Estimation Method, Device, Medium and Electronic Device for Advertising Inventory

Abstract

The present disclosure provides a method, device, medium and electronic device for estimating advertising inventory. The method includes: constructing a sampling pool based on preset detailed data of advertising exposure, and obtaining the content dimension and at least two data dimensions other than the content dimension in the historical exposure log; counting the daily number of users under at least two data dimensions, and using the daily number of users to estimate the daily estimated number under at least two data dimensions; clustering the content dimension to obtain at least two categories, and calculating the number of at least two categories to obtain the content ratio; calculating the daily estimated number and the content ratio to obtain the sampling number, and sampling in the sampling pool according to the sampling number to obtain advertising inventory. On the one hand, the present disclosure improves the accuracy of advertising inventory estimation and saves the computational cost of estimation; on the other hand, it enriches the refinement of the estimated advertising inventory and provides good support for subsequent targeted dimension expansion.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method for estimating advertising inventory, an apparatus for estimating advertising inventory, a computer-readable medium, and an electronic device. Background Art

[0002] With the increasing number of smart devices appearing in people's lives, advertising push through these devices has become a common scenario. Contract advertising has become a mainstream form of advertising sales, whereby media outlets are required to achieve a specified number of impressions within a given timeframe, under specified targeting conditions, and in accordance with advertiser requirements.

[0003] Typically, time series estimation or sampling methods can be used to estimate future ad inventory from historical ad inventory. However, time series estimation methods cannot handle large amounts of time series and cannot accurately measure user granularity, resulting in insufficient accuracy in ad inventory estimates. Sampling methods also suffer from insufficient estimation accuracy. Summary of the Invention

[0004] The present disclosure aims to provide an advertising inventory estimation method, an advertising inventory estimation device, a computer-readable medium, and an electronic device, thereby at least to some extent solving technical problems such as inaccurate advertising inventory estimation.

[0005] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by practice of the present disclosure.

[0006] According to one aspect of an embodiment of the present disclosure, a method for estimating advertising inventory is provided, the method comprising: constructing a sampling pool based on preset detailed data of advertising exposures, and obtaining a content dimension and at least two data dimensions other than the content dimension from a historical exposure log;

[0007] Counting the number of daily users in the at least two data dimensions, and estimating the estimated daily number in the at least two data dimensions using the number of daily users;

[0008] Clustering the content dimensions to obtain at least two categories, and calculating the number of the at least two categories to obtain a content ratio;

[0009] The daily estimated quantity and the content ratio are calculated to obtain a sampling quantity, and the advertising inventory is obtained by sampling in the sampling pool according to the sampling quantity.

[0010] According to one aspect of an embodiment of the present disclosure, there is provided an apparatus for estimating advertising inventory, the apparatus comprising: a data acquisition module configured to construct a sampling pool based on preset detailed data of advertising exposures, and to acquire a content dimension and at least two data dimensions other than the content dimension from a historical exposure log;

[0011] a quantity estimation module, configured to count the daily number of users in the at least two data dimensions, and estimate the daily estimated quantity in the at least two data dimensions using the daily number of users;

[0012] a dimension clustering module configured to cluster the content dimensions to obtain at least two categories, and calculate the number of the at least two categories to obtain a content ratio;

[0013] The data sampling module is configured to calculate the daily estimated quantity and the content ratio to obtain a sampling quantity, and to perform sampling in the sampling pool according to the sampling quantity to obtain advertising inventory.

[0014] In some embodiments of the present disclosure, based on the above technical solution, the quantity estimation module includes: a quantity statistics submodule configured to count the historical number of users in the at least two data dimensions during the historical duration;

[0015] The joint estimation submodule is configured to estimate the daily estimated number under the at least two data dimensions by using the daily number of users and the historical number of users.

[0016] In some embodiments of the present disclosure, based on the above technical solution, the quantity statistics submodule includes: a duration determination unit configured to obtain an estimated duration of advertisement exposure and determine a historical duration corresponding to the estimated duration;

[0017] The on-time statistics unit is configured to collect the historical number of users in the at least two data dimensions according to the historical duration.

[0018] In some embodiments of the present disclosure, based on the above technical solution, the joint estimation submodule includes: a time series feature unit configured to extract the daily time series features of the daily number of users and extract the historical time series features of the historical number of users;

[0019] The connection processing unit is configured to perform full connection processing on the daily time series features and the historical time series features to obtain the daily estimated quantity under the at least two data dimensions.

[0020] In some embodiments of the present disclosure, based on the above technical solution, the time series feature unit includes: a first extraction subunit, configured to extract daily shallow features of the daily number of users, and extract daily time series features of the daily shallow features;

[0021] The second extraction subunit is configured to extract the historical shallow features of the historical number of users and extract the historical time series features of the historical shallow features.

[0022] In some embodiments of the present disclosure, based on the above technical solution, the first extraction subunit includes: a first network microunit, configured to perform feature extraction on the daily number of users using a convolutional neural network to obtain daily shallow features;

[0023] The second network micro-unit is configured to use a recurrent neural network to extract features from the daily shallow features to obtain daily time series features.

[0024] In some embodiments of the present disclosure, based on the above technical solution, the second extraction subunit includes: a third network micro-unit, configured to extract features of the historical number of users using a convolutional neural network to obtain historical shallow features;

[0025] The fourth network micro-unit is configured to use a recurrent neural network to extract features from the historical shallow features to obtain historical time series features.

[0026] In some embodiments of the present disclosure, based on the above technical solution, the connection processing unit includes: a combination feature subunit, configured to perform feature extraction on the at least two data dimensions to obtain a combination feature, and perform full connection processing on the daily time series feature and the historical time series feature to obtain a connection feature;

[0027] The multiplication processing subunit is configured to multiply the combined feature and the connection feature to obtain the daily estimated quantity under the at least two data dimensions.

[0028] In some embodiments of the present disclosure, based on the above technical solution, the combined feature subunit includes: an encoding processing microunit, configured to encode the at least two data dimensions using an attention encoding network to obtain at least two encoding vectors;

[0029] The feature combination micro unit is configured to perform feature extraction on the at least two coding vectors to obtain a combination feature between the at least two coding vectors.

[0030] In some embodiments of the present disclosure, based on the above technical solution, the data acquisition module includes: a log acquisition submodule configured to determine the advertisement exposure duration and acquire preset exposure logs within the advertisement exposure duration;

[0031] The detail acquisition submodule is configured to acquire preset detail data in the preset exposure log and construct a sampling pool using the preset detail data.

[0032] In some embodiments of the present disclosure, based on the above technical solution, the data sampling module includes: a data classification submodule configured to divide the preset detailed data in the sampling pool according to the at least two categories to obtain classified detailed data;

[0033] The random arrangement submodule is configured to randomly arrange the classification detailed data to obtain random detailed data, and to sample the random detailed data according to the sampling quantity to obtain advertising inventory.

[0034] In some embodiments of the present disclosure, based on the above technical solution, the dimension clustering module includes: a vector construction submodule configured to construct a content vector corresponding to the content dimension and cluster the content vector to obtain at least two categories;

[0035] The number calculation submodule is configured to count the number of the at least two categories and calculate the number to obtain the content ratio.

