Artificial intelligence-based cross-regional advertisement placement analysis decision method and system

Abstract

The application relates to the technical field of computers and discloses a cross-region advertisement putting analysis and decision method and system based on artificial intelligence, which comprises the following steps: responding to a cross-region advertisement putting instruction, analyzing the cross-region advertisement putting instruction, determining a target advertisement putting region in the cross-region advertisement putting instruction, obtaining multidimensional data of the target advertisement putting region, performing semantic analysis and coding conversion on the multidimensional data, constructing a structured regional data matrix, performing feature extraction on the regional data matrix based on a preset deep learning algorithm, mining potential features related to advertisement putting effects, inputting the potential features into an advertisement putting decision model, and obtaining an advertisement putting strategy for the target advertisement putting region output by the advertisement putting decision model. The application solves the problems that market dynamic changes and potential trends cannot be accurately predicted and advertisement putting strategies lack precision, and improves the precision and predictability of cross-region advertisement putting analysis and decision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a cross-regional advertisement putting analysis and decision method and system based on artificial intelligence. BACKGROUND

[0002] In today's business advertising field, advertisement putting analysis and decision plays a crucial role in the marketing effect and resource utilization efficiency of enterprises. The traditional advertisement putting analysis and decision method mainly relies on historical data statistical analysis and simple market research.

[0003] The method based on historical data statistical analysis usually collects the advertisement putting data of different regions in the past period, calculates the average putting effect index of each region according to these data, and formulates the future advertisement putting strategy based on this. However, this method only reflects the past situation and cannot accurately predict the dynamic changes and potential trends of the market, making it difficult to detect and adjust the putting strategy in advance and easily leading to the waste of resources in ineffective putting. The simple market research method focuses on understanding the basic characteristics, consumption habits and attitudes towards advertisements of consumers in different regions through questionnaire survey, interview and other ways. However, on the one hand, the sample size of market research is often limited and difficult to fully represent the consumer groups in the whole region, especially in the case of large geographical scope and large population, sample deviation may lead to misjudgment of the market situation. On the other hand, the information obtained by market research is mostly qualitative data, which lacks precision when converted into quantitative advertisement putting strategy. SUMMARY

[0004] The present application provides a cross-regional advertisement putting analysis and decision method and system based on artificial intelligence to solve the problem of inaccurate prediction of dynamic changes and potential trends of the market and lack of precision in advertisement putting strategy, and improve the precision and predictability of cross-regional advertisement putting analysis and decision.

[0005] In the first aspect, the present application provides a cross-regional advertisement putting analysis and decision method based on artificial intelligence, comprising:

[0006] In response to the cross-regional advertisement putting instruction, the cross-regional advertisement putting instruction is analyzed to determine the target advertisement putting region in the cross-regional advertisement putting instruction;

[0007] Multi-dimensional data of the target advertisement putting region is obtained; the multi-dimensional data includes historical advertisement putting data, regional population data, regional economic data, regional cultural data, regional media data and real-time market dynamic data;

[0008] The multi-dimensional data is subjected to semantic analysis and coding conversion to construct a structured regional data matrix;

[0009] extracting potential features related to the advertising effect based on a preset deep learning network on the regional data matrix; the potential features include regional consumption potential features, regional cultural preference features, regional media communication features and market trend features;

[0010] inputting the potential features into an advertising decision model to obtain an advertising strategy for the target advertising region output by the advertising decision model; the advertising strategy includes advertising creative adjustment suggestions, advertising channel selection, advertising time planning and advertising budget allocation; the advertising decision model is obtained based on sample features and corresponding advertising strategy label results.

[0011] In a second aspect, the present application further provides a cross-regional advertising analysis and decision system based on artificial intelligence, which is applied to the cross-regional advertising analysis and decision method based on artificial intelligence as described in the first aspect. The cross-regional advertising analysis and decision system based on artificial intelligence comprises:

[0012] The analysis module is configured to analyze the cross-regional advertising instruction and determine the target advertising region in the cross-regional advertising instruction in response to the cross-regional advertising instruction.

[0013] The acquisition module is configured to acquire multi-dimensional data of the target advertising region; the multi-dimensional data includes historical advertising data, regional population data, regional economic data, regional cultural data, regional media data and real-time market dynamic data.

[0014] The data processing module is configured to perform semantic analysis and coding conversion on the multi-dimensional data to construct a structured regional data matrix.

[0015] The feature extraction module is configured to extract potential features related to the advertising effect based on a preset deep learning network on the regional data matrix; the potential features include regional consumption potential features, regional cultural preference features, regional media communication features and market trend features.

[0016] The analysis and decision module is configured to input the potential features into an advertising decision model to obtain an advertising strategy for the target advertising region output by the advertising decision model; the advertising strategy includes advertising creative adjustment suggestions, advertising channel selection, advertising time planning and advertising budget allocation; the advertising decision model is obtained based on sample features and corresponding advertising strategy label results.

[0017] In a third aspect, the present application further provides an electronic device comprising: a memory for storing a computer software program; and a processor for reading and executing the computer software program, thereby implementing the method for analyzing and deciding on cross-regional advertisement placement based on artificial intelligence according to the first aspect above.

