices according to the similarity of each advertising device, sorting them from largest to smallest by a preset number (such as 5 or 8), and trigger the training module 205 to perform the above-mentioned operation of training a pre-determined basic category analysis model based on the features of each advertising sample in all advertising samples. The advertising categories corresponding to all advertising samples include the advertising categories corresponding to all target advertising devices.

[0141] It is evident that implementation Figure 3 The described embodiments enable category recall analysis by sampling the location characteristics of advertising devices before training the category analysis model. This reduces the number of advertising categories and the amount of data, thus facilitating the training of the category analysis model. While ensuring accurate training, this also improves training efficiency, thereby enhancing the application efficiency of the category analysis model.

[0142] Example 3

[0143] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of another intelligent recommendation device for advertising placement categories disclosed in an embodiment of the present invention. The device is applied in scenarios where advertising placement is required and advertising placement equipment is installed. These scenarios include one or more of the following: shopping malls, schools, factories, commercial areas, etc. The device may include a data processing chip, a processing terminal, or a processing server (wherein the processing server can be a local server or a cloud server). Figure 4 As shown, the device may include:

[0144] Memory 301 storing executable program code;

[0145] Processor 302 coupled to memory 301;

[0146] The processor 302 calls the executable program code stored in the memory 301 to execute the steps in the intelligent recommendation method for advertising placement categories disclosed in Embodiment 1 of the present invention.

[0147] Example 4

[0148] This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to implement the steps in the intelligent recommendation method for advertising placement categories disclosed in Embodiment 1 of this invention.

[0149] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0150] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0151] Finally, it should be noted that the intelligent recommendation method and apparatus for advertising placement categories disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0152] Finally, it should be noted that the intelligent recommendation method and apparatus for advertising placement categories disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent recommendation of product categories for advertising placement, characterized in that, The method includes: When there is a target advertising device that needs to be advertised, the location characteristics of the target advertising device are determined. The location characteristics of the target advertising device are determined based on the latitude and longitude of the actual geographical location of the target advertising device. The location features of the target advertising placement device are input into a pre-trained category analysis model for analysis, and the analysis results output by the category analysis model are obtained as the advertising category results for recommending placement on the target advertising placement device. The advertising category results include all target advertising categories and the probability of each target advertising category. The method further includes: The characteristics of each of the multiple advertising placement samples are analyzed. The characteristics of each advertising placement sample include the location characteristics of the advertising placement device corresponding to the advertising placement sample, the category of the advertisement placed on the advertising placement device, the industry to which the advertisement placed on the advertising placement device belongs, and the matching parameters between the category of the advertisement placed on the advertising placement device and the advertising placement device. Based on the features of each of the advertising samples in all the advertising samples, a pre-determined basic category analysis model is trained, and the trained basic category analysis model is determined as the pre-trained category analysis model. Wherein, all the advertising placement samples include positive advertising placement samples and negative advertising placement samples; wherein, each of the advertising placement devices in the positive advertising placement samples is an advertising placement device that has already placed an advertisement, and the negative advertising placement samples include negative samples of all the advertising placement devices in the positive advertising placement samples and / or negative samples of the advertising categories placed by all the advertising placement devices in the positive advertising placement samples; Wherein, when the negative example ad delivery sample includes negative example samples of all the ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner: For any of the advertising devices in the positive example advertising sample, based on the first advertising category that has been advertised on the advertising device, determine the second advertising category that has not been advertised on the advertising device, and determine the second advertising category corresponding to all the advertising devices as negative example advertising samples; When the negative example ad delivery sample includes negative example samples of all ad categories delivered by the ad delivery devices of the positive example ad delivery sample, the negative example ad delivery sample is constructed in the following manner: For any of the advertising categories that have been advertised in the positive example advertising sample, select advertising devices that have not advertised that advertising category from all the advertising devices in the positive example advertising sample, and determine the advertising devices that have not advertised that advertising category as negative example advertising samples.

2. The intelligent recommendation method for advertising placement categories according to claim 1, characterized in that, Each of the aforementioned advertising delivery samples includes the location characteristics of the advertising delivery device for that advertising delivery sample; The analysis of the characteristics of each of the multiple advertising samples includes: For any of the aforementioned advertising placement samples, obtain the latitude and longitude of the advertising placement device corresponding to the advertising placement sample, and perform a preset level of geographic hash analysis on the latitude and longitude of the advertising placement device to obtain the location identifier of the advertising placement device in the geographic hash coordinates. Based on the location identifier corresponding to the advertising device, the grid of the advertising device in the geographic hash coordinates is determined, and one or more of the traffic conditions, scene conditions and user conditions of the grid where the advertising device is located are extracted as the location features of the advertising device. The feature dimensions of the location features of each advertising device are equal.

3. The intelligent recommendation method for advertising placement categories according to claim 1 or 2, characterized in that, Before training the pre-determined basic category analysis model based on the features of each of the advertising samples in all the advertising samples, the method further includes: Determine the location features of the sampling advertising delivery devices, and calculate the similarity between the location features of the sampling advertising delivery devices and the location features of each advertising delivery device in each advertising delivery sample; Based on the similarity of each of the advertising delivery devices, a predetermined number of target advertising delivery devices are selected from all the advertising delivery devices, sorted by similarity from largest to smallest. Then, the operation of training a predetermined basic category analysis model based on the features of each of the advertising delivery samples in all the advertising delivery samples is performed. The advertising categories corresponding to all the advertising delivery samples include the advertising categories corresponding to all the target advertising delivery devices.

4. The intelligent recommendation method for advertising placement categories according to claim 1 or 2, characterized in that, The step of training a pre-determined basic category analysis model based on the features of each of the advertising samples in all the advertising samples includes: Based on a pre-determined basic category analysis model, the interaction information between the location characteristics, advertising category, industry, and matching parameters of the advertising device in each of the advertising placement samples is learned, and the learning result of the basic category analysis model is obtained. The learning result is the probability value of placing the advertising category at the location of the advertising device. Calculate the loss between the probability value of the learned advertising category and the determined true label of the advertising category. If the loss is less than or equal to the preset loss, the basic category analysis model is considered to have been trained successfully. If the current loss is greater than the preset loss, the network parameters of the basic category analysis model trained in the current iteration are adjusted based on the gradient descent method. Then, the advertising samples that did not participate in the training are selected from all the advertising samples to participate in the next training operation of the basic category analysis model. This process continues until the current loss is less than or equal to the preset loss. At this point, the training of the basic category analysis model is considered complete, and the next basic category analysis model is the one obtained in the previous training.

5. An intelligent recommendation device for advertising placement categories, characterized in that, The apparatus is used to implement the intelligent recommendation method for advertising placement categories as described in any one of claims 1-4, the apparatus comprising: The determination module is used to determine the location characteristics of the target advertising device when there is a target advertising device that needs to be advertised. The location characteristics of the target advertising device are determined based on the latitude and longitude of the actual geographical location of the target advertising device. The analysis module is used to input the location characteristics of the target advertising device into a pre-trained category analysis model for analysis; The acquisition module is used to acquire the analysis results output by the category analysis model, which are used as the advertising category results for recommending placement on the target advertising device. The advertising category results include all target advertising categories and the probability of each target advertising category.

6. An intelligent recommendation device for advertising placement categories, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the intelligent recommendation method for advertising placement categories as described in any one of claims 1-4.

Images