[0036] In some embodiments of the present disclosure, based on the above technical solution, the data acquisition module includes: a content acquisition submodule configured to acquire historical exposure logs and acquire content dimensions in the historical exposure logs;

[0037] The data dimension submodule is configured to obtain the platform dimension, region dimension, age dimension and gender dimension in the historical exposure log as at least two data dimensions in addition to the content dimension.

[0038] According to one aspect of an embodiment of the present disclosure, a computer-readable medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for estimating advertising inventory as in the above technical solution is implemented.

[0039] According to one aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the advertising inventory estimation method as in the above technical solution by executing the executable instructions.

[0040] In the technical solution provided by the embodiment of the present disclosure, the advertising inventory is obtained by sampling in a sampling pool constructed from preset detailed data based on the estimated daily estimated quantity. On the one hand, the estimation based on the daily estimated quantity can take into account the changes in the advertising inventory in the time series, and because the number of combinations of at least two data dimensions is limited, it is affordable and achievable in engineering. Therefore, while improving the accuracy of the advertising inventory estimation, it saves the estimated computing cost and optimizes the efficiency of the advertising inventory estimation. On the other hand, sampling in the sampling pool to obtain the advertising inventory makes the advertising inventory accurate to the user granularity, enriches the refinement of the estimated advertising inventory, and provides good support for the subsequent expansion of the targeted dimension.

[0041] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort. In the drawings:

[0043] Figure 1 Schematically shows an exemplary system architecture diagram applying the technical solution of the present disclosure;

[0044] Figure 2 A flowchart schematically illustrates the steps of a method for estimating advertising inventory in some embodiments of the present disclosure;

[0045] Figure 3 Schematically showing a flowchart of the steps of a method for constructing a sampling pool in some embodiments of the present disclosure;

[0046] Figure 4 Schematically illustrates a flowchart of steps of a method for obtaining content dimensions and data dimensions in some embodiments of the present disclosure;

[0047] Figure 5 A flowchart schematically illustrates the steps of a method for estimating daily estimated quantities in some embodiments of the present disclosure;

[0048] Figure 6 Schematically shows a flowchart of the steps of a method for counting the number of historical users in some embodiments of the present disclosure;

[0049] Figure 7 Schematically showing a flowchart of steps of a method for further estimating daily estimated quantities in some embodiments of the present disclosure;

[0050] Figure 8 Schematically illustrates a flowchart of steps of a method for extracting daily time series features and historical time series features in some embodiments of the present disclosure;

[0051] Figure 9 Schematically showing a flowchart of steps of a method for extracting daily time series features in some embodiments of the present disclosure;

[0052] Figure 10 Schematically illustrates a flowchart of steps of a method for extracting historical time series features in some embodiments of the present disclosure;

[0053] Figure 11 Schematically shows a flowchart of steps of a method for full connection processing in some embodiments of the present disclosure;

[0054] Figure 12 Schematically showing a flow chart of steps of a method for obtaining combined features in some embodiments of the present disclosure;

[0055] Figure 13 Schematically showing a flowchart of steps of a method for calculating content ratio in some embodiments of the present disclosure;

[0056] Figure 14 Schematically showing a flowchart of steps of a method for sampling and obtaining advertising inventory in some embodiments of the present disclosure;

[0057] Figure 15 The following schematically illustrates a flowchart of the steps of the advertising inventory estimation method in an application scenario according to an embodiment of the present disclosure;

[0058] Figure 16 The following schematically illustrates a detailed flow chart of the advertising inventory estimation method according to an embodiment of the present disclosure in an application scenario;

[0059] Figure 17 A schematic diagram of a model of a method for estimating daily estimated quantity in an application scenario according to an embodiment of the present disclosure is schematically shown;

[0060] Figure 18 The following is a flowchart schematically illustrating the steps of the sampling method in an application scenario according to an embodiment of the present disclosure;

[0061] Figure 19 Schematically shows a structural block diagram of an apparatus for estimating advertising inventory in some embodiments of the present disclosure;

[0062] Figure 20 The structure diagram of a computer system suitable for implementing the electronic device of the embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0063] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0064] In addition, the described features, structures or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, well-known methods, devices, implementations or operations are not shown or described in detail to avoid blurring various aspects of the present disclosure.

[0065] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0066] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0067] Traditional time series estimation methods in this field primarily rely on the Autoregressive Integrated Moving Average (ARIMA) model. The ARIMA model has three parameters: p, d, and q. p represents the autoregressive term, or the number of lags in the time series data used in the ARIMA model; d represents the number of differences required to make the time series data stationary, or the integration term; and q represents the number of sliding average terms.

[0068] The ARIMA model is formulated as formula (1):

[0069] (1)

[0070] Where L is the lag operator, .

[0071] The basic process of ARIMA model building is to first obtain time series data, observe whether the time series data is a stationary series, and perform stationary processing on the non-stationary series data. Furthermore, the autocorrelation coefficient and partial autocorrelation coefficient of the stationary series are obtained, and p and q are obtained.

[0072] In addition, a time series forecasting model (Trigonometric seasonality, Box-Cox transformation, ARMA errors, Trend and Seasonal components, or TBATS) can also be used for advertising inventory estimation. This method combines Fourier terms with exponential smoothing spatial models and the Box-Cox transformation, resulting in a fully automated forecasting approach.

[0073] Recently, the more commonly used time series prediction methods include deep learning models such as long short-term memory artificial neural networks (LSTM) and convolutional neural networks.

[0074] However, traditional ad inventory structures typically contain enormous amounts of combined data, potentially encompassing hundreds of millions of time series data points, and this data continues to expand as the business grows. Therefore, traditional time series estimation methods are difficult to apply in engineering applications. Furthermore, while deep learning-based models can handle massive amounts of time series data, they struggle to accurately measure ad inventory at the user level, making them incapable of supporting refined ad inventory estimation.

[0075] Based on the problems existing in the above solutions, the present disclosure provides an advertising inventory estimation method involving artificial intelligence, an advertising inventory estimation device, a computer-readable medium, and an electronic device.

[0076] Artificial Intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI also involves studying the design principles and implementation methods of various intelligent machines, enabling them to possess the capabilities of perception, reasoning, and decision-making.

[0077] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0078] Machine learning (ML) is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0079] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, autonomous driving, drones, robots, smart medical care, smart customer service, etc. It is believed that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0080] Figure 1 A schematic diagram of an exemplary system architecture applying the technical solution of the present disclosure is shown.

[0081] like Figure 1 As shown, the system architecture 100 may include a terminal 110 , a network 120 , and a server 130 . The terminal 110 and the server 130 are connected via the network 120 .

[0082] The terminal 110 can specifically be a desktop terminal or a mobile terminal. The mobile terminal can specifically be at least one of a smart phone, a tablet computer, a laptop computer, a smart speaker, a smart watch, etc. The desktop terminal can specifically be a desktop computer, but is not limited to this; the network 120 can be a communication medium of various connection types that can provide a communication link between the terminal 110 and the server side 130, such as a wired communication link, a wireless communication link or an optical fiber cable, etc., which is not specifically limited in this disclosure; the server side 130 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and basic cloud computing services such as big data and artificial intelligence platforms.

[0083] Specifically, the server side 130 constructs a sampling pool based on preset detailed data of advertisements exposed on the terminal 110, and obtains the content dimension and at least two data dimensions other than the content dimension in the historical exposure data; counts the daily number of users under at least two data dimensions, and uses the daily number of users to estimate the daily estimated number under at least two data dimensions; clusters the content dimension to obtain at least two categories, and calculates the number of at least two categories to obtain the content ratio; calculates the daily estimated number and the content ratio to obtain the sampling number, and samples in the sampling pool according to the sampling number to obtain advertising inventory for subsequent exposure on the terminal 110.