[0018] In a fourth aspect, the present application further provides a non-transitory computer readable storage medium having a computer software program stored therein, the computer software program being executed by a processor to implement the method for analyzing and deciding on cross-regional advertisement placement based on artificial intelligence according to the first aspect above.

[0019] In a fifth aspect, the present application further provides a computer program product comprising a computer program, the computer program being executed by a processor to implement the method for analyzing and deciding on cross-regional advertisement placement based on artificial intelligence according to the first aspect above.

[0020] The method for analyzing and deciding on cross-regional advertisement placement based on artificial intelligence provided by the present application, on one hand, collects multi-dimensional data of different regions, covering historical advertisement placement data, regional population data, regional economic data, regional cultural data, regional media data and real-time market dynamic data, so that the advertisement placement strategy can be quickly adjusted, and the accuracy of cross-regional advertisement placement analysis and decision-making is greatly improved. The preset deep learning algorithm is used for feature extraction to mine potential features related to advertisement placement effect, which cannot be found by traditional analysis methods, thereby providing a basis for early layout of advertisement placement. In combination with the advertisement placement decision-making model of artificial intelligence, the dynamic changes and potential trends of the market can be accurately predicted, so that the placement strategy can be adjusted in advance, strong support is provided for formulating accurate advertisement placement strategies, and the predictability of cross-regional advertisement placement analysis and decision-making is improved. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is a flowchart of the method for analyzing and deciding on cross-regional advertisement placement based on artificial intelligence provided by the present application;

[0022] Figure 2 is a structural schematic diagram of the system for analyzing and deciding on cross-regional advertisement placement based on artificial intelligence provided by the present application;

[0023] Figure 3 is an embodiment schematic diagram of the electronic device provided by the present application;

[0024] Figure 4 is an embodiment schematic diagram of the computer readable storage medium provided by the present application. DETAILED DESCRIPTION

[0025] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described in the description of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0026] In the description of the present application, the terms "first", "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0027] In the description of the present application, the term "for example" is used to mean "serving as an example, instance, or illustration". Any embodiment described as "for example" in the present application is not necessarily to be construed as more preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to practice the present application. In the following description, details are set forth in order to provide a thorough understanding of the present application. It will be apparent to those skilled in the art that the present application can be practiced without using these specific details. In other instances, well-known structures and processes have not been described in detail in order to avoid unnecessarily obscuring the description of the present application. Therefore, the present application is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0028] Optionally, with reference to Figure 1 , Figure 1 is a flowchart of the method for analyzing and deciding cross-regional advertisement putting provided by the present application. The method for analyzing and deciding cross-regional advertisement putting based on artificial intelligence provided by the present application is applied to a putting analysis and decision system. Therefore, the method for analyzing and deciding cross-regional advertisement putting based on artificial intelligence comprises the following steps:

[0029] Step 10, in response to the cross-regional advertisement putting instruction, the cross-regional advertisement putting instruction is analyzed to determine the target advertisement putting region in the cross-regional advertisement putting instruction.

[0030] Optionally, when a user needs to analyze advertisement putting in a certain region, an instruction needs to be sent to the putting analysis and decision system, and the instruction carries the region to be analyzed for advertisement putting. The region can be a local region or a cross-region. The region in the embodiment of the present application is a cross-region.

[0031] Therefore, the delivery analysis and decision-making system responds to the cross-regional advertising delivery instruction, parses the cross-regional advertising delivery instruction, and determines the target advertising delivery area in the cross-regional advertising delivery instruction. Specifically, the cross-regional advertising delivery instruction is subjected to text parsing, and the instruction text is segmented and analyzed according to lexical and syntactic rules. In one embodiment, the instruction text is represented as Text, and the instruction text represented as Text is matched with words through the set word library W to find the keyword set K={k1, k2, k n}. Further, for each keyword, calculate its regional matching value F match (k i ), the specific calculation formula is as follows:

[0032]

[0033] Among them, C ij Indicates keyword k i and the jth known region identifier L in the vocabulary W j The association weight can be calculated by cosine similarity, and the value range is between 0 and 1; S ij Indicates keyword k i The jth known region identifier L in the vocabulary W j The frequency statistics of the occurrence in the relevant corpus. By calculating F match (k i ) For all keywords k i Then select F match The regions with the largest sum of values ​​are used as target advertising delivery areas, where the target advertising delivery area can be expressed as R = {r1, r2, ..., r p}.

[0034] Step 20: Obtain multi-dimensional data of the target advertising delivery area.

[0035] Furthermore, the placement analysis and decision-making system acquires multi-dimensional data on the target advertising area. This multi-dimensional data includes historical advertising placement data, regional demographic data, regional economic data, regional cultural data, regional media data, and real-time market dynamics data. Regional demographic data includes information such as the population size, age structure, and gender ratio of different regions. Regional economic data includes information such as per capita income, consumption levels, and industrial structure. Regional cultural data includes information such as local consumers' values, aesthetic concepts, and consumption habits. Regional media data includes the coverage, audience characteristics, and communication effects of various media types in different regions. Real-time market dynamics data captures changing market trends, such as new moves by competitors and sudden shifts in consumer demand.

