Traffic estimation method and device, equipment and storage medium

Abstract

According to the embodiment of the invention, the invention provides a method and device for traffic estimation, equipment and a storage medium. The method comprises the following steps: in response to a received advertisement putting request of a user for a target object, determining an interest point location corresponding to the advertisement putting request; determining feature data corresponding to the point of interest, wherein the feature data comprises at least one of the following items: historical delivery data corresponding to the point of interest, user behavior data and warehouse logistics data associated with the target object; and determining a traffic estimation result of the point of interest corresponding to the advertisement putting request based on the feature data by using a trained traffic estimation model, the initialization weight of the traffic estimation model being determined by using a meta-learning model. Therefore, by introducing the warehouse logistics data associated with the target object and the meta-learning method, factors influencing the advertisement traffic can be considered more comprehensively, so that the accuracy of traffic estimation is improved.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of information technology, and in particular, to a method, an apparatus, a device, and a computer-readable storage medium for traffic estimation. Background Art

[0002] In the marketing field, advertising is an important means to increase brand exposure and market influence. Logistics advertising usually relies on material carriers such as express packaging boxes (such as stickers, box advertising and direct mail (DM) sheets) to achieve widespread advertising and support corporate marketing activities.

[0003] In the related art, the traffic estimation of advertisements mainly relies on the experience of the delivery personnel to make decisions. However, this empirical estimation method has obvious limitations: on the one hand, due to the reliance on human judgment, the accuracy of traffic estimation is low, and there will often be insufficient or excessive delivery, resulting in limited coverage of advertising or increased costs. On the other hand, the related art lacks comprehensive consideration of multiple factors related to the target objects that affect the traffic of advertisements, which further limits the accuracy of traffic estimation.

[0004] Therefore, a solution for traffic estimation is needed to improve the accuracy of advertising traffic estimation and delivery efficiency, thereby helping users optimize advertising delivery strategies and maximize return on investment. Summary of the invention

[0005] In a first aspect of the present disclosure, a method for traffic estimation is provided. The method includes: in response to receiving a user's advertisement delivery request for a target object, determining a point of interest corresponding to the advertisement delivery request; determining feature data corresponding to the point of interest, the feature data including at least one of the following: historical delivery data corresponding to the point of interest, user behavior data, and warehouse logistics data associated with the target object; using a trained traffic estimation model, based on the feature data, determining a traffic estimation result for the point of interest corresponding to the advertisement delivery request, wherein the initialization weight of the traffic estimation model is determined using a meta-learning model.

[0006] In the second aspect of the present disclosure, a device for traffic estimation is provided. The device includes: a point of interest determination module, configured to determine the point of interest corresponding to the advertisement delivery request in response to receiving an advertisement delivery request from a user for a target object; a feature data determination module, configured to determine feature data corresponding to the point of interest, the feature data including at least one of the following: historical delivery data corresponding to the point of interest, user behavior data, and warehouse logistics data associated with the target object; and a traffic estimation module, configured to use a trained traffic estimation model to determine the traffic estimation result for the point of interest corresponding to the advertisement delivery request based on the feature data, and the initialization weight of the traffic estimation model is determined using a meta-learning model.

[0007] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory, the at least one memory is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit. When the instructions are executed by the at least one processing unit, the device executes the method of the first aspect.

[0008] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and the computer program can be executed by a processor to implement the method of the first aspect.

[0009] In a fifth aspect of the present disclosure, a computer program product is provided, which is tangibly stored in a computer storage medium and includes computer executable instructions, which when executed by a device cause the device to perform the method according to the first aspect of the present disclosure.

[0010] It should be understood that the content described in this content section is not intended to limit the key features or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0012] Figure 1 A schematic diagram showing an example environment in which embodiments of the present disclosure can be implemented;

[0013] Figure 2 A flow chart showing a flow estimation process according to some embodiments of the present disclosure is shown;

[0014] Figure 3AA schematic diagram showing a process of determining feature data in some embodiments of the present disclosure;

[0015] Figure 3B A schematic diagram showing a process of determining a flow estimation result in some embodiments of the present disclosure;

[0016] Figure 3C A schematic diagram showing a process of training a traffic prediction model in some embodiments of the present disclosure is shown;

[0017] Figure 3D A schematic diagram showing a process of training a meta-learning model according to some embodiments of the present disclosure;

[0018] Figure 4 A block diagram showing an apparatus for traffic estimation according to some embodiments of the present disclosure; and

[0019] Figure 5 A block diagram of an electronic device capable of implementing one or more embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0020] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0021] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0022] As an optional but non-limiting implementation, in response to receiving an active request from the user, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0023] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet the relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0024] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and relevant provisions.

[0025] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0026] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may also be included below.

[0027] As briefly mentioned above, in the logistics industry, advertising is a common marketing method, which usually relies on the material carriers of logistics packages (for example, stickers, box advertisements, and DM leaflets) to carry out high-exposure advertising. Users hope to maximize the exposure of advertisements through these material carriers, thereby improving marketing effectiveness. The traffic of these advertisements (which can also be understood as exposure, the number of times the advertisement is viewed and viewed) is an important indicator of the effectiveness of advertising. The return on investment (ROI) can usually be calculated based on the estimated results of advertising traffic, thereby determining the efficiency of advertising and measuring the amount of advertising.

[0028] In the related technology, current advertising traffic estimation mainly relies on the estimation of the delivery personnel based on experience. For example, delivery personnel will decide how many physical carriers such as DMs to deliver based on the estimation of the user's recent ordering behavior. However, this method is not accurate and can easily lead to deviations in delivery decisions. For example, the estimated traffic is smaller than the actual traffic: the scale of advertising delivery is too small, which may result in failure to cover all target groups and insufficient advertising effect. The estimated traffic is larger than the actual traffic: it will cause waste of physical media (such as DMs) and increase the cost of advertising delivery.