[0084] In addition, the advertising inventory estimation method in the embodiment of the present disclosure can be applied to the terminal 110 or the server 130, and the present disclosure does not specifically limit this. The embodiment of the present disclosure mainly uses the application of the advertising inventory estimation method to the server 130 as an example.

[0085] The following describes in detail the advertising inventory estimation method, advertising inventory estimation device, computer-readable medium, and electronic device provided by the present disclosure in conjunction with specific embodiments.

[0086] Figure 2 The following schematically illustrates a flowchart of the steps of the method for estimating advertising inventory in some embodiments of the present disclosure. Figure 2 As shown, the advertising inventory estimation method can mainly include the following steps:

[0087] Step S210: Build a sampling pool based on preset detailed data of advertisement exposure, and obtain the content dimension and at least two data dimensions other than the content dimension in the historical exposure log.

[0088] Step S220: Count the number of daily users in at least two data dimensions, and use the number of daily users to estimate the daily estimated number in at least two data dimensions.

[0089] Step S230: Cluster the content dimension to obtain at least two categories, and calculate the number of the at least two categories to obtain the content ratio.

[0090] Step S240: Calculate the daily estimated quantity and content ratio to obtain a sampling quantity, and perform sampling in the sampling pool according to the sampling quantity to obtain advertising inventory.

[0091] In an exemplary embodiment of the present disclosure, advertising inventory is obtained by sampling from a sampling pool constructed from preset detailed data based on the estimated daily quantity. On the one hand, estimating based on the daily estimated quantity can take into account the changes in advertising inventory over time series, and because the number of combinations of at least two data dimensions is limited, it is affordable and achievable in engineering. Therefore, while improving the accuracy of advertising inventory estimation, it also saves the estimated computing cost and optimizes the efficiency of advertising inventory estimation. On the other hand, sampling from the sampling pool to obtain advertising inventory makes the advertising inventory accurate to the user granularity, enriches the refinement of the estimated advertising inventory, and provides good support for subsequent expansion of targeted dimensions.

[0092] The following is a detailed description of each step in the advertising inventory estimation method.

[0093] In step S210, a sampling pool is constructed according to preset detailed data of advertisement exposure, and the content dimension and at least two data dimensions other than the content dimension in the historical exposure log are obtained.

[0094] In an exemplary embodiment of the present disclosure, the sampling pool may be constructed based on preset detailed data of recent advertisement exposure. The recent time period may be set based on actual circumstances and may be a few days, a week, or a month, etc., and this exemplary embodiment does not impose any particular limitation on this.

[0095] In an alternative embodiment, Figure 3 A flow chart showing the steps of the method for constructing a sampling pool is shown. Figure 3 As shown, the method includes at least the following steps: in step S310, determining the advertisement exposure duration, and obtaining a preset exposure log within the advertisement exposure duration.

[0096] The ad exposure duration can be the length of time the preset detailed data is collected to construct the sampling pool. Since most user behavior can be reflected in a week's worth of historical data, and a week's worth of data is sufficient for subsequent sampling tasks, the typical ad exposure duration can be set at seven days.

[0097] During these seven days, every time an ad is exposed, an exposure log is generated, also known as the preset exposure log. This preset exposure log includes both targeting information and content information. Targeting information includes user information, such as age, gender, region, and platform used; content information includes information related to the content of the ad exposure, such as whether it was a movie or TV series.

[0098] In step S320, the preset detailed data in the preset exposure log is obtained, and a sampling pool is constructed using the preset detailed data.

[0099] When a preset exposure log is obtained, the preset detailed data included therein can be extracted, such as detailed data such as the user's age, gender, region, and platform, as well as detailed data such as the content of the ad exposure. In addition, to allow for continued sampling based on advertisers' needs after this sampling, other detailed data may also be included, which is not specifically limited in this exemplary embodiment.

[0100] Furthermore, the acquired preset detailed data can be used to construct a sampling pool for subsequent sampling in preparation.

[0101] In this exemplary embodiment, a sampling pool is constructed based on preset detailed data in a preset exposure log. The construction method is simple and feasible, the data dimensions are rich and accurate, and it can provide good support for the expansion of directional dimensions.

[0102] Ad inventory estimates use historical information directly transferred to a future date to represent future impressions. This means that ad inventory estimates are derived based on trends in historical detail data. Therefore, obtaining historical detail data is crucial.

[0103] In an alternative embodiment, Figure 4 A flowchart showing the steps of the method for obtaining content dimension and data dimension is shown as follows: Figure 4 As shown, the method includes at least the following steps: in step S410, a historical exposure log is obtained, and content dimensions in the historical exposure log are obtained.

[0104] It is worth noting that historical exposure logs can be exposure logs acquired within a past period of time. The past period is generally set to 56 days, but other days can be set based on actual needs, and this exemplary embodiment does not specifically limit this. Furthermore, these 56 days can include not only the recent 56 days of this year, but also the 56 days of the same date last year, or even the 56 days of the same date the year before, etc. This exemplary embodiment specifically limits this.

[0105] Historical exposure logs may include content information. This information includes information related to the content of the ad exposure, such as whether the content is a movie or TV series. Because user content preferences are relatively stable over time, only the content related to the historical exposure logs is collected.

[0106] In step S420 , the platform dimension, region dimension, age dimension, and gender dimension in the historical exposure log are obtained as at least two data dimensions in addition to the content dimension.

[0107] In addition, historical exposure logs also include targeting information. This targeting information includes user information, such as age, gender, region, platform, and other dimensions. The platform dimension can be the exposure platform, such as the application or browser. Because merging data dimensions can make engineering easier, at least two of these dimensions can be obtained. Generally, platform, region, age, and gender are used as at least two data dimensions to provide feedback on user distribution.

[0108] In this exemplary embodiment, the content dimension, platform dimension, region dimension, age dimension and gender dimension can be obtained based on the recorded content in the historical exposure log, so as to facilitate the subsequent estimation of the distribution of future users and ensure the accuracy of the estimation.

[0109] In step S220 , the daily number of users in at least two data dimensions is counted, and the daily estimated number in at least two data dimensions is estimated using the daily number of users.

[0110] In an exemplary embodiment of the present disclosure, after obtaining records of at least two data dimensions in 56 days of historical exposure logs, the daily number of users in each of the at least two data dimensions may be counted.

[0111] For example, statistics can be collected for the following 56 days: the platform dimension is Android phone, the region dimension is Shanghai, the age dimension is 18, and the gender dimension is male; the daily user count can be collected for the platform dimension is iOS phone, the region dimension is Shanghai, the age dimension is 18, and the gender dimension is male; the daily user count can be collected for the platform dimension is Android phone, the region dimension is Shanghai, the age dimension is 18, and the gender dimension is female; the daily user count can be collected for the platform dimension is iOS phone, the region dimension is Shanghai, the age dimension is 18, and the gender dimension is female; etc. Furthermore, the daily estimated number of users can be estimated based on the daily user counts under multiple data dimensions.

[0112] It is worth noting that the daily number of users may be counted from the 56 days of this year, or may include the 56 days with the same dates from multiple years such as last year and the year before last. This exemplary embodiment does not impose any special limitation on this.