[0036] In one embodiment, the acquisition of historical advertising delivery data:i The corresponding advertisement delivery platform database establishes a connection. Optionally, each advertisement delivery information record in the database is A ij (contains exposure, click volume, conversion rate, etc. indicators), time dimension is t, and is integrated into a historical advertisement delivery data vector

[0037]

[0038] Where Q represents the number of entries related to advertisement delivery information recorded in the database, ω ij is a weight coefficient set according to the advertisement type, delivery target, etc., used to measure the contribution of different advertisement delivery records A ij to the overall historical data, and the sum is 1.

[0039] For the acquisition of regional population data: from the interface of authoritative population statistics departments or professional data suppliers to obtain population-related data of each target advertisement delivery area r i , such as total population Age distribution (a represents age interval), gender ratio , etc., integrated into a regional population data vector V pop (r i ) :

[0040]

[0041] For the acquisition of regional economic data: collect economic indicators such as GDPE i (r GDP ), per capita income E income (r i ), industrial structure proportion E struct (r i , s) (s represents different industry categories) of each target advertisement delivery area r eco , and integrate them into a regional economic data vector V i (r eco ) :

[0042] V i (r GDP ) = [E i (r income ), E i (r struct ), E i (r i , s),...].

[0043] For the acquisition of regional cultural data: through multi-source data mining of regional cultural research agency reports, social media regional topic heat, etc.i The frequency of occurrence of cultural elements, such as cultural activity type C act (r i ,c)(c represents different cultural activities), cultural tradition preference C pref (r i ,t)(t represents different traditional elements), forming a regional cultural data vector V cult (r i ):

[0044] V cult (r i )=[C act (r i ,c1),C act (r i ,c2),C pref (r i ,t1),C pref (r i ,t2),...].

[0045] Acquisition of regional media data: Analyze each target advertising area i The coverage rate of various media (such as TV stations, newspapers, websites, social media platforms, etc.) cov (r i ,m)(m represents different media), user activity M act (r i ,m), forming a regional media data vector V media (r i ):

[0046] V media (r i )=[M cov (r i ,m1),M cov (r i ,m2),M act (r i ,m1),M act (r i ,m2)),...].

[0047] Acquisition of real-time market dynamic data: by real-time monitoring of market-related information platforms, e-commerce platform transaction data, etc., to obtain the target advertising area i Such as the current popular product category S prod (r i ,p)(p represents different products), price fluctuation coefficient S price (r i ,g)(g represents different commodity categories), integrated into real-time market dynamic data vector V market (ri ):

[0048] V market (r i )=[S prod (r i ,p1),S prod (r i ,p2),S price (r i ,g1),S price (r i ,g2),...].

[0049] Step 30: Perform semantic analysis and encoding conversion on the multi-dimensional data to construct a structured regional data matrix.

[0050] Furthermore, the placement analysis and decision-making system performs semantic analysis and encoding conversion on the multi-dimensional data. For the semantic analysis process: semantic annotation is performed on the data vector of each dimension, for example, for the historical advertising placement data vector Each element in the advertisement is labeled with its meaning according to the semantic definition related to the advertising effect, such as the high exposure element is labeled as "high spreadability", etc. Therefore, the historical advertising data vector can be Each element in is transformed into a semantic label set L x ={l x1 ,l x2 ,...}, the specific conversion formula is as follows:

[0051]

[0052] Where x represents the historical advertising data vector Each element in F represents the total number of semantic feature judgment functions; k (x) is the kth predefined semantic feature judgment function (such as judging whether the exposure is higher than a certain threshold, etc.), α k is the corresponding weight coefficient.

[0053] For the encoding conversion process: the delivery analysis and decision-making system converts the semantically annotated data elements into numerical codes E code (l ij ), mapped to specific numerical ranges according to different semantic labels, such as the "high spread" label corresponds to the numerical range [0.8, 1], "medium spread" corresponds to [0.4, 0.8), etc. After semantic analysis and encoding conversion of each element of all dimensional data vectors, the target advertising delivery area r is selected. i Arrange the encoded data into a structured regional data matrix ; Indicates the separation of data rows from different regions.

[0054] Step 40, based on the preset deep learning network, the regional data matrix is extracted, and the potential features related to the advertising effect are mined.

[0055] Further, the advertising analysis and decision system extracts features from the regional data matrix through the embedded preset deep learning network, and mines potential features related to the advertising effect, wherein the potential features include regional consumer potential features, regional cultural preference features, regional media communication features and market trend features.

[0056] Among them, the regional consumer potential feature can analyze the subtle differences in the potential consumption ability and consumption tendency of consumers in different regions. The regional cultural preference feature can analyze the potential connection between local cultural elements and advertising elements, such as the acceptance difference of certain colors, patterns, language styles in a specific region. The regional media communication feature can analyze the communication effect difference of different media in different regions, such as the active time distribution of social media users in different regions, the regional user stickiness difference of different media platforms, etc. The market trend feature can analyze the development direction of the market, for example, the development speed and potential of a new industry in different regions.