[0029] In addition, due to the particularity of advertisements for target objects, the estimation of advertisement traffic is also based on the estimation of the situation of the target objects, such as changes in warehouse capacity, the impact of promotional activities, etc. Since material packages are sent directly from the warehouse, warehouse logistics data (also known as warehouse capacity) (such as inventory and shipment volume) has an important impact on advertisement traffic. If the estimation is based solely on experience, it cannot accurately reflect the impact of warehouse logistics data on traffic, resulting in inaccurate estimation results.

[0030] Therefore, in order to solve the above problems, we can introduce warehouse logistics data and meta-learning methods, comprehensively consider multiple factors affecting advertising traffic, and build a machine learning model for advertising traffic estimation, thereby improving the accuracy of traffic estimation.

[0031] In an embodiment of the present disclosure, a solution for traffic estimation is provided. According to an embodiment of the present disclosure, in response to receiving a user's advertising delivery request for a target object, a point of interest corresponding to the advertising delivery request is determined; feature data corresponding to the point of interest is determined, and the feature data includes at least one of the following: historical delivery data corresponding to the point of interest, user behavior data, and warehouse logistics data associated with the target object; using a trained traffic estimation model, based on the feature data, a traffic estimation result for the point of interest corresponding to the advertising delivery request is determined, and the initialization weight of the traffic estimation model is determined using a meta-learning model.

[0032] Therefore, by combining the historical delivery data corresponding to the points of interest, user behavior data, and warehouse logistics data associated with the target object, the machine learning model is used to estimate traffic, which greatly improves the accuracy of advertising traffic estimation, thereby effectively preventing insufficient or excessive advertising delivery and avoiding the problem of resource waste caused by inaccurate traffic estimation. In addition, the meta-learning model is used to provide good initialization weights for the traffic estimation model, so that the model can converge faster and adapt to new tasks, improve the generalization ability of the model, and ensure that high prediction performance can be maintained in different tasks.

[0033] Figure 1 A schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented is shown. In the environment 100, in response to receiving a user's advertisement delivery request 120 for a target object, the electronic device 110 determines feature data 130 corresponding to the advertisement delivery request 120. The electronic device 110 can then use the trained traffic estimation model 140 to determine a traffic estimation result 150 for the advertisement delivery request based on the feature data 130.

[0034] In the environment 100, the electronic device 110 can be any type of device with computing capabilities, including a terminal device or a server device. The terminal device can be any type of mobile terminal, fixed terminal or portable terminal, including a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an e-book device, a gaming device, or any combination of the foregoing, including accessories and peripherals of these devices or any combination thereof. The server device can include, for example, a computing system / server, such as a mainframe, an edge computing node, a computing device in a cloud environment, and the like.

[0035] It should be understood that the structure and functionality of environment 100 are described for exemplary purposes only and does not imply any limitation on the scope of the present disclosure.

[0036] Figure 2 1 shows a flow chart of a process 200 for traffic estimation according to some embodiments of the present disclosure. The process 200 may be implemented in the environment 100, for example, the process 200 may be implemented at the electronic device 110. In the following, for ease of discussion, reference will be made to Figure 1 The process 200 is described below.

[0037] In block 210 , in response to receiving an advertisement delivery request 120 from a user for a target object, the electronic device 110 determines a point of interest corresponding to the advertisement delivery request 120 .

[0038] In some embodiments, the electronic device 110 receives an advertisement delivery request 120 from a user. Typically, these advertisement delivery requests may be for specific target objects, such as express packaging boxes, vehicle advertisements, leaflets or brochure advertisements attached to packages, or other advertising media. The electronic device 110 determines the user's advertising needs through the user's advertisement delivery request, such as the point of interest corresponding to the advertisement delivery request, the expected advertising traffic, etc.

[0039] In some embodiments, a point of interest (POI) refers to a location that is considered interesting or useful in a map or geographic information system, usually represented by geographic coordinates. In the user's advertising needs, a POI can be understood as a specific location or area where the user wants to place an advertisement. The POI may be some specific cities, communities, business districts, etc. These locations are where the user believes that their target consumer groups often appear or may be exposed to advertisements.

[0040] As an example, in the context of logistics advertising, POIs can refer not only to a certain geographical location, but also to specific links related to the logistics process. For example, the shipping point of a specific warehouse, the distribution station, the route of a courier vehicle, etc. can all be considered POIs. These points are key contact points in logistics activities, and advertising exposure can be effectively achieved through these points.

[0041] Continue to refer Figure 2 In box 220, the electronic device 110 determines the feature data 130 corresponding to the point of interest.

[0042] In some embodiments, after determining the POI corresponding to the advertisement delivery request 120, the electronic device 110 needs to collect and determine the feature data 130 related to the POI. These feature data will be used in the machine learning model for advertisement traffic estimation to more accurately predict the advertisement delivery effect.

[0043] In some embodiments, the feature data 130 includes at least one of the following: historical delivery data corresponding to the point of interest, user behavior data, and warehouse logistics data associated with the target object.

[0044] Figure 3A A schematic diagram of a process 300A for determining feature data 130 according to some embodiments of the present disclosure is shown. The process 300A may be implemented at the electronic device 110.

[0045] In block 311, the electronic device 110 determines historical delivery data of the point of interest within a predetermined time period. The historical delivery data at least includes the category and quantity of the advertisement delivery requests received within the predetermined time period.