[0113] In an alternative embodiment, Figure 5 A flow chart showing the steps of a method for estimating daily estimated quantities is shown in FIG. Figure 5 As shown, the method includes at least the following steps: in step S510, in the historical time length, the number of historical users in at least two data dimensions is counted.

[0114] In an alternative embodiment, Figure 6 A flow chart showing a method for counting the number of historical users is shown in FIG. Figure 6 As shown, the method includes at least the following steps: in step S610, obtaining an estimated duration of advertisement exposure, and determining a historical duration corresponding to the estimated duration.

[0115] The estimated duration can be the duration for estimating the daily estimated number of users using the daily number of users. Generally, since an ad exposure cycle is 28 days, the estimated duration can be determined as 28 days. In addition, other estimated durations can be set according to actual circumstances, and this exemplary embodiment does not specifically limit this.

[0116] After determining that the 28 days to be estimated in the future are the estimated duration, the historical duration can be determined accordingly. The historical duration can be the 28-day dates before last year, the year before last, or the year before last. For example, when the estimated duration is 28 days from July 1, 2020 to July 28, 2020, the 28 days from July 1, 2019 to July 28, 2019 can be determined as the historical duration, and the 28 days from July 1, 2018 to July 28, 2018 can also be determined as the historical duration. The 28 days of 2019 and 2018 can also be determined as the historical duration at the same time. There can also be historical durations of other years, and this exemplary embodiment does not specifically limit this.

[0117] In step S620, the historical number of users in at least two dimensions is collected according to the historical duration.

[0118] After the historical duration is determined, the number of historical users in at least two dimensions within this historical duration may be further counted.

[0119] In this exemplary embodiment, the number of historical users under historical duration is collected to determine the periodic characteristics within the estimated duration, thereby improving the accuracy of subsequent estimation.

[0120] In step S520 , the daily user count and the historical user count are used to estimate the daily estimated count in at least two data dimensions.

[0121] In an alternative embodiment, Figure 7 A flowchart showing the steps of a method for further estimating the daily estimated quantity is shown in FIG. Figure 7 As shown, the method includes at least the following steps: in step S710, extracting the daily time series features of the daily number of users, and extracting the historical time series features of the historical number of users.

[0122] In an alternative embodiment, Figure 8 A flowchart showing the steps of the method for extracting daily time series features and historical time series features is shown in FIG. Figure 8 As shown, the method includes at least the following steps: in step S810, extracting daily shallow features of the daily number of users, and extracting daily time series features of the daily shallow features.

[0123] In an alternative embodiment, Figure 9 A flowchart showing the steps of the method for extracting daily time series features is shown in FIG. Figure 9 As shown, the method includes at least the following steps: in step S910, a convolutional neural network is used to extract features of the daily number of users to obtain daily shallow features.

[0124] A convolutional neural network (CNN) can consist of an input layer, convolutional layers, activation functions, pooling layers, and fully connected layers. The input layer inputs the daily user count for 56 days of one or more years; the convolutional layers extract features from the daily user count; the activation functions add nonlinear factors to the extracted feature data to improve its representation; the pooling layers compress the feature data to simplify computational complexity; and the fully connected layers output the compressed feature data, also known as the daily shallow features. These daily shallow features can be a 56×64-dimensional vector.

[0125] In step S920, a recurrent neural network is used to extract daily shallow features to obtain daily time series features.

[0126] Recurrent Neural Networks (RNNs) are neural networks that incorporate directed loops, meaning that the current output of a sequence is related to the previous output. RNNs consist of three layers: an input layer, a hidden layer, and an output layer. The hidden layer is a ring-shaped structure, connecting the input layer to the hidden layer and vice versa. The hidden layers of RNNs have self-recursive connections, meaning that the hidden layer input includes not only the current input layer output but also the previous hidden layer output. Therefore, RNNs are also called multi-layer feedback neural networks.

[0127] Specifically, since the daily shallow features received by the input layer of the recurrent neural network are data sorted according to a certain time series, that is, the daily shallow features received by the input layer are sequence data. For the convenience of explanation, the daily shallow features are marked as 、 、……、 The dates corresponding to the daily shallow features are 、 、……、 , will be 、 、……、 The output data obtained by processing them separately, that is, the daily time series features are marked as 、 、……、 , and the steps in which the recurrent neural network processes each daily shallow feature in chronological order can be called time steps. Where n is the number of times the recurrent neural network processes daily shallow features.

[0128] In an unrolled recurrent neural network, The input data received by the input layer is , and Transmitted to the hidden layer, the hidden layer After processing, the data is transferred to the output layer to obtain Output data when . The input data received by the input layer is , the hidden layer can receive , and according to Output data when right Processed to transmit to the output layer, we get Output data when By continuously processing in this way, the daily time series features corresponding to the daily shallow features outputted are also a 56×64-dimensional vector.

[0129] It is worth noting that when the number of daily users for more than 56 days of a year is used for feature extraction, it means that data from the same period in history will be used. At this time, the convolutional neural network and the recurrent neural network can be called long-term models; and when only the number of daily users for 56 days of this year is used for feature extraction, the convolutional neural network and the recurrent neural network can be called basic models.

[0130] In this exemplary embodiment, convolutional neural networks and recurrent neural networks can be used to extract daily time series features, that is, the corresponding feature representation can be extracted by reconstructing the number of daily users in the past, so that multi-level features are better maintained during the extraction process, ensuring the accuracy of subsequent estimates.

[0131] In step S820, historical shallow features of the historical number of users are extracted, and historical time series features of the historical shallow features are extracted.

[0132] In an alternative embodiment, Figure 10 A flowchart showing the steps of the method for extracting historical time series features is shown in FIG. Figure 10 As shown, the method includes at least the following steps: in step S1010, a convolutional neural network is used to extract features of the historical number of users to obtain historical shallow features.

[0133] The convolutional neural network's input layer takes in historical user counts for 28 days over one or more years. The convolutional layer extracts features from these historical user counts. The activation function adds nonlinearity to the extracted feature data to improve its representation. The pooling layer compresses the feature data to simplify computational complexity. The fully connected layer outputs the compressed feature data, also known as the historical shallow features. These shallow features can be a 28×64-dimensional vector.

[0134] In step S1020, a recurrent neural network is used to extract historical shallow features to obtain historical time series features.

[0135] In the recurrent neural network, through multi-layer feedback processing in the input layer, hidden layer, and output layer, the historical time series features corresponding to the historical shallow features are finally output as a 28×64-dimensional vector. The specific processing method is the same as step S920 and will not be repeated here.

[0136] In this exemplary embodiment, convolutional neural networks and recurrent neural networks can be used to extract historical time series features, which can be used as a supplement to daily time series features to ensure the accuracy and efficiency of subsequent estimates.

[0137] In step S720 , full connection processing is performed on the daily time series features and the historical time series features to obtain the daily estimated quantity in at least two data dimensions.

[0138] In an alternative embodiment, Figure 11 A flowchart showing the steps of the method for full connection processing is shown in FIG. Figure 11 As shown, the method includes at least the following steps: in step S1110, feature extraction is performed on at least two data dimensions to obtain combined features, and full connection processing is performed on daily time series features and historical time series features to obtain connection features.

[0139] In an alternative embodiment, Figure 12 A flow chart showing the steps of the method for obtaining the combined features is shown in FIG. Figure 12 As shown, the method includes at least the following steps: in step S1210, at least two data dimensions are encoded using an attention encoding network to obtain at least two encoding vectors.