[0057] In an embodiment, the advertising analysis and decision system inputs the regional data matrix into the preset deep learning network, determines the contribution degree coefficient of each data feature of each data type in the regional data matrix to the advertising effect through convolution feature operation, and the specific convolution feature operation formula is as follows:

[0058]

[0059] Among them, A abt (z) represents the contribution degree coefficient of each data feature of each data type to the advertising effect, and z represents the feature value of each data feature of each data type.

[0060] Further, the advertising analysis and decision system sorts the contribution degree coefficient of each data feature of each data type to the advertising effect in descending order of numerical value, and extracts the target data feature of each data type whose contribution degree coefficient is ranked in the front preset position, wherein the front preset position is, for example, the top 10, that is, the data feature whose contribution degree coefficient is ranked in the top 10 in each data type is determined as the target data feature of each data type.

[0061] Further, the delivery analysis decision system classifies the target data features of each data type according to preset types, wherein the preset types of the embodiment of the present application include regional consumption potential, regional cultural preference, regional media communication and market trend, and thus the target data features of each data type are classified according to the types of regional consumption potential, regional cultural preference, regional media communication and market trend, to obtain regional consumption potential features, regional cultural preference features, regional media communication features and market trend features.

[0062] Step 50: inputting the potential features into the advertisement delivery decision model to obtain an advertisement delivery strategy for the target advertisement delivery area output by the advertisement delivery decision model.

[0063] Further, the delivery analysis decision system inputs the potential features into the embedded advertisement delivery decision model, and outputs an advertisement delivery strategy for the target advertisement delivery area through the advertisement delivery decision model, wherein the advertisement delivery decision model is obtained by training a preset neural network based on sample features and corresponding advertisement delivery strategy label results, and the advertisement delivery strategy includes advertisement creative adjustment suggestions, delivery channel selection, delivery time planning and delivery budget allocation.

[0064] In an embodiment, the advertisement delivery decision model includes a feature processing layer, an input layer, a plurality of hidden layers, an output layer and a strategy prediction layer, and thus the input is a potential feature vector F out with a dimension of q, wherein the potential features in the embodiment of the present application include regional consumption potential features, regional cultural preference features, regional media communication features and market trend features, and thus the dimension q = 4, the input layer of the advertisement delivery decision model directly receives the feature vector F out and transmits it to the subsequent hidden layers.

[0065] In order to better focus on the part of the potential features that has a stronger correlation with the advertisement delivery strategy, the embodiment of the present application introduces an attention mechanism, and the specific analysis is as follows: the input potential feature vector F att1 is transformed through a preset linear transformation matrix W out with a dimension of q*d, d is a preset intermediate dimension, for example, d = 8, to obtain a transformed feature vector H att1 = F out *W att1 .

[0066] Further, the attention weight S att1 (h i ) of each element h att ,i = 1, 2, 3, 4 in the transformed feature vector H i is calculated, and the specific calculation formula is as follows:

[0067]

[0068] Furthermore, the attention weight S att (h i ) performs softmax normalization to obtain the attention weight vector A att1 =[a1,a2,...,a q ],in:

[0069]

[0070] Furthermore, the attention weight A att1 =[a1,a2,...,a q ] applied to the input latent feature vector F out On the above, we get the weighted feature representation F att , where the weighted feature representation F att It can be expressed as:

[0071] F att =[a1*F out *(1),a2*F out (2),...,a q *F out (q)];

[0072] Optionally, the hidden layer of the advertising placement decision model in the embodiment of the present invention is composed of multiple fully connected layers, which are used to further extract features and build a mapping relationship with the advertising placement strategy. The first hidden layer in the advertising placement decision model has h1 neurons, such as h1 = 32, and the weight matrix (dimension is q*h1), weighted feature representation F is obtained through the first hidden layer att Processing is performed to obtain the output of the first hidden layer Among them, the output of the first hidden layer It can be expressed as:

[0073]

[0074] in, represents the output of the kth neuron in the first hidden layer, Indicates the initialization function, F att (i) represents the weighted feature representation F att The i-th element in Represents the weight matrix of the first hidden layer The weight element with index i,k in , represents the bias term of the kth neuron in the first hidden layer.

[0075]

[0076] Optionally, the output of the first hidden layer is input to the subsequent hidden layer in the ad serving decision model, and so on, for example, the second hidden layer has h2 neurons, such as h2 = 16, the weight matrix bias term is the same as the weight matrix bias term of the first hidden layer, to obtain the output of the last hidden layer in the ad serving decision model.

[0077] Optionally, the number of nodes in the output layer of the ad serving decision model corresponds to each dimension of the ad serving strategy. The ad serving strategy in the embodiment of the present application includes ad creative adjustment suggestion, delivery channel selection, delivery time planning and delivery budget allocation, a total of 4 dimensions, so the output layer has 4 nodes.

[0078] The connection weight W of the output layer out is a matrix of dimension h last *4 (h last is the number of neurons in the last hidden layer), and the bias term b out is a vector of length 4, so the initialization and the output S of the output layer = [S creat , S channel , S time , S budget ] respectively correspond to the predicted probability or predicted probability distribution of the ad creative adjustment suggestion, delivery channel selection, delivery time planning and delivery budget allocation.