[0046] In some embodiments, the historical delivery data corresponding to the POI may refer to the relevant data generated when advertising is delivered at a specific POI point within a certain period of time in the past. For example, the historical delivery data may include the category and quantity of advertising delivery requests for the POI point within a predetermined time period.

[0047] In some embodiments, the historical delivery data corresponding to the POI may also include the content, category and corresponding quantity of the advertisement, the exposure of the advertisement on the day of delivery, whether there is a promotion activity, whether the day is a legal holiday, etc. The electronic device 110 can understand the changing trend of the advertisement delivery effect of the POI at different time periods and different activity nodes by analyzing the historical delivery data, and provide a reference for future advertisement traffic estimation.

[0048] In block 312 , the electronic device 110 determines user behavior data of a user whose delivery location is a point of interest within a predetermined time period.

[0049] In some embodiments, user behavior data may refer to the behavior data of people whose delivery destination is the corresponding POI point within a predetermined time period, including but not limited to the user's search behavior, browsing behavior, add-to-cart behavior, purchase behavior, stay time, etc. These feature data reflect the online activity and purchase intention of users related to the POI point. The electronic device 110 can use user behavior data to help the traffic prediction model identify potential advertising audiences and predict future advertising traffic, thereby optimizing the targeting of advertising content.

[0050] In block 313, the electronic device 110 determines warehouse logistics data associated with the target object in the warehouse corresponding to the point of interest within a predetermined time period. The warehouse logistics data associated with the target object at least includes the inventory and shipment volume corresponding to the target object.

[0051] In some embodiments, warehouse logistics data (also referred to as warehouse capacity data) may refer to data such as warehouse inventory, logistics transportation, and distribution efficiency related to specific POIs and advertising targets. Warehouse logistics data may be pre-stored in the electronic device 110 or the electronic device 110 receives data provided by an external system such as a warehouse management system or a logistics management system.

[0052] In some embodiments, the electronic device 110 can determine whether the logistics system is capable of processing and delivering products related to the advertisement in a timely manner based on the warehouse logistics data. For example, if the target of the advertisement is express packaging of a certain type of goods, the warehouse logistics data can help evaluate whether the warehouse has sufficient inventory and logistics capabilities to meet the required flow of advertising. If the warehouse logistics data indicates that the warehouse capacity is insufficient, it may be necessary to adjust the advertising delivery strategy to avoid poor advertising results due to insufficient logistics capabilities.

[0053] In the disclosed embodiment, by collecting and using historical delivery data, user behavior data and warehouse logistics data in multiple dimensions, the characteristics and trends of the location can be captured more accurately, thereby more comprehensively reflecting the various factors affecting advertising traffic, reducing deviations in predictions, and improving the accuracy of traffic estimates.

[0054] Continue to refer Figure 2 In box 230, the electronic device 110 uses the trained traffic estimation model 140 to determine the traffic estimation result 150 for the point of interest corresponding to the advertisement delivery request based on the feature data.

[0055] In some embodiments, the initialization weights of the traffic prediction model 140 can be determined using a meta-learning model. The meta-learning model is a special machine learning model whose core lies in "learning-to-learn". The meta-learning model not only focuses on the ability to solve specific tasks, but also focuses on improving the generalization ability and learning efficiency on different tasks by learning and optimizing its own learning process. The meta-learning model can learn how to generate effective initialization weights for new learning tasks. These initialization weights enable the model to converge to the optimal solution faster when facing new tasks, thereby improving learning efficiency. Therefore, the electronic device 110 can use the meta-learning model to generate customized initialization weights for different POI locations based on the feature data 130 to optimize the performance of the traffic prediction model and improve the accuracy of the prediction.

[0056] Figure 3B FIG. 3 is a schematic diagram of a process 300B for determining a traffic estimation result 150 according to some embodiments of the present disclosure. The process 300B may be implemented at the electronic device 110 .

[0057] In box 321, the electronic device 110 performs a feature engineering operation on the feature data 130 to pre-process the feature data.

[0058] In some embodiments, feature engineering operations can significantly improve the performance of machine learning models. By converting, extracting, and selecting the original feature data, more effective and discriminative features can be generated. These features can better reflect the inherent laws and potential information of the data. Therefore, performing feature engineering operations before inputting the data into the machine learning model can make it easier for the model to learn the essential characteristics of the data, thereby improving the accuracy and efficiency of prediction or classification. The electronic device 110 can perform feature engineering operations on the feature data 130 to pre-process the feature data, thereby standardizing or normalizing the feature data 130 to ensure that all features are on the same scale, which helps the model converge faster and improves training stability.

[0059] In some embodiments, the electronic device 110 may convert at least part of the feature data of discrete type in the feature data into discrete features using one-hot encoding.

[0060] In some embodiments, discrete features refer to features with limited values, such as "whether it is a legal holiday" and "whether there is a promotion activity" included in historical delivery data. The electronic device 110 can use one-hot encoding to convert these discrete category features into binary vectors for easy processing by machine learning models.

[0061] As an example, for the feature of whether it is a statutory holiday: suppose the feature "statutory holiday" has two values: yes (Yes) or no (No). One-hot encoding converts this feature into two binary features: yes (Yes): [1,0], no (No): [0,1]. For the feature of whether there is a promotion: if the feature "promotion" has three possible values: no promotion (None), discount promotion (Discount), gift promotion (Gift), one-hot encoding converts it into a three-bit binary vector: no promotion (None): [1,0,0], discount promotion (Discount): [0,1,0], gift promotion (Gift): [0,0,1].

[0062] In some embodiments, the electronic device 110 may perform a normalization operation on at least part of the feature data that is of a continuous type in the feature data to convert at least part of the feature data of the continuous type into a continuous feature.