[0140] The attention encoding network can be a self-attention encoding network. The self-attention encoding network includes a self-attention layer and a forward propagation layer. Therefore, the at least two data dimensions input to the self-attention layer can include four discrete values: platform dimension, region dimension, age dimension, and gender dimension. The self-attention layer can fuse the four discrete values to obtain the word vectors of each discrete value, and the forward propagation layer performs forward propagation processing on the word vector to obtain the processed vector, which can be 、 、 and .in, is the encoding vector of the platform dimension, is the encoding vector of the regional dimension, is the encoding vector of age dimension, is the encoding vector of the gender dimension. This encoding vector can capture the correlation between individual dimensions.

[0141] In step S1220, feature extraction is performed on at least two coding vectors to obtain a combined feature between the at least two coding vectors.

[0142] Specifically, a deep cross network (DCN) can be used to extract features from at least two coding vectors. The deep cross network consists of an embedding and stacking layer, an intermediate layer consisting of a cross network and a deep network parallel to the cross network, and a combination layer. Whether the coding vector is a discrete feature or the coding vector is connected to form a continuous feature, it can be input into the embedding and stacking layer of the deep cross network. Furthermore, the cross network can perform pairwise crossover on the coding vectors, and the deep network can process the coding vectors as a fully connected feedforward neural network. Then, the combination layer connects the outputs of the two parallel cross networks and the deep network, and outputs a combined feature through a fully connected layer. The combined feature can be a vector of length 64.

[0143] In this exemplary embodiment, a combined feature between at least two coding vectors is extracted, which can capture the correlation relationship between the at least two coding vectors, has low computational cost, and does not require human intervention.

[0144] Furthermore, a full connection process is performed on the daily time series features and the historical time series features. The full connection process is to connect each node with all nodes in the previous layer, that is, to integrate the daily time series features and the historical time series features extracted previously.

[0145] Therefore, when the daily time series feature is a 56×64-dimensional vector and the historical time series feature is a 28×64-dimensional vector, connecting the two can obtain a 28×64-dimensional vector, which is the connection feature.

[0146] In step S1120 , the combined feature and the connected feature are multiplied to obtain the daily estimated quantity in at least two data dimensions.

[0147] After obtaining the combined features and the connection features, the combined features are used as a reference for the connection features. Therefore, the combined features and the connection features are multiplied together to perform the inner product, resulting in a 28×64-dimensional vector as the daily estimated quantity for the next 28 days.

[0148] In this exemplary embodiment, the accuracy of the daily estimated quantity can be ensured by using the combined features as time series features to simultaneously estimate the daily estimated quantity.

[0149] In step S230 , the content dimensions are clustered to obtain at least two categories, and the number of the at least two categories is calculated to obtain the content ratio.

[0150] In an exemplary embodiment of the present disclosure, before clustering the content dimensions, the content dimensions of each user may be constructed into a vector to calculate the content ratio.

[0151] In an alternative embodiment, Figure 13 A flowchart showing the steps of a method for calculating content ratio is shown in FIG. Figure 13 As shown, the method includes at least the following steps: in step S1310, constructing a content vector corresponding to the content dimension, and clustering the content vector to obtain at least two categories.

[0152] For example, a method for constructing a content vector corresponding to a content dimension is to assume that a user has viewed movie content 5 times and television content 10 times. Therefore, a content vector of length 2 can be constructed with the content of 5 and 10. Similarly, corresponding content vectors can be constructed for the content dimension of all users over a historical period, such as a month. Furthermore, the content vectors are clustered, for example, using unsupervised clustering, to obtain cluster centers.

[0153] The clustering method may be a K-means clustering algorithm. The K-means clustering algorithm is a distance-based clustering algorithm that uses clustering as a similarity evaluation metric. Specifically, it assumes that the closer the distance between two objects, the greater the similarity. The K-means clustering algorithm considers clusters to be composed of closely spaced objects, and therefore aims to obtain compact and independent clusters.

[0154] The specific process of the algorithm is as follows: first, randomly select K content vectors from N content vectors as centers; then, measure the distances of the remaining content vectors to each center content vector, and assign the remaining content vectors to the cluster with the nearest center; further, recalculate the centers of each cluster obtained until the new center is equal to the original center or less than the threshold, and the clustering is completed.

[0155] Typically, clustering content vectors yields 15 centers. Furthermore, weekly verification based on the content dimension over two months of usage confirms that at least 12 centers overlap between any two weeks, demonstrating that user preference centers are relatively stable and that content dimensions can be used to effectively represent user preferences.

[0156] It is worth noting that, given that the deviation of the content dimension in weekly units is not large, the update of the cluster center can be carried out on a larger time scale, such as a quarter or a year, etc., and this exemplary embodiment does not impose any special limitation on this.

[0157] In addition, other clustering methods may be selected for clustering according to actual conditions. This exemplary embodiment does not impose any special limitation on the clustering algorithm.

[0158] In step S1320, the number of at least two categories is counted, and the number is calculated to obtain the content ratio.

[0159] In this exemplary embodiment, after clustering to obtain multiple cluster centers, that is, classification, the number of classifications can be counted. Generally, 15 classifications can be counted.

[0160] Furthermore, the number of categories is calculated to obtain the content ratio of a content dimension. For example, when there are 15 categories and the content ratio corresponding to a content dimension is calculated, a division operation can be performed between 1 and 15 to obtain 1 / 15 as the content ratio. In addition, other calculation methods are also possible and are not specifically limited in this exemplary embodiment.

[0161] In this exemplary embodiment, content vectors corresponding to content dimensions are clustered to obtain the content proportions occupied by the content dimensions. The calculation method is simple and accurate, and can save computing costs.

[0162] In step S240 , the daily estimated quantity and the content ratio are calculated to obtain a sampling quantity, and sampling is performed in the sampling pool according to the sampling quantity to obtain advertising inventory.

[0163] In an exemplary embodiment of the present disclosure, after obtaining the daily estimated quantity and content ratio, the number of samples to be sampled in the sampling pool may be calculated and determined.

[0164] For example, the daily estimated quantity and the content ratio can be multiplied to obtain the corresponding sampling quantity. Specifically, when the daily estimated quantity for the next 28 days is: 1500, 3000, 1500, 4500, ..., 1500, and the content ratio is 1 / 15, 1 / 15 can be multiplied by the daily estimated quantity for each of the 28 days to obtain the sampling quantity to be sampled in the next 28 days, which is: 100, 200, 100, 300, ..., 100, respectively. In addition, other calculation methods can also be used, and this exemplary embodiment does not specifically limit this.

[0165] Furthermore, the advertising inventory is obtained by sampling in the constructed sampling pool according to the sampling quantity.

[0166] In an alternative embodiment, Figure 14 A flow chart showing the steps of a method for sampling and obtaining advertising inventory is shown in FIG. Figure 14 As shown, the method includes at least the following steps: in step S1410, the preset detailed data in the sampling pool is divided into at least two categories to obtain classified detailed data.

[0167] For example, when clustering the content dimension yields 15 categories, the preset detailed data in the sampling pool can be categorized to facilitate rapid and efficient sampling. That is, the preset detailed data is divided into 15 types of categorized detailed data according to the 15 categories. Other categorization methods are also possible, and this exemplary embodiment does not specifically limit this.