[0079] Further, according to the probability interval in which the predicted probability of the ad creative adjustment suggestion, delivery channel selection, delivery time planning and delivery budget allocation is located, or the probability value in the predicted probability distribution is sorted, the ad creative adjustment suggestion, delivery channel selection, delivery time planning and delivery budget allocation of the target ad serving area are determined.

[0080] In the output layer, the calculation formula of the predicted probability of the ad creative adjustment suggestion is as follows:

[0081]

[0082] Where S creat represents the predicted probability of the ad creative adjustment suggestion; O last (j) represents the jth output of the last hidden layer, W out (j,0) represents the weight element with index j in the weight matrix W out of the last hidden layer, and b out(0) represents the bias term of the first neuron in the last hidden layer. Therefore, the corresponding advertising creative adjustment suggestion is matched according to the probability interval of the predicted probability of the advertising creative adjustment suggestion.

[0083] Among them, the predicted probability distribution S of the delivery channel selection channel The calculation formula is as follows:

[0084]

[0085] Among them, the predicted probability distribution of delivery channel selection includes the probability values ​​corresponding to channels such as social media platforms, TV advertising, and outdoor advertising. The final selected delivery channel is determined by sorting the probability values.

[0086] Among them, the predicted probability distribution S of the delivery time plan time The calculation formula is as follows:

[0087]

[0088] Among them, the predicted probability distribution of the delivery time planning includes the delivery weights or heat values ​​of different time periods (hours, days, weeks, etc.), and specific delivery time plans are formulated based on these values.

[0089] Among them, the predicted probability of budget allocation S budget The calculation formula is as follows:

[0090]

[0091] Therefore, according to the probability interval of the predicted probability of the advertising budget allocation, the corresponding advertising budget, that is, the specific advertising budget amount, is matched.

[0092] Furthermore, the loss function of the advertising decision model is:

[0093] loss total =α*loss creat +β*loss channel +γ*loss time +δ*loss budget ;

[0094] Among them, loss total Represents the loss function of the advertising decision model, loss creat Represents the loss function of the creative adjustment suggestion, loss channel Represents the loss function for channel selection, loss time Represents the loss function of time planning, loss budgetA loss function representing the allocation of the delivery budget, and alpha, beta, gamma, and delta represent preset important coefficients of the loss function.

[0095]

[0096] Wherein, C represents the number of categories of the creative adjustment suggestion, y creat (i) represents the true label of the creative adjustment suggestion, S creat (i) represents the predicted value of the creative adjustment suggestion.

[0097]

[0098] Wherein, T represents the dimension of the delivery time planning, S time (i) represents the predicted value of the delivery time planning, y time (i) represents the true label of the delivery time planning.

[0099] loss budget =|S budget -y budget |.

[0100] Wherein, S budget represents the predicted value of the delivery budget allocation, y budget represents the true label of the delivery budget allocation.

[0101] The embodiment of the present application can quickly adjust the advertising delivery strategy by collecting multi-dimensional data of different regions, covering historical advertising delivery data, regional population data, regional economic data, regional cultural data, regional media data, and real-time market dynamic data, thereby greatly improving the accuracy of cross-regional advertising delivery analysis and decision-making. The preset deep learning algorithm is used for feature extraction to mine potential features related to advertising delivery effect. These potential features are difficult to find by traditional analysis methods, thereby providing a basis for early layout of advertising delivery. In combination with the artificial intelligence advertising delivery decision-making model, the dynamic changes and potential trends of the market can be accurately predicted, so that the delivery strategy can be adjusted in advance, thereby providing strong support for formulating accurate advertising delivery strategies and improving the predictability of cross-regional advertising delivery analysis and decision-making.

[0102] The cross-regional advertising delivery analysis and decision-making system based on artificial intelligence provided by the present application is described below. The cross-regional advertising delivery analysis and decision-making system based on artificial intelligence described below can be correspondingly referred to the cross-regional advertising delivery analysis and decision-making method based on artificial intelligence described above. Optionally, refer to Figure 2 , Figure 2 The structure diagram of the cross-regional advertising delivery analysis and decision-making system based on artificial intelligence provided by the embodiment of the present application is shown in the figure. The cross-regional advertising delivery analysis and decision-making system based on artificial intelligence comprises:

[0103] The analysis module 210 is configured to analyze the cross-regional advertisement launching instruction in response to the cross-regional advertisement launching instruction, and determine a target advertisement launching region in the cross-regional advertisement launching instruction;

[0104] The acquisition module 220 is configured to acquire multi-dimensional data of the target advertisement launching region, wherein the multi-dimensional data comprises historical advertisement launching data, regional population data, regional economic data, regional cultural data, regional media data and real-time market dynamic data.

[0105] The data processing module 230 is configured to perform semantic analysis and coding conversion on the multi-dimensional data, and construct a structured regional data matrix.