[0063] In some embodiments, continuous features refer to numerical data with a wide and continuous value range, such as “the number of ad delivery requests.” Normalization scales the values ​​of these features to a smaller range (usually [0, 1]) to reduce the magnitude differences between features.

[0064] As an example, for the feature of the number of ad delivery requests: suppose the number of ad delivery requests on a certain day is 1000, but the number of ad delivery requests on other days is between 50 and 500, which will result in different weights of feature values ​​in different dimensions. Through normalization, the number of ad delivery requests is scaled to the range of [0,1]. Assuming that the number of ad delivery requests on a certain day is 450, the value after normalization may be 0.9 (depending on the minimum and maximum values ​​of the number of ad delivery requests). For the feature of active user behavior corresponding to POI points: Assuming that the value of the feature "daily views" is between 100 and 10,000, after the electronic device 110 performs a normalization operation, the feature value of a day with 5,000 views may be scaled to 0.5.

[0065] In the disclosed embodiments, by performing feature engineering operations on feature data for preprocessing and conversion, more effective and discriminative features can be generated, thereby improving the performance of the traffic estimation model and removing redundancy and noise, adapting to model requirements, improving interpretability, and accelerating model training.

[0066] Continue to refer Figure 3B In box 322, the electronic device 110 inputs the preprocessed feature data 130 into the traffic estimation model 140 to determine the traffic estimation result 150 for the point of interest corresponding to the advertisement delivery request.

[0067] In some embodiments, the electronic device 110 inputs the preprocessed feature data 130 into the trained traffic estimation model 140, and uses the traffic estimation model 130 to calculate the advertising traffic estimation result 150 for a specific POI based on the input feature data. The initialization weight 323 of the traffic estimation model 140 is determined using a meta-learning model.

[0068] In some embodiments, the electronic device 110 may construct a traffic prediction model based on a long short-term memory neural network (LSTM). LSTM is a special recurrent neural network designed to process and predict important events with very long intervals and delays in time series. LSTM can effectively capture long-term dependencies in sequence data through its unique design structure, thereby performing well in processing long sequence data.

[0069] In some embodiments, advertising traffic has obvious time series characteristics, and the estimation of advertising traffic depends on time series data, because advertising traffic will be affected by factors such as user behavior data and delivery data in the past period of time. Therefore, when predicting advertising traffic, it is necessary to consider the changes of these factors over time. LSTM, as a special recurrent neural network, is good at processing this type of data with time series characteristics. It can capture long-term dependencies in the data and use these relationships to predict future traffic changes. For example, historical data from the past five days (such as the number of advertising delivery requests, user behavior, etc.) is used as input features to predict advertising traffic on the sixth day. This method predicts future traffic by establishing a time series model and using past data as the input features of the model to capture patterns and trends in the time series.

[0070] Figure 3C A schematic diagram of a process 300C of training a traffic prediction model 140 according to some embodiments of the present disclosure is shown. The process 300C may be implemented at the electronic device 110 .

[0071] In some embodiments, the traffic estimation model 140 may include a first long short-term memory neural network 331 and a second long short-term memory neural network 332 .

[0072] In some embodiments, the electronic device 110 uses the first long short-term memory neural network 331 to determine a first feature vector corresponding to the historical delivery data and the user behavior data based on the initialization weights determined using the meta-learning model.

[0073] In some embodiments, the electronic device 110 may concatenate the historical delivery data and user behavior data obtained by performing feature engineering operations and input them into the first long short-term memory neural network 331 to obtain a corresponding first feature vector.

[0074] In some embodiments, the historical delivery data and user behavior data corresponding to the POI points mainly reflect the dynamic changes in user behavior and market demand, and have a direct and significant impact on predicting future advertising traffic. The historical delivery data provides the advertising delivery situation in the past period of time, while the user behavior data indicates the preferences and activity level of the target population for the advertisement. By splicing these two types of feature data together and extracting the first feature vector features through the first long short-term memory neural network 331, it is possible to comprehensively consider historical trends and current market dynamics, identify patterns and dependencies in time series, and better understand the dynamic changes of advertising traffic.

[0075] In some embodiments, the parameters of LSTM (including weights and biases) are usually randomly initialized at the beginning of training. This is because the network has not done any learning on the data before training, so the initial parameters can be set in a random way. The parameter initialization method will affect the convergence speed and final effect of the model. Therefore, good initialization weights can make the model converge to a better local optimal solution faster, while poor initialization weights may lead to slow convergence or fall into a poor local optimal solution.

[0076] In some embodiments, the electronic device 110 may use a meta-learning model to determine the initialization weights of the first long short-term memory neural network 331. The meta-learning model may be used to determine the initialization weights of the LSTM so that the model converges faster in subsequent training processes and can better adapt to the characteristics of specific tasks. Therefore, the electronic device 110 may use the first long short-term memory neural network to determine the first feature vector corresponding to the historical delivery data and the user behavior data based on the initialization weights determined using the meta-learning model.

[0077] In some embodiments, the electronic device 110 uses the second long short-term memory neural network 332 to determine a second feature vector corresponding to the warehouse logistics data associated with the target object.

[0078] In some embodiments, warehouse logistics data represents the supply capacity of the supply chain, which is an important factor in determining the material medium that supports advertising traffic. By extracting the features of warehouse logistics data using the second long short-term memory neural network 332, the status of the supply chain can be understood, providing a reference for subsequent traffic estimation and inventory management.

[0079] In some embodiments, if all features are input into the same long short-term memory neural network, information redundancy and excessive model complexity may result. Through division of labor, unnecessary calculations can be reduced, and each long short-term memory neural network can process related features more centrally. The electronic device 110 is responsible for processing different types of features through the first long short-term memory neural network 331 and the second long short-term memory neural network 332, respectively, to achieve division of labor and cooperation of the model. This division of labor not only helps the model to focus more on the data features in its respective fields, but also improves the training efficiency and prediction accuracy of the model.