[0168] In step S1420, the classified detailed data is arranged in random order to obtain random detailed data, and the random detailed data is sampled according to the sampling quantity to obtain the advertising inventory.

[0169] In order to randomly sample the classified detailed data, the classified detailed data can be further arranged in random order to obtain random detailed data to avoid inaccurate sampling.

[0170] After obtaining the out-of-order detailed data, sampling may be performed in the out-of-order detailed data according to the sampling quantity to obtain the advertising inventory.

[0171] Since the final advertising inventory is generated by sampling, the advertising inventory not only meets the estimated quantity, but also includes all the detailed data generated by users in the exposure log, which can provide good support for the expansion of targeting dimensions.

[0172] The following describes in detail the advertising inventory estimation method provided in the embodiments of the present disclosure in conjunction with a specific application scenario.

[0173] Figure 15 A flowchart of the steps of the advertising inventory estimation method in an application scenario is shown. Figure 15 As shown, in step S1510, the original exposure log is obtained.

[0174] For further explanation Figure 15 For details of the data obtained in the step flow chart, you can continue to refer to Figure 16 .

[0175] Figure 16 The flowchart of the advertising inventory estimation method in the application scenario is shown as follows: Figure 16 As shown, in step S1610, original exposure data is obtained.

[0176] The detailed data recorded in the original exposure log is the original exposure data. The original exposure data may include preset detailed data, and may also include the content dimension in the historical exposure log and at least two data dimensions other than the content dimension.

[0177] In step S1520, a sampling pool is constructed.

[0178] This step is the same as step S1620. Specifically, the sampling pool can be constructed based on preset detailed data of recent advertisement exposure. The recent time period can be set based on actual conditions and can be a few days, a week, or a month, etc. This exemplary embodiment does not specifically limit this.

[0179] For example, the user's age, gender, region, and platform used can be obtained from the preset exposure logs within the past seven days. Furthermore, information related to the ad exposure content, such as whether the content is a movie or a TV series, can be obtained from the content information. This allows the sample pool to be constructed based on the targeting information and content information. Other detailed data may also be included, and this exemplary embodiment does not specifically limit this.

[0180] In step S1530, user data is counted.

[0181] Specifically, after obtaining records of at least two data dimensions in the historical exposure logs for 56 days, the daily number of users in each of the at least two data dimensions may be counted.

[0182] The number of daily users is as shown in step S1630 .in, For the platform dimension, For the regional dimension, For the age dimension, is the gender dimension. Specifically, the number of daily users can be counted as is 319; is 100; ...; It is 134.

[0183] In step S1540 , estimation is performed using an estimation algorithm.

[0184] To illustrate the principle of the estimation algorithm, you can refer to formula (2):

[0185] (2)

[0186] in, The probability of further dividing according to the content dimension under at least two data dimensions. Referring to step S1640, the probability of dividing according to the platform dimension, the region dimension, the age dimension and the gender dimension can be obtained. is 10, is 30, is 40, is 20. Correspondingly, is 0.1, is 0.3, is 0.4, is 0.2. Indicates the estimated number of users under all content dimensions.

[0187] Therefore, referring to step S1650, the convolutional neural network can estimate the future probability Furthermore, sampling is performed in the sampling pool according to the estimated future probability.

[0188] But in practice, the estimated and the total future ad inventory data is equivalent to estimating each User data under the combination.

[0189] Therefore, in step S1550, you can use Figure 17 The model in estimates the daily estimated quantity in at least two dimensions, that is, obtains future user data.

[0190] Figure 17 A schematic diagram of a model showing a method for estimating daily estimated quantities in an application scenario is shown in FIG. Figure 17 As shown, in step S1710, the number of daily users is obtained, where The number of daily users counted over M days in the past Y years is generally 56 days. The number of years Y can be determined based on the actual number of daily users collected, and can be one year or multiple years. This exemplary embodiment does not impose any particular limitation on this.

[0191] In step S1711, a convolutional neural network is used to extract features of the daily number of users to obtain daily shallow features.

[0192] For example, the daily shallow feature can be a 56×64 dimensional vector.

[0193] In step S1712, a recurrent neural network is used to extract daily shallow features to obtain daily time series features.

[0194] Corresponding to the daily shallow features, the daily time series features are also a 56×64-dimensional vector.

[0195] In step S1720, the historical number of users in at least two dimensions is collected according to the historical duration.

[0196] After determining the estimated duration of the next 28 days, the historical duration can be determined. The historical duration can be the 28 days before last year, the year before last, or the year before last. Indicates the number of historical users counted over N days in the past year Y-1. Typically, N is 28 days.

[0197] In step S1721, a convolutional neural network is used to extract features of the historical number of users to obtain historical shallow features.

[0198] For example, the historical shallow feature can be a 28×64-dimensional vector.

[0199] In step S1722, a recurrent neural network is used to extract historical shallow features to obtain historical time series features.

[0200] Corresponding to the historical shallow features, the historical time series feature is also a 28×64-dimensional vector.

[0201] In step S1723 , full connection processing is performed on the daily time series features and the historical time series features to obtain the daily estimated quantity in at least two data dimensions.

[0202] When the daily time series feature is a 56×64-dimensional vector and the historical time series feature is a 28×64-dimensional vector, connecting the two can obtain a 28×64-dimensional vector, which is the connection feature.

[0203] To obtain a more accurate daily estimate, you can also refer to the combined features of at least two data dimensions.

[0204] Specifically, in step S1730, at least two data dimensions are obtained, which may be four discrete values of platform dimension, region dimension, age dimension and gender dimension.

[0205] In step S1731, at least two data dimensions are encoded using an attention encoding network to obtain at least two encoding vectors.

[0206] Correspondingly, the encoding vector can be 、 、 and .in, is the encoding vector of the platform dimension, is the encoding vector of the regional dimension, is the encoding vector of age dimension, is the encoding vector of the gender dimension. This encoding vector can capture the correlation between individual dimensions.

[0207] In step S1732, feature extraction is performed using a deep cross network to obtain a combined feature between at least two encoding vectors.

[0208] Therefore, the combined feature can be a vector of length 64.

[0209] After obtaining the connection feature and the combination feature, in step S1724 , the combination feature and the connection feature may be multiplied to obtain the daily estimated quantity in at least two data dimensions.

[0210] By using the combined features as a reference for the connected features, a 28×64-dimensional vector can be obtained as the daily estimated quantity for the next 28 days.

[0211] In addition, after obtaining the combined feature, in step S1713 , the combined feature can be used as a reference for the daily time series feature and multiplied with the daily time series feature to obtain a daily time series feature with a better reconstruction effect.

[0212] In step S1560 , sampling is performed based on the forecast data, ie, the daily estimated quantity.

[0213] Figure 18 A flow chart showing the steps of the sampling method in the application scenario is shown, such as Figure 18 As shown, in step S1810, a sampling pool constructed according to preset detailed data is displayed.

[0214] The sampling pool may include: IOS system mobile phone, Beijing, male, 25 years old, content dimension 1; IOS system mobile phone, Beijing, male, 25 years old, content dimension 2; Android system mobile phone, Guangzhou, female, 18 years old, content dimension 1;...

[0215] In step S1820, the detailed data in the sampling pool are classified into IOS system mobile phone, Beijing, male, 25 years old, and content dimension 1, and are arranged in random order to obtain random detailed data.