[0106] The feature extraction module 240 is configured to perform feature extraction on the regional data matrix based on a preset deep learning network, and mine potential features related to advertisement launching effect, wherein the potential features comprise regional consumption potential feature, regional cultural preference feature, regional media communication feature and market trend feature.

[0107] The analysis and decision module 250 is configured to input the potential features into an advertisement launching decision model, and obtain an advertisement launching strategy for the target advertisement launching region output by the advertisement launching decision model, wherein the advertisement launching strategy comprises advertisement creative adjustment suggestion, launching channel selection, launching time planning and launching budget allocation, and the advertisement launching decision model is trained based on sample features and corresponding advertisement launching strategy label results.

[0108] The cross-regional advertisement launching analysis and decision system based on artificial intelligence provided by the application can rapidly adjust the advertisement launching strategy by collecting multi-dimensional data of different regions, thereby greatly improving the accuracy of cross-regional advertisement launching analysis and decision. The preset deep learning algorithm is used for feature extraction to mine potential features related to advertisement launching effect, which cannot be found by traditional analysis methods, thereby providing a basis for early advertisement launching layout. In combination with the advertisement launching decision model based on artificial intelligence, the dynamic changes and potential trends of the market can be accurately predicted, so that the launching strategy can be adjusted in advance, thereby providing strong support for formulating accurate advertisement launching strategy and improving the predictability of cross-regional advertisement launching analysis and decision.

[0109] Please refer to Figure 3 , Figure 3 The embodiment of the electronic device provided by the application is shown in the figure. Figure 3As shown, the embodiment of the present application provides an electronic device 300, which comprises a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and capable of running on the processor 320, and the processor 320 implements the following steps when executing the computer program 311:

[0110] In response to the cross-regional advertisement putting instruction, the cross-regional advertisement putting instruction is parsed to determine the target advertisement putting region in the cross-regional advertisement putting instruction;

[0111] Multi-dimensional data of the target advertisement putting region is acquired; the multi-dimensional data comprises historical advertisement putting data, regional population data, regional economic data, regional cultural data, regional media data, and real-time market dynamic data;

[0112] The multi-dimensional data is subjected to semantic analysis and coding conversion to construct a structured regional data matrix;

[0113] Based on a preset deep learning network, the regional data matrix is subjected to feature extraction to mine potential features related to advertisement putting effect; the potential features comprise regional consumption potential feature, regional cultural preference feature, regional media communication feature, and market trend feature;

[0114] The potential features are input into an advertisement putting decision model to obtain an advertisement putting strategy for the target advertisement putting region output by the advertisement putting decision model; the advertisement putting strategy comprises advertisement creative adjustment suggestion, putting channel selection, putting time planning, and putting budget allocation; the advertisement putting decision model is trained based on sample features and corresponding advertisement putting strategy label results.

[0115] Please refer to Figure 4 , Figure 4 An embodiment of the computer readable storage medium provided by the embodiment of the present application is shown in the figure. Figure 4 As shown, the embodiment provides a computer readable storage medium 400, which stores a computer program 311, and the computer program 311 is executed by a processor to implement the following steps:

[0116] In response to the cross-regional advertisement putting instruction, the cross-regional advertisement putting instruction is parsed to determine the target advertisement putting region in the cross-regional advertisement putting instruction;

[0117] Multi-dimensional data of the target advertisement putting region is acquired; the multi-dimensional data comprises historical advertisement putting data, regional population data, regional economic data, regional cultural data, regional media data, and real-time market dynamic data;

[0118] The multi-dimensional data is subjected to semantic analysis and coding conversion to construct a structured regional data matrix;

[0119] The preset deep learning network is used for feature extraction on the regional data matrix, and potential features related to the advertising effect are mined, including regional consumption potential features, regional cultural preference features, regional media communication features and market trend features.

[0120] The potential features are input into an advertising decision model to obtain an advertising strategy for the target advertising region output by the advertising decision model, including advertising creative adjustment suggestions, advertising channel selection, advertising time planning and advertising budget allocation.

[0121] In another aspect, the present application also provides a computer program product, which comprises a computer program that can be stored on a non-transitory computer readable storage medium, and the computer program can be executed by a processor to enable the computer to execute the above-mentioned artificial intelligence-based cross-regional advertising analysis and decision-making method, which comprises:

[0122] In response to the cross-regional advertising instruction, the cross-regional advertising instruction is analyzed to determine the target advertising region in the cross-regional advertising instruction.

[0123] Obtain multi-dimensional data of the target advertising region, including historical advertising data, regional population data, regional economic data, regional cultural data, regional media data and real-time market dynamic data.

[0124] Perform semantic analysis and coding conversion on the multi-dimensional data to construct a structured regional data matrix.

[0125] Feature extraction is performed on the regional data matrix based on a preset deep learning network to mine potential features related to the advertising effect, including regional consumption potential features, regional cultural preference features, regional media communication features and market trend features.

[0126] The potential features are input into an advertising decision model to obtain an advertising strategy for the target advertising region output by the advertising decision model, including advertising creative adjustment suggestions, advertising channel selection, advertising time planning and advertising budget allocation.