[0080] Continue to refer Figure 3C In box 333, the electronic device 110 performs a pooling operation on the first feature vector and the second feature vector to determine the target weight of the first long short-term memory neural network.

[0081] In some embodiments, the electronic device 110 aggregates the feature vectors obtained by the first LSTM neural network 331 and the second LSTM neural network 332 through a pooling method to obtain a comprehensive feature vector. This aggregated feature vector is used as a comprehensive feature input of the first LSTM neural network 331 to determine the configuration of the target weight of the first LSTM neural network 331, so that it can better integrate and utilize all feature information for advertising traffic estimation.

[0082] In some embodiments, the electronic device 110 uses the target weight of the first long short-term memory neural network 331 that is finally determined to perform traffic estimation and determine a traffic estimation result 334 for the point of interest corresponding to the advertisement delivery request.

[0083] In some embodiments, the electronic device 110 may use an optimization algorithm (such as a back propagation algorithm) to calculate the gradient of the loss function during the process of performing traffic estimation using the traffic estimation model after training, and fine-tune and verify the model to further improve the generalization ability of the model.

[0084] In the disclosed embodiment, by using the initialization weights provided by the meta-learning model, the training efficiency of traffic estimation can be improved, the model can be helped to converge faster, and it can better adapt to the time series characteristics of historical delivery data and user behavior data, thereby improving prediction accuracy. Moreover, by processing warehouse logistics data through the second long short-term memory neural network, we can deeply understand the impact of warehouse capacity on advertising traffic and improve the modeling ability of warehouse-related features. This separate processing method ensures that the characteristics of warehouse data and advertising data are fully mined and utilized, further improving the accuracy of traffic estimation.

[0085] In some embodiments, in addition to the long short-term memory neural network, the electronic device 110 may also use other time series prediction algorithm models to estimate the advertising traffic, for example, the following models may be used as alternatives to the long short-term memory neural network model:

[0086] AutoRegressive Integrated Moving Average (ARIMA): used to analyze and predict time series data. It transforms the time series into a stationary series and combines autoregressive and moving average components to build a model.

[0087] Gated Recurrent Unit (GRU): A simpler recurrent neural network model than LSTM. GRU controls the flow of information through gating mechanisms (such as reset gates and update gates) to capture dependencies in time series data.

[0088] In some embodiments, the electronic device 110 may generate initialization weights for the traffic estimation model by inputting feature data corresponding to the POI into a meta-learning model.

[0089] In the disclosed embodiment, the initialization weights generated by the meta-learning model can provide an optimization starting point for a specific task and reduce the time required for model training. Moreover, the meta-learning model can improve the generalization ability of the model by sharing knowledge between multiple tasks. The generated initialization weights are not only applicable to the current task, but also can use the knowledge learned in different tasks to improve the model's adaptability to new data.

[0090] Figure 3D FIG. 3 is a schematic diagram of a process 300D for training a meta-learning model according to some embodiments of the present disclosure. The process 300D may be implemented at the electronic device 110 .

[0091] In block 341 , the electronic device 110 acquires a training data set.

[0092] In some embodiments, the training data set includes feature data and advertising traffic corresponding to at least one task, and at least one task indicates a traffic estimation operation for at least one point of interest and / or at least one time period.

[0093] As an example, each point of interest (POI) can be considered as an independent task. For example, different POIs may have different user behavior patterns, different historical delivery data, etc. The meta-learning model can generate initialization weights adapted to different POIs by learning from multiple POI tasks. The training dataset corresponding to each task may contain feature data from different POIs.

[0094] As another example, ad traffic estimation in different time periods can also be considered as different tasks. For example, the ad traffic patterns may be different on weekends and weekdays, or the ad traffic characteristics may be different on holidays and non-holidays. By learning ad traffic estimation tasks for different time periods, the meta-learning model can better adapt to the impact of time changes. The training dataset corresponding to each task may also contain feature data based on different time periods. For example, the feature data for weekdays and weekends, holidays and non-holidays may be different. These feature data help the model capture the impact of time factors on ad traffic.

[0095] In some embodiments, the training data set may further include advertisement traffic corresponding to at least one task. The advertisement traffic is used as output data of the model to guide the model to learn the mapping relationship from feature data to advertisement traffic.

[0096] In some embodiments, after determining the training data set, the electronic device 110 can determine a meta-learning model that can learn to generate initialization weights, and design the structure of the meta-learning model according to task requirements and data characteristics. The meta-learning model usually includes a "meta-network" for generating task-specific parameters and one or more "sub-networks" for actual task processing.

[0097] In some embodiments, the electronic device 110 can train the model based on a suitable meta-learning algorithm. For example, commonly used meta-learning algorithms include: Model-Agnostic Meta-Learning (MAML), a gradient-based meta-learning method that updates network parameters through multiple iterations so that it can quickly adapt to new tasks on a small amount of sample data and improve its performance; a memory-based meta-learning method that guides the learning process of the current task by storing and reusing the experience of previous tasks, etc.

[0098] In box 342, the electronic device 110 uses the weight generator included in the meta-learning model to generate an initialization weight corresponding to at least one task based on feature data corresponding to at least one task included in the training data set.

[0099] In some embodiments, the electronic device 110 may introduce a weight generator, which may be a neural network or other types of generative models. The electronic device 110 may use the generator to accept feature data corresponding to the task as input and output the initialization weight corresponding to the task.

