Intelligent advertising content recommendation method and system using prediction model

Abstract

The present application provides an intelligent advertising content recommendation method and system using an application prediction model. First, a historical interaction data set of a target user in an advertising interaction scenario is obtained, which contains multiple advertising interaction sequences composed of user identifiers, etc. Then, cross-dimensional feature fusion processing is performed on the historical interaction data set to generate a combined behavior feature set and an advertising content association feature set. A preset deep prediction model is called to perform time-series dependency modeling on the combined behavior feature set to obtain an advertising content preference prediction result including a click probability distribution and an interaction intention classification label. Then, dynamic weight matching is performed based on this advertising content preference prediction result and the advertising content association feature set to determine a candidate advertising content priority ranking list. Finally, an adapted advertising content set is pushed based on the candidate advertising content priority ranking list, and the deep prediction model parameter configuration is updated according to real-time feedback data to achieve intelligent and accurate advertising recommendation.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a method and system for intelligent advertising content recommendation using a prediction model. Background Art

[0002] In today's digital advertising landscape, with the explosive growth of information and the increasing emphasis on personalized user needs, traditional ad content recommendation methods are struggling to meet the market's high standards. Traditional methods often rely on simple user behavior statistics or rule-based matching strategies, which are unable to cope with complex and changing user behavior patterns.

[0003] Specifically, existing advertising recommendation systems can usually only process features of a single dimension or a limited number of dimensions, and are unable to effectively integrate user behavior features, advertising content features, and the complex relationships between them. This limits the accuracy and personalization of the recommendation system, making advertising content often disconnected from users' actual needs, and reducing the effective reach of advertising and user satisfaction.

[0004] In addition, traditional advertising recommendation methods lack dynamic adjustment and optimization mechanisms. For example, they often adopt fixed recommendation strategies and cannot be flexibly adjusted based on real-time user feedback and dynamic market changes. In the rapidly changing digital marketing environment, this static recommendation model is obviously unable to adapt to the competitive needs of the market. Summary of the Invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present application, an embodiment of the present application provides an intelligent advertising content recommendation method using a prediction model, the method comprising:

[0006] Acquire a historical interaction data set of a target user in an advertising interaction scenario, wherein the historical interaction data set includes multiple advertising interaction sequences, each of which is composed of a user identifier, an advertising content identifier, an interaction behavior type, and an interaction timestamp;

[0007] Performing cross-dimensional feature fusion processing on the historical interaction data set to generate a combined behavior feature set and an advertisement content-related feature set for each advertisement interaction sequence;

[0008] Calling a preset deep prediction model to perform temporal dependency modeling on the combined behavioral feature set to generate an advertising content preference prediction result, wherein the advertising content preference prediction result includes a user's click probability distribution for candidate advertising content and an interaction intention classification label;

[0009] performing dynamic weight matching based on the advertising content preference prediction result and the advertising content associated feature set to determine a priority ranking list of candidate advertising contents;

[0010] Based on the priority sorting list, a set of adapted advertising content is pushed to target users, and parameter configuration of the depth prediction model is updated according to real-time feedback data.

[0011] On the other hand, an embodiment of the present application also provides an advertising service system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to run the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0012] Based on the above aspects, the embodiment of the present application not only captures the temporal characteristics of user behavior by integrating the multi-dimensional historical interaction data of the target user in the advertising interaction scenario, but also deeply associates the user behavior pattern with the advertising content characteristics through cross-dimensional feature fusion processing, forming a comprehensive and detailed combination behavior feature set and advertising content association feature set. The introduction of the deep prediction model further explores the temporal dependency relationship in the user behavior sequence, effectively predicts the user's click probability distribution and interaction intention classification label for the candidate advertising content, and greatly improves the accuracy and personalization level of advertising content recommendation. On this basis, the dynamic weight matching mechanism intelligently determines the priority ranking list of candidate advertising content based on the deep fusion of the advertising content preference prediction results and the advertising content association feature set, ensuring the timeliness and relevance of advertising push. Finally, based on the priority ranking list, the adapted advertising content set is pushed to the target user, and the deep prediction model is continuously optimized based on real-time feedback data, forming a self-iterative and continuously evolving intelligent recommendation system, which not only significantly improves the click-through rate and conversion rate of advertising recommendations, but also enhances the satisfaction and loyalty of the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a schematic diagram of the execution flow of the intelligent advertising content recommendation method using the prediction model provided in an embodiment of the present application.

[0014] Figure 2 This is a schematic diagram of the hardware architecture of the advertising service system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0015] The present application will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of an intelligent advertising content recommendation method using an application prediction model provided by an embodiment of the present application. The intelligent advertising content recommendation method using the application prediction model is introduced in detail below.

[0016] Examples section:

[0017] Step S110: Acquire a historical interaction data set of the target user in an advertising interaction scenario, wherein the historical interaction data set includes multiple advertising interaction sequences, and each advertising interaction sequence consists of a user identifier, an advertising content identifier, an interaction behavior type, and an interaction timestamp.

[0018] In this embodiment, to implement intelligent advertising content recommendation, using an online shopping platform's advertising interaction scenario as an example, the platform's server can continuously record a large amount of user-advertising interaction information. Assume that the online shopping platform has multiple users, such as User A and User B. Each user sees and interacts with different advertisements while browsing product pages or searching for products. For example, if User A interacts with multiple advertisements within a week, the server will record these interactions to form an ad interaction sequence. Each ad interaction sequence includes a user identifier, such as User A's unique identification number "12345"; an ad content identifier, such as the number "AD001" for a mobile phone ad; the type of interaction, such as clicking on the ad, adding an item in the ad to a shopping cart, or purchasing the advertised item; and an interaction timestamp. For example, if User A clicks on an ad numbered "AD001" at 10:30 AM on July 1, 2024, the time "2024-07-01 10:30:00" is the interaction timestamp. The server integrates the above-mentioned advertising interaction sequences collected from all users to form a historical interaction data set. The data in the historical interaction data set is the basis for subsequent advertising recommendations. By analyzing the above data, the user's advertising interaction habits and preferences can be understood.