[0127] The system embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0128] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0129] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A cross-regional advertising analysis and decision-making method based on artificial intelligence, characterized by: include: Responding to a cross-region advertisement delivery instruction, parsing the cross-region advertisement delivery instruction, and determining a target advertisement delivery area in the cross-region advertisement delivery instruction; Acquire multi-dimensional data of the target advertisement delivery area; The multi-dimensional data includes historical advertising data, regional population data, regional economic data, regional cultural data, regional media data and real-time market dynamics data; Performing semantic analysis and encoding conversion on the multi-dimensional data to construct a structured regional data matrix; Extract features from the regional data matrix based on a preset deep learning network to mine potential features related to advertising effectiveness; the potential features include regional consumption potential features, regional cultural preference features, regional media communication features, and market trend features; Inputting the potential features into an advertising placement decision model, the advertising placement decision model outputting an advertising placement strategy for the target advertising placement area; the advertising placement strategy includes suggestions for adjusting advertising creatives, selecting placement channels, planning placement time, and allocating placement budgets; the advertising placement decision model is trained based on sample features and their corresponding advertising placement strategy label results; The advertising placement decision model includes a feature processing layer, an input layer, multiple hidden layers, an output layer, and a strategy prediction layer. The process of outputting an advertising placement strategy for a target advertising placement area includes: Inputting the potential features into the advertising placement decision model, performing feature processing on the potential features based on the feature processing layer combined with the attention mechanism to obtain a weighted feature representation; Inputting the weighted feature representation to the multiple hidden layers based on the input layer, and processing the weighted feature representation based on the multiple hidden layers to obtain an intermediate output; Processing the intermediate output based on the output layer to obtain the predicted probability or predicted probability distribution of the advertising creative adjustment suggestions, delivery channel selection, delivery time planning and delivery budget allocation output by the output layer; Based on the strategy prediction layer, the predicted probability or predicted probability distribution of the advertising creative adjustment suggestions, delivery channel selection, delivery time planning and delivery budget allocation is used to perform strategy prediction to obtain the advertising delivery strategy for the target advertising delivery area; The loss function of the advertising decision model is: loss total =α*loss creat +β*loss channel +γ*loss time +δ*loss budget ; loss budget =|S budget -y budget |; Among them, loss total Represents the loss function of the advertising decision model, loss creat Represents the loss function of the creative adjustment suggestion, loss channel Represents the loss function for channel selection, loss time Represents the loss function of time planning, loss budget represents the loss function of the budget allocation, α, β, γ, and δ represent the important coefficients of the preset loss function, C represents the number of categories of advertising creative adjustment suggestions, and y creat (i) represents the true label of the creative adjustment suggestion, S creat (i) represents the predicted value of the creative adjustment suggestion, T represents the dimension of the delivery time planning, S time (i) represents the predicted value of the delivery time plan, y time (i) represents the true label of the delivery time plan, S budget represents the predicted value of the delivery budget allocation, t budget Indicates the true label of the delivery budget allocation.

2. The cross-regional advertising analysis and decision-making method based on artificial intelligence according to claim 1 is characterized in that: The processing process of the feature processing layer is as follows: Through the preset linear transformation matrix W att1 , the dimension is q*d, d is the preset intermediate dimension, for the input q-dimensional potential feature vector F out Perform the transformation and obtain the transformed eigenvector H att1 =F out *W att1 ; Calculate the transformed eigenvector H att1 Each element h in i , i=1,2,...,q attention weight S att (h i ), the specific calculation formula is as follows: For the attention weight S att (h i ) performs softmax normalization to obtain the attention weight vector A att1 =[a1,a2,...,a q ],in: The attention weight A att1 =[a1,a2,...,a q ] applied to the input latent feature vector F out On the above, we get the weighted feature representation F att , where the weighted feature representation F att It can be expressed as: F att =[a1*F out *(1),a2*F out (2),...,a q *F out (q)]。 3. The cross-regional advertising analysis and decision-making method based on artificial intelligence according to claim 1 is characterized in that: The processing of the multiple hidden layers is as follows: Based on a weight matrix of h1 neurons with a dimension of q*h1 The first hidden layer of the weighted feature representation F att Processing is performed to obtain the output of the first hidden layer Expressed as: in, represents the output of the kth neuron in the first hidden layer, Indicates the initialization function, F att (i) represents the weighted feature representation F att The i-th element in Represents the weight matrix of the first hidden layer The weight element with index i,k in , represents the bias term of the kth neuron in the first hidden layer; The output of the first hidden layer is used as the input of the second hidden layer, and the output of each hidden layer is used as the input of the next hidden layer in turn, until the output of the last hidden layer is obtained.