[0100] In some embodiments, by introducing a weight generator, the model can learn shared features from multiple different tasks. This learning method is called multi-task learning. Because each task contains different feature data and target outputs, the model can improve its performance on new tasks (unseen POIs or time periods) by learning the common features of different tasks, thereby improving the generalization ability of the model, so that the meta-learning model can generate more efficient initialization weights when reasoning about real feature data.

[0101] In box 343, the electronic device 110 uses a back-propagation algorithm to train the meta-learning model based on the initialization weight corresponding to at least one task generated by the weight generator and the advertisement traffic corresponding to at least one task.

[0102] In some embodiments, the electronic device 110 trains the meta-learning model using a prepared training data set. In each training iteration, the feature data of a task is provided as input and the corresponding advertising traffic is provided as output to the meta-learning model, and the parameters of the meta-learning model are updated through the back propagation algorithm so that it can better generate the initialization weights.

[0103] In the disclosed embodiment, by training the meta-learning model on multiple tasks, the common features between different tasks can be captured. This way of sharing knowledge makes the generated initialization weights not only applicable to the current task, but also better adaptable to other similar tasks that may be encountered in the future, improving the performance of the model in different task scenarios. This is because the meta-learning model takes into account the commonalities of multiple tasks during the training process, and the weight generation has higher robustness.

[0104] In some embodiments, the electronic device 110 may measure the effect of the traffic prediction model through the following indicators:

[0105] Mean Absolute Error (MAE): The model effect is evaluated by calculating the average value of the absolute error between the traffic estimation result and the actual traffic.

[0106] Mean Squared Error (MSE): The model effect is evaluated by calculating the average of the squares of the difference between the estimated flow rate and the actual flow rate. The evaluation results can also provide direction for further optimization of the model. For example, if the MSE is high, it means that the model has a large error in predicting high flow values, and it may be necessary to adjust the model structure or training strategy to improve the prediction ability of high flow values.

[0107] Mean Absolute Percentage Error (MAPE): The model effect is evaluated by calculating the average value of the percentage error between the predicted traffic estimate and the actual traffic relative to the true value.

[0108] As an example, Table 1 shows the model evaluation effect verification results of some embodiments of the present disclosure. Among them, the logistics scenario package external advertising business is used as the verification target object, and the AB test method is used to verify the effect of the traffic estimation solution in the embodiments of the present disclosure by comparing the data of the experimental group and the control group.

[0109]

[0110] Table 1 Model effect verification

[0111] As shown in Table 1, service+ scenarios generally refer to scenarios in which advertising is carried out in the service industry or in business scenarios involving customer service. The experimental group is a group that applies the traffic estimation scheme in the embodiment of the present disclosure, and optimizes the advertising delivery strategy based on the estimation results, and adjusts the advertising content, location or budget. The control group is a group that continues to use the existing advertising delivery method and / or traffic estimation scheme. By comparing the traffic estimation accuracy of the experimental group with that of the control group, it can be seen that the traffic estimation accuracy of the experimental group has increased by 10%. This shows that the traffic estimation scheme in the embodiment of the present disclosure is more accurate, the traffic estimation results are closer to the actual traffic, and can significantly improve the reliability of the estimation.

[0112] Table 2 shows the rationality verification results of the experimental group and the control group of some embodiments of the present disclosure.

[0113]

[0114] Table 2 Verification of rationality of experimental control group

[0115] As shown in Table 2, this process is to verify whether the distribution of the experimental group and the control group is reasonable, as well as the scientific design of the model effect verification method. Specifically, the purpose of rationality verification is to confirm whether there is a significant difference in the amount of advertising between the experimental group and the control group, so as to ensure that the two groups are at the same starting point when the experiment begins.

[0116] Among them, the alternative hypothesis H1 is a concept in statistics, which represents the hypothesis that the researcher wants to prove or accept. In this verification, the alternative hypothesis H1 may be a hypothesis about whether there is a significant difference between the experimental group and the control group in a certain specific indicator (such as the average daily delivery). The p-value is an important indicator in statistics used to determine whether a hypothesis is valid. It indicates the probability of observing the current data or more extreme data when the null hypothesis (H0, usually indicating that there is no significant difference between the two groups) is true. In Table 2, the P value is used to evaluate whether there is a significant difference in the average daily delivery between the experimental group and the control group. Specifically, if the P value is less than the preset significance level (usually 0.05), the null hypothesis H0 is rejected and the alternative hypothesis H1 is accepted, which means that there is a significant difference between the two groups.

[0117] By counting the data from the first 28 days, we obtained the P value conclusion, which showed that there was no significant difference in the baseline delivery and variance between the experimental group and the control group before the experiment. This result shows that at the beginning of the verification in Table 1, the initial conditions of the experimental group and the control group were similar, ensuring the reliability of the data.

[0118] In some embodiments, the electronic device 110 determines the traffic demand of the user corresponding to the advertisement delivery request for the target object.

[0119] In some embodiments, the user's ad delivery request may also indicate the user's traffic demand, that is, the user's expectation of the ad effect and traffic. When making an ad delivery request, the user usually has an expected traffic target, such as how many views, impressions, or conversions a certain ad is expected to receive within a specific time period.

[0120] In some embodiments, the electronic device 110 determines an advertisement delivery strategy for an advertisement delivery request based on the user's traffic demand and traffic estimation results.

[0121] In some embodiments, after the electronic device 110 generates the traffic estimation result based on the model, it can further determine the advertising delivery strategy for the target object based on the traffic demand corresponding to the user's advertising delivery request and the traffic estimation result, for example, determining the advertising delivery time, advertising delivery frequency, adjusting the advertising content, and optimizing the advertising budget allocation.