[0216] The disordered detailed data may be user 1: attributes 1-n; user 2: attributes 1-n; user 3: attributes 1-n; and so on.

[0217] In step S1830, the estimation model is used to sample the disordered data to obtain a sampling result.

[0218] The sampling results obtained by this sampling may be two: User 1: Attributes 1-n and User 2: Attributes 1-n. Attributes 1-n represent dimensions other than the platform, region, age, and gender dimensions. Attributes 1-n may be all dimensions in the detailed data, and this exemplary embodiment does not impose any particular limitation on this.

[0219] In step S1570, the final estimated inventory is obtained.

[0220] Similar to the sampling results in step S1830, the number of users sampled and their corresponding attributes across various dimensions constitute the final estimated ad inventory. Therefore, the ad inventory not only meets the estimated quantity but also includes all detailed data generated by users in their exposure logs, providing strong support for the expansion of targeting dimensions.

[0221] Based on the above application scenarios, it can be seen that the advertising inventory estimation method provided by the embodiment of the present disclosure obtains advertising inventory by sampling in a sampling pool constructed from preset detailed data based on the estimated daily estimated quantity. On the one hand, estimating by using the daily estimated quantity can take into account the changes in advertising inventory in the time series, and since the number of combinations of at least two data dimensions is limited, it can be afforded and implemented in engineering. Therefore, while improving the accuracy of advertising inventory estimation, it saves the estimated computing cost and optimizes the efficiency of advertising inventory estimation. On the other hand, sampling in the sampling pool to obtain advertising inventory makes the advertising inventory accurate to the user granularity, enriches the degree of refinement of the estimated advertising inventory, and provides good support for the subsequent expansion of targeted dimensions.

[0222] It should be noted that although the steps of the method of the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all steps must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0223] The following describes an apparatus embodiment of the present disclosure, which can be used to execute the advertising inventory estimation method in the above-mentioned embodiment of the present disclosure. For details not disclosed in the apparatus embodiment of the present disclosure, please refer to the embodiment of the advertising inventory estimation method in the above-mentioned embodiment of the present disclosure.

[0224] Figure 19 The structure block diagram of the device for estimating advertising inventory in some embodiments of the present disclosure is schematically shown. Figure 19 As shown, the advertising inventory estimation device 1900 may mainly include: a data acquisition module 1910 , a quantity estimation module 1920 , a dimensional clustering model 1930 and a quantity sampling module 1940 .

[0225] The data acquisition module 1910 is configured to construct a sampling pool based on preset detailed data of advertisement exposure, and obtain the content dimension and at least two data dimensions other than the content dimension from the historical exposure log;

[0226] The number estimation module 1920 is configured to count the number of daily users in at least two data dimensions and estimate the daily estimated number in at least two data dimensions using the daily number of users;

[0227] The dimension clustering model 1930 is configured to cluster the content dimension to obtain at least two categories, and calculate the number of the at least two categories to obtain the content ratio;

[0228] The quantity sampling module 1940 is configured to calculate the daily estimated quantity and content ratio to obtain a sampling quantity, and perform sampling in the sampling pool according to the sampling quantity to obtain advertising inventory.

[0229] In some embodiments of the present disclosure, the quantity estimation module includes: a quantity statistics submodule configured to count the number of historical users in at least two data dimensions during a historical period;

[0230] The joint estimation submodule is configured to estimate the daily estimated number in at least two data dimensions by using the daily number of users and the historical number of users.

[0231] In some embodiments of the present disclosure, the quantity statistics submodule includes: a duration determination unit configured to obtain an estimated duration of advertisement exposure and determine a historical duration corresponding to the estimated duration;

[0232] The time statistics unit is configured to collect historical user numbers in at least two data dimensions according to historical duration.

[0233] In some embodiments of the present disclosure, the joint estimation submodule includes: a time series feature unit configured to extract daily time series features of the number of daily users and extract historical time series features of the number of historical users;

[0234] The connection processing unit is configured to perform full connection processing on the daily time series features and the historical time series features to obtain the daily estimated quantity in at least two data dimensions.

[0235] In some embodiments of the present disclosure, the time series feature unit includes: a first extraction subunit configured to extract daily shallow features of the daily number of users and extract daily time series features of the daily shallow features;

[0236] The second extraction subunit is configured to extract historical shallow features of the historical number of users and extract historical time series features of the historical shallow features.

[0237] In some embodiments of the present disclosure, the first extraction subunit includes: a first network microunit configured to perform feature extraction on the daily number of users using a convolutional neural network to obtain daily shallow features;

[0238] The second network micro-unit is configured to use a recurrent neural network to extract daily shallow features to obtain daily time series features.

[0239] In some embodiments of the present disclosure, the second extraction subunit includes: a third network microunit configured to perform feature extraction on the historical number of users using a convolutional neural network to obtain historical shallow features;

[0240] The fourth network micro-unit is configured to use a recurrent neural network to extract historical shallow features to obtain historical time series features.

[0241] In some embodiments of the present disclosure, the connection processing unit includes: a combination feature subunit configured to perform feature extraction on at least two data dimensions to obtain a combination feature, and perform full connection processing on the daily time series features and the historical time series features to obtain a connection feature;

[0242] The multiplication processing subunit is configured to multiply the combined feature and the connection feature to obtain the daily estimated quantity in at least two data dimensions.

[0243] In some embodiments of the present disclosure, the combined feature subunit includes: an encoding processing microunit configured to encode at least two data dimensions using an attention encoding network to obtain at least two encoding vectors;

[0244] The feature combination micro unit is configured to perform feature extraction on at least two coding vectors to obtain a combination feature between the at least two coding vectors.

[0245] In some embodiments of the present disclosure, the data acquisition module includes: a log acquisition submodule configured to determine the advertisement exposure duration and acquire a preset exposure log within the advertisement exposure duration;

[0246] The detail acquisition submodule is configured to obtain preset detail data in a preset exposure log and construct a sampling pool using the preset detail data.

[0247] In some embodiments of the present disclosure, the data sampling module includes: a data classification submodule configured to divide the preset detailed data in the sampling pool into at least two categories to obtain classified detailed data;

[0248] The shuffle submodule is configured to shuffle the classified detail data to obtain shuffled detail data, and to sample the shuffled detail data according to the sampling quantity to obtain the advertising inventory.

[0249] In some embodiments of the present disclosure, the dimension clustering module includes: a vector construction submodule configured to construct a content vector corresponding to a content dimension and cluster the content vector to obtain at least two categories;

[0250] The number calculation submodule is configured to count the number of at least two categories and calculate the number to obtain the content ratio.

[0251] In some embodiments of the present disclosure, the data acquisition module includes: a content acquisition submodule configured to acquire historical exposure logs and obtain content dimensions in the historical exposure logs;

[0252] The data dimension submodule is configured to obtain the platform dimension, region dimension, age dimension and gender dimension in the historical exposure log as at least two data dimensions in addition to the content dimension.

[0253] The specific details of the advertising inventory estimation device provided in each embodiment of the present disclosure have been described in detail in the corresponding method embodiment, and will not be repeated here.

[0254] Figure 20 A schematic diagram of the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present disclosure is shown.