4. The cross-regional advertising analysis and decision-making method based on artificial intelligence according to claim 1 is characterized in that: The processing of the output layer is as follows: In the output layer, a probability prediction is made based on the output of the last hidden layer; For the predicted probability S of the advertising creative adjustment suggestion creat The calculation formula is as follows: Among them, h last Indicates the number of neurons in the last hidden layer; O last (j) represents the j-th output of the last hidden layer, W out (j,0) represents the weight matrix W of the last hidden layer out The weight element with index j,0 in b out (0) represents the bias term of the first neuron in the last hidden layer; For the predicted probability distribution S of the delivery channel selection channel The calculation formula is as follows: Among them, W out (j,1) represents the weight matrix W of the last hidden layer out The weight element with index j,1 in b out (1) represents the bias term of the second neuron in the last hidden layer; For the predicted probability distribution S of the delivery time plan time The calculation formula is as follows: Among them, W out (j,2) represents the weight matrix W of the last hidden layer out The weight element with index j,2 in b out (2) represents the bias term of the third neuron in the last hidden layer; For the predicted probability S of budget allocation budget The calculation formula is as follows: Among them, W out (j,3) represents the weight matrix W of the last hidden layer out The weight element with index j,3 in b out (3) represents the bias term of the fourth neuron in the last hidden layer.

5. The cross-regional advertising analysis and decision-making method based on artificial intelligence according to any one of claims 1 to 4, characterized in that: The feature extraction of the regional data matrix based on the preset deep learning network to mine potential features related to the advertising effect includes: Determining, based on the convolution feature operation in the preset deep learning network, the contribution coefficient of each data feature of each data type in the regional data matrix to the advertising delivery effect; Sort the contribution coefficient of each data feature in each data type to the advertising effect, and extract the target data features with the highest contribution coefficient in each data type; The target data features in each data type are classified according to preset types to obtain the regional consumption potential features, the regional cultural preference features, the regional media communication features and the market trend features.

6. An artificial intelligence-based cross-regional advertising analysis and decision-making system, characterized by: Applied to the cross-regional advertising analysis and decision-making method based on artificial intelligence as claimed in any one of claims 1 to 5, the cross-regional advertising analysis and decision-making system based on artificial intelligence comprises: a parsing module, configured to respond to a cross-region advertisement delivery instruction, parse the cross-region advertisement delivery instruction, and determine a target advertisement delivery area in the cross-region advertisement delivery instruction; An acquisition module is used to acquire multi-dimensional data of the target advertising delivery area; the multi-dimensional data includes historical advertising delivery data, regional population data, regional economic data, regional cultural data, regional media data and real-time market dynamics data; A data processing module is used to perform semantic analysis and code conversion on the multi-dimensional data to construct a structured regional data matrix; A feature extraction module is used to extract features from the regional data matrix based on a preset deep learning network to mine potential features related to the effectiveness of advertising; the potential features include regional consumption potential features, regional cultural preference features, regional media communication features, and market trend features; An analysis and decision module is configured to input the potential features into an advertising placement decision model, and obtain an output from the advertising placement decision model of an advertising placement strategy for the target advertising placement area; the advertising placement strategy includes suggestions for adjusting advertising creatives, selection of placement channels, placement time planning, and placement budget allocation; the advertising placement decision model is trained based on sample features and their corresponding advertising placement strategy label results; The advertising placement decision model includes a feature processing layer, an input layer, multiple hidden layers, an output layer, and a strategy prediction layer. The process of outputting an advertising placement strategy for a target advertising placement area includes: Inputting the potential features into the advertising placement decision model, performing feature processing on the potential features based on the feature processing layer combined with the attention mechanism to obtain a weighted feature representation; Inputting the weighted feature representation to the multiple hidden layers based on the input layer, and processing the weighted feature representation based on the multiple hidden layers to obtain an intermediate output; Processing the intermediate output based on the output layer to obtain the predicted probability or predicted probability distribution of the advertising creative adjustment suggestions, delivery channel selection, delivery time planning and delivery budget allocation output by the output layer; Based on the strategy prediction layer, the predicted probability or predicted probability distribution of the advertising creative adjustment suggestions, delivery channel selection, delivery time planning and delivery budget allocation is used to perform strategy prediction to obtain the advertising delivery strategy for the target advertising delivery area; The loss function of the advertising decision model is: loss total =α*loss creat +β*loss channel +γ*loss time +δ*loss budget ; loss budget =|S budget -y budget |; Among them, loss total Represents the loss function of the advertising decision model, loss creat Represents the loss function of the creative adjustment suggestion, loss channel Represents the loss function for channel selection, loss time Represents the loss function of time planning, loss budget represents the loss function of the budget allocation, α, β, γ, and δ represent the important coefficients of the preset loss function, C represents the number of categories of advertising creative adjustment suggestions, and y creat (i) represents the true label of the creative adjustment suggestion, S creat (i) represents the predicted value of the creative adjustment suggestion, T represents the dimension of the delivery time planning, S time (i) represents the predicted value of the delivery time plan, y time (i) represents the true label of the delivery time plan, S budget represents the predicted value of the delivery budget allocation, y budget Indicates the true label of the delivery budget allocation.

7. An electronic device comprising: Memory for storing computer software programs; A processor for reading and executing the computer software program, wherein the computer software program, when executed by the processor, implements the cross-regional advertising analysis and decision-making method based on artificial intelligence as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium storing a computer software program, characterized in that: When the computer software program is executed by a processor, the cross-regional advertising analysis and decision-making method based on artificial intelligence as described in any one of claims 1 to 5 is implemented.

Images