[0122] In some embodiments, the electronic device 110 may also determine an advertising pricing strategy, that is, the cost of setting an advertising spot, based on the traffic estimation results generated by the model and the user's advertising budget, so as to maximize the return on investment (ROI) of the advertisement.

[0123] In the disclosed embodiment, after determining the traffic estimation result, the advertising delivery strategy for the target object can be determined based on the traffic demand of the user corresponding to the user's advertising delivery request and the traffic estimation result. This process helps to optimize the effect of advertising delivery, so that the advertisement can more effectively cover the target audience, thereby improving the conversion rate and return on investment of the advertisement. By accurately predicting the traffic demand, advertisers can formulate more accurate delivery plans.

[0124] Figure 4 A block diagram of an apparatus 400 for traffic estimation according to some embodiments of the present disclosure is shown. The apparatus 400 may be implemented as or include the electronic device 110.

[0125] The device 400 includes an interest point determination module 410, which is configured to determine the interest point corresponding to the advertisement delivery request in response to receiving an advertisement delivery request from a user for a target object; a feature data determination module 420, which is configured to determine feature data corresponding to the interest point, the feature data including at least one of the following: historical delivery data corresponding to the interest point, user behavior data and warehouse logistics data associated with the target object; and a traffic estimation module 430, which is configured to use a trained traffic estimation model to determine the traffic estimation result for the interest point corresponding to the advertisement delivery request based on the feature data, and the initialization weight of the traffic estimation model is determined using a meta-learning model.

[0126] In some embodiments, the feature data determination module 420 is further configured to determine the historical delivery data of the point of interest within a predetermined time period, the historical delivery data at least including the category and quantity of advertising delivery requests received within the predetermined time period; determine the user behavior data of users whose delivery address is the point of interest within the predetermined time period; and determine the warehouse logistics data associated with the target object of the warehouse corresponding to the point of interest within the predetermined time period, the warehouse logistics data associated with the target object includes the inventory and shipment volume corresponding to the target object.

[0127] In some embodiments, the traffic estimation module 430 is further configured to perform feature engineering operations on the feature data to pre-process the feature data; and input the pre-processed feature data into the traffic estimation model to determine the traffic estimation result for the point of interest corresponding to the advertising delivery request.

[0128] In some embodiments, the traffic estimation module 430 is further configured to use one-hot encoding to convert at least part of the feature data of discrete type in the feature data into discrete features; and perform a normalization operation on at least part of the feature data of continuous type in the feature data to convert at least part of the feature data of continuous type into continuous features.

[0129] In some embodiments, the traffic estimation model includes a first long short-term memory neural network and a second long short-term memory neural network.

[0130] In some embodiments, the device 400 also includes a traffic estimation model training module 440, which is configured to use a first long short-term memory neural network to determine a first feature vector corresponding to historical delivery data and user behavior data based on the initialization weights determined using a meta-learning model; use a second long short-term memory neural network to determine a second feature vector corresponding to warehouse logistics data associated with a target object; and perform an aggregation and pooling operation on the first feature vector and the second feature vector to determine the target weight of the first long short-term memory neural network.

[0131] In some embodiments, the initialization weights of the traffic prediction model are determined by the following operation: inputting feature data corresponding to the point of interest into a meta-learning model to generate the initialization weights of the traffic prediction model.

[0132] In some embodiments, the device 400 also includes a meta-learning model training module 440, which is configured to obtain a training data set, the training data set including feature data and advertising traffic corresponding to at least one task, at least one task indicating a traffic estimation operation for at least one point of interest and / or at least one time period; using a weight generator included in the meta-learning model to generate an initialization weight corresponding to at least one task based on the feature data corresponding to at least one task included in the training data set; and using a back-propagation algorithm to train the meta-learning model based on the initialization weight corresponding to at least one task generated by the weight generator and the advertising traffic corresponding to at least one task.

[0133] In some embodiments, the device 400 also includes a delivery strategy determination module 460, which is configured to determine the user's traffic demand corresponding to the advertising delivery request for the target object; and determine the advertising delivery strategy for the advertising delivery request based on the user's traffic demand and traffic estimation results.

[0134] The modules included in the device 400 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units can be implemented using software and / or firmware, such as machine executable instructions stored on a storage medium. In addition to or as an alternative to machine executable instructions, some or all of the units in the device 400 can be implemented at least in part by one or more hardware logic components. As an example and not limitation, exemplary types of hardware logic components that can be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0135] Figure 5 1 shows a block diagram of an electronic device 500 in which one or more embodiments of the present disclosure may be implemented. It should be understood that Figure 5 The electronic device 500 shown is merely exemplary and should not constitute any limitation on the functionality and scope of the embodiments described herein.

[0136] like Figure 5 As shown, the electronic device 500 is in the form of a general electronic device. The components of the electronic device 500 may include, but are not limited to, one or more processors or processing units 510, a memory 520, a storage device 530, one or more communication units 540, one or more input devices 550, and one or more output devices 560. The processing unit 510 may be an actual or virtual processor and is capable of performing various processes according to a program stored in the memory 520. In a multi-processor system, multiple processing units execute computer executable instructions in parallel to improve the parallel processing capability of the electronic device 500.

[0137] The electronic device 500 typically includes a plurality of computer storage media. Such media may be any available media accessible to the electronic device 500, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 520 may be a volatile memory (e.g., registers, caches, random access memory (RAM)), a non-volatile memory (e.g., a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 530 may be a removable or non-removable medium, and may include a machine-readable medium, such as a flash drive, a disk, or any other medium, which may be capable of being used to store information and / or data and may be accessed within the electronic device 500.