[0255] It should be noted that Figure 20 The computer system 2000 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0256] like Figure 20 As shown, computer system 2000 includes a central processing unit (CPU) 2001, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 2002 or programs loaded from storage 2008 into random access memory (RAM) 2003. RAM 2003 also stores various programs and data required for system operation. CPU 2001, ROM 2002, and RAM 2003 are interconnected via bus 2004. An input / output (I / O) interface 2005 is also connected to bus 2004.

[0257] The following components are connected to the I / O interface 2005: an input section 2006 including a keyboard, a mouse, and the like; an output section 2007 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and speakers; a storage section 2008 including a hard disk; and a communication section 2009 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 2009 performs communication processing via a network such as the Internet. A drive 2010 is also connected to the I / O interface 2005 as needed. Removable media 2011, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 2010 as needed, so that computer programs read from the removable media can be installed in the storage section 2008 as needed.

[0258] In particular, according to embodiments of the present disclosure, the processes described in the various method flowcharts can be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for executing the methods illustrated in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 2009 and / or installed from removable media 2011. When executed by the central processing unit (CPU) 2001, the computer program performs the various functions defined in the system of the present application.

[0259] It should be noted that the computer-readable medium described in the embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0260] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0261] It should be noted that although several modules or units of the device for action execution are mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0262] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented through software or through a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of instructions to enable a computing device (such as a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0263] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein.

[0264] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for estimating advertising inventory, characterized in that: The method comprises: A sampling pool is constructed based on preset detailed data of advertising exposures, and a content dimension and at least two data dimensions other than the content dimension are obtained from historical exposure logs; wherein the content dimension is information related to the content of the advertising exposures, and the preset detailed data includes the at least two data dimensions and other data dimensions other than the at least two data dimensions; Counting the daily number of users in the at least two data dimensions in the historical exposure log, and estimating the daily estimated number in the at least two data dimensions using the daily number of users; Clustering the content dimensions to obtain at least two categories, and calculating the number of the at least two categories to obtain a content ratio; The daily estimated number and the content ratio are calculated to obtain a sampling number, and sampling is performed in the sampling pool according to the sampling number to obtain advertising inventory, wherein the advertising inventory is obtained by sampling users corresponding to the at least two data dimensions and the content dimension contained in the sampling pool according to the sampling number, and the advertising inventory includes the sampled number of users and attributes of various dimensions corresponding to the users.

2. The method for estimating advertising inventory according to claim 1, wherein: The estimating the daily estimated number of users in the at least two data dimensions by using the daily number of users includes: During the historical period, counting the number of historical users under the at least two data dimensions; The daily estimated number of users and the historical number of users are used to estimate the daily estimated number under the at least two data dimensions.

3. The method for estimating advertising inventory according to claim 2, wherein: The counting of the number of historical users in the at least two data dimensions during the historical period includes: Obtain an estimated duration of ad exposure and determine a historical duration corresponding to the estimated duration; The historical number of users in the at least two data dimensions is collected according to the historical duration.

4. The method for estimating advertising inventory according to claim 2, wherein: The estimating the daily estimated number of users in the at least two data dimensions by using the daily number of users and the historical number of users includes: Extracting daily time series features of the daily number of users, and extracting historical time series features of the historical number of users; Fully connect the daily time series features and the historical time series features to obtain daily estimated quantities in the at least two data dimensions.

5. The method for estimating advertising inventory according to claim 4, wherein: The extracting of the daily time series features of the daily number of users and the extracting of the historical time series features of the historical number of users includes: Extracting daily shallow features of the daily number of users, and extracting daily time series features of the daily shallow features; Extract the historical shallow features of the historical number of users, and extract the historical time series features of the historical shallow features.

6. The method for estimating advertising inventory according to claim 5, characterized in that: The extracting of daily shallow features of the daily number of users and extracting daily time series features of the daily shallow features includes: Using a convolutional neural network to extract features from the daily number of users to obtain daily shallow features; A recurrent neural network is used to extract the daily shallow features to obtain daily time series features.

7. The method for estimating advertising inventory according to claim 5, wherein: The extracting of the historical shallow features of the historical number of users and the extracting of the historical time series features of the historical shallow features include: Using a convolutional neural network to extract features of the historical number of users to obtain historical shallow features; A recurrent neural network is used to extract the historical shallow features to obtain historical time series features.

8. The method for estimating advertising inventory according to claim 4, wherein: The performing full connection processing on the daily time series features and the historical time series features to obtain the daily estimated quantity in the at least two data dimensions includes: Performing feature extraction on the at least two data dimensions to obtain a combined feature, and performing full connection processing on the daily time series feature and the historical time series feature to obtain a connection feature; The combined feature and the connection feature are multiplied to obtain the daily estimated quantity in the at least two data dimensions.

9. The method for estimating advertising inventory according to claim 8, wherein: The extracting features from the at least two data dimensions to obtain a combined feature includes: Encoding the at least two data dimensions using an attention encoding network to obtain at least two encoding vectors; Feature extraction is performed on the at least two encoding vectors to obtain a combined feature between the at least two encoding vectors.

10. The method for estimating advertising inventory according to claim 1, wherein: The step of constructing a sampling pool based on preset detailed data of advertisement exposure includes: Determine the exposure duration of the advertisement, and obtain preset exposure logs within the exposure duration of the advertisement; Obtain preset detailed data in the preset exposure log, and construct a sampling pool using the preset detailed data.

11. The method for estimating advertising inventory according to claim 1, wherein: Sampling the advertising inventory in the sampling pool according to the sampling quantity includes: Dividing the preset detailed data in the sampling pool according to the at least two categories to obtain classified detailed data; The classified detailed data are arranged in random order to obtain random detailed data, and the random detailed data are sampled according to the sampling quantity to obtain the advertising inventory.

12. The method for estimating advertising inventory according to claim 1, wherein: Clustering the content dimension to obtain at least two categories, and calculating the number of the at least two categories to obtain the content ratio, includes: Constructing a content vector corresponding to the content dimension, and clustering the content vector to obtain at least two categories; The number of the at least two categories is counted, and the number is calculated to obtain a content ratio.

13. An advertising inventory estimation device, characterized in that: The device comprises: a data acquisition module configured to construct a sampling pool based on preset detailed data of advertising exposures, and acquire a content dimension and at least two data dimensions other than the content dimension from historical exposure logs; wherein the content dimension is information related to the content of the advertising exposures, and the preset detailed data includes the at least two data dimensions and other data dimensions other than the at least two data dimensions; a quantity estimation module configured to count the daily number of users in the at least two data dimensions in the historical exposure log, and estimate the daily estimated quantity in the at least two data dimensions using the daily number of users; a dimension clustering module configured to cluster the content dimensions to obtain at least two categories, and calculate the number of the at least two categories to obtain a content ratio; The data sampling module is configured to calculate the daily estimated number and the content ratio to obtain a sampling number, and to sample the sampling pool according to the sampling number to obtain advertising inventory, wherein the advertising inventory is obtained by sampling the users corresponding to the at least two data dimensions and the content dimension contained in the sampling pool according to the sampling number, and the advertising inventory includes the sampled number of users and attributes of various dimensions corresponding to the users.

14. A computer-readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for estimating advertising inventory according to any one of claims 1 to 12 is implemented.

15. An electronic device, characterized in that: include: processor; as well as a memory for storing executable instructions of the processor; The processor is configured to perform the advertising inventory estimation method according to any one of claims 1 to 12 by executing the executable instructions.

16. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the advertising inventory estimation method according to any one of claims 1 to 12.

Images