[0138] The electronic device 500 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 5 As shown in , a disk drive for reading or writing from a removable, non-volatile disk (e.g., a "floppy disk") and an optical drive for reading or writing from a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to the bus (not shown) by one or more data media interfaces. The memory 520 may include a computer program product 525 having one or more program modules that are configured to perform various methods or actions of various embodiments of the present disclosure.

[0139] The communication unit 540 implements communication with other electronic devices through a communication medium. Additionally, the functions of the components of the electronic device 500 can be implemented in a single computing cluster or multiple computing machines that can communicate through a communication connection. Therefore, the electronic device 500 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0140] The input device 550 may be one or more input devices, such as a mouse, a keyboard, a tracking ball, etc. The output device 560 may be one or more output devices, such as a display, a speaker, a printer, etc. The electronic device 500 may also communicate with one or more external devices (not shown) through the communication unit 540 as needed, such as a storage device, a display device, etc., communicate with one or more devices that allow a user to interact with the electronic device 500, or communicate with any device that allows the electronic device 500 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0141] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which one or more computer instructions are stored, wherein the one or more computer instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0142] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products implemented according to the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0143] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0144] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0145] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple implementations of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of a module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some implementations as replacements, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0146] The above descriptions of various implementations of the present disclosure are exemplary, non-exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The selection of terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to the technology in the marketplace, or to enable other persons of ordinary skill in the art to understand the implementations disclosed herein.

Claims

1. A method for traffic estimation, comprising: In response to receiving a user's advertisement delivery request for a target object, determining a point of interest corresponding to the advertisement delivery request; Determine feature data corresponding to the point of interest, the feature data including at least one of the following: historical delivery data corresponding to the point of interest, user behavior data, and warehouse logistics data associated with the target object; as well as The trained traffic prediction model is used to determine the traffic prediction result of the point of interest corresponding to the advertisement delivery request based on the feature data, and the initialization weight of the traffic prediction model is determined by using a meta-learning model.

2. The method according to claim 1, wherein determining the feature data corresponding to the point of interest comprises: Determine historical delivery data of the point of interest within a predetermined time period, the historical delivery data at least including the category and quantity of advertisement delivery requests received within the predetermined time period; Determine user behavior data of users whose delivery location is the point of interest within the predetermined time period; as well as Determine warehouse logistics data associated with the target object in the warehouse corresponding to the point of interest within the predetermined time period, wherein the warehouse logistics data associated with the target object includes inventory and shipment corresponding to the target object.

3. The method according to claim 1, wherein using a trained traffic estimation model and based on the feature data, determining a traffic estimation result for the point of interest corresponding to the advertisement delivery request comprises: Performing feature engineering operations on the feature data to pre-process the feature data; as well as The preprocessed feature data is input into the traffic estimation model to determine the traffic estimation result for the point of interest corresponding to the advertisement delivery request.

4. The method according to claim 3, wherein performing feature engineering operations on the feature data comprises at least one of the following: Converting at least part of the feature data of discrete type in the feature data into discrete features using one-hot encoding; and A normalization operation is performed on at least part of the feature data that is of a continuous type in the feature data to convert at least part of the feature data of the continuous type into a continuous feature.

5. The method according to claim 1, wherein the flow prediction model comprises a first long short-term memory neural network and a second long short-term memory neural network, wherein the flow prediction model is trained by: Determine, using the first long short-term memory neural network, a first feature vector corresponding to the historical delivery data and the user behavior data based on the initialization weights determined using the meta-learning model; Determine, using the second long short-term memory neural network, a second feature vector corresponding to the warehouse logistics data associated with the target object; as well as A pooling operation is performed on the first feature vector and the second feature vector to determine a target weight of the first long short-term memory neural network.

6. The method according to claim 1, wherein the initialization weight of the traffic estimation model is determined by the following operations: The feature data corresponding to the point of interest is input into the meta-learning model to generate the initialization weights of the traffic estimation model.

7. The method according to claim 1, wherein the meta-learning model is trained by: Acquire a training data set, the training data set including feature data and advertising traffic corresponding to at least one task, the at least one task indicating a traffic estimation operation for at least one point of interest and / or at least one time period; Using a weight generator included in the meta-learning model, based on feature data corresponding to at least one task included in the training data set, generate an initialization weight corresponding to the at least one task; as well as The meta-learning model is trained by using a back-propagation algorithm based on the initialization weight corresponding to the at least one task generated by the weight generator and the advertisement traffic corresponding to the at least one task.

8. The method according to claim 1, further comprising: Determine the traffic demand of the user corresponding to the advertisement delivery request for the target object; as well as An advertisement delivery strategy for the advertisement delivery request is determined based on the traffic demand of the user and the traffic estimation result.

9. A device for flow estimation, comprising: An interest point determination module is configured to, in response to receiving a user's advertisement delivery request for a target object, determine an interest point corresponding to the advertisement delivery request; A feature data determination module is configured to determine feature data corresponding to the point of interest, wherein the feature data includes at least one of the following: historical delivery data corresponding to the point of interest, user behavior data, and warehouse logistics data associated with the target object; as well as The traffic estimation module is configured to use a trained traffic estimation model to determine a traffic estimation result for the point of interest corresponding to the advertising delivery request based on the feature data, and the initialization weight of the traffic estimation model is determined using a meta-learning model.

10. An electronic device, comprising: at least one processing unit; as well as At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 8 when executed by the at least one processing unit.

11. A computer-readable storage medium having a computer program stored thereon, wherein the computer program can be executed by a processor to implement the method according to any one of claims 1 to 8.

12. A computer program product tangibly stored in a computer storage medium and comprising computer executable instructions which, when executed by a device, cause the device to perform the method according to any one of claims 1 to 8.

Images