[0019] Step S120: performing cross-dimensional feature fusion processing on the historical interaction data set to generate a combined behavior feature set and an advertisement content-related feature set for each advertisement interaction sequence.

[0020] In this embodiment, after obtaining the historical interaction data set, it is necessary to perform cross-dimensional feature fusion processing on it to generate a useful feature set. Continuing with the above example of the online shopping platform, each advertising interaction sequence in the historical interaction data set is deeply analyzed and processed.

[0021] Step S121: performing time window segmentation processing on the interactive behavior types in the advertisement interactive sequence to generate a plurality of behavior segment units, each behavior segment unit including continuous interactive behaviors within a preset time length.

[0022] In this embodiment, taking the advertising interaction sequence of user A as an example, in order to better analyze the user's behavior pattern, the interaction behavior type is segmented into time windows. Assuming that the preset duration is 1 day, the advertising interaction behavior of user A within a week is divided into daily periods. For example, on July 1, 2024, user A clicked on the mobile phone advertisement "AD001" and viewed the details page of the computer advertisement "AD002". The above continuous interaction behaviors within one day constitute a behavior segment unit. Similarly, the interaction behaviors on July 2, July 3, 2024, etc. each constitute a different behavior segment unit. In this way, through time window segmentation, the user's advertising interaction behavior is divided into multiple segments with temporal continuity, which facilitates the subsequent extraction and analysis of the features of each segment.

[0023] Step S122: extracting the behavior density feature and interaction intensity feature of each behavior segment unit, wherein the behavior density feature is determined by the interaction frequency per unit time, and the interaction intensity feature is determined by the weighted value of the interaction duration and the operation depth.

[0024] In this embodiment, behavior density and interaction intensity features are further extracted for each behavior segment. Taking the behavior segment on July 1, 2024, as an example, the behavior density feature can be determined by calculating the interaction frequency per unit time (here, one day). If user A clicked on an ad three times and viewed the ad details page twice during that day, the interaction frequency for that day would be five, which serves as an important reference value for the behavior density feature for that behavior segment. The interaction intensity feature, on the other hand, comprehensively considers interaction duration and engagement depth. Suppose user A clicks on the mobile ad "AD001," stays on the ad page for three minutes, and performs actions such as viewing product specifications and comparing different styles. These actions can be quantified as engagement depth. Assuming a weight of 0.6 for interaction duration and 0.4 for engagement depth, interaction duration and engagement depth are weighted according to these weights to produce the interaction intensity feature for that behavior segment. For example, if the interaction duration is quantified as 3 points (out of 10 points) and the operation depth is quantified as 6 points (out of 10 points), then the interaction intensity feature = 3×0.6+6×0.4=4.2 points.

[0025] Step S123: performing semantic parsing on the advertisement material corresponding to the advertisement content identifier, and extracting semantic topic distribution features of the advertisement text and attention weight features of the visual elements.

[0026] In this embodiment, semantic parsing is performed on the advertising material corresponding to the advertising content identifier in each advertising interaction sequence. Taking the advertisement "AD001" as an example, this advertising material may include advertising text and visual elements. Semantic parsing of the advertising text is performed. For example, if the advertisement text is "New smartphone, superb performance, clear photos, affordable price," natural language processing techniques are used to analyze the semantic topic distribution characteristics of the advertisement text. Semantic topics can be categorized into "performance," "photography," "price," and so on, and the proportion of each topic in the text is calculated. Assuming that the "performance" topic accounts for 0.3, the "photography" topic accounts for 0.4, and the "price" topic accounts for 0.3, these percentage values ​​constitute the semantic topic distribution characteristics of the advertisement text. For the visual elements of the advertisement, such as images and videos, their attention weight characteristics are analyzed using computer vision techniques. For example, if the phone's camera is prominently displayed in the advertisement image, analysis shows that the camera element has an attention weight of 0.7, while the total attention weight of the other elements is 0.3. These attention weight values ​​represent the attention weight characteristics of the visual elements.

[0027] Step S124: performing feature splicing on the behavior density features, interaction intensity features, semantic topic distribution features, and attention weight features corresponding to the same advertising interaction sequence to generate the combined behavior feature set.

[0028] In this embodiment, taking the advertisement interaction sequence between user A and the advertisement "AD001" as an example, the behavioral density features, interaction intensity features, semantic topic distribution features and attention weight features extracted above are spliced. Assume that the behavioral density feature is 5 times / day, the interaction intensity feature is 4.2 points, the semantic topic distribution feature is (the "performance" theme accounts for 0.3, the "photographing function" theme accounts for 0.4, and the "price" theme accounts for 0.3), and the attention weight feature is (the camera element attention weight is 0.7, and the other elements attention weight is 0.3). The above features are spliced ​​in a certain order, for example, the behavioral density features and the interaction intensity features are spliced ​​first, and then the semantic topic distribution features and the attention weight features are spliced ​​to obtain a multi-dimensional combined behavioral feature set. This combined behavioral feature set contains information on multiple aspects of the user's interaction with the advertisement, and can more comprehensively reflect the user's interactive behavior characteristics for the advertisement.

[0029] Step S125: calling the advertising knowledge base to match the product category tree corresponding to the advertising content identifier, extracting the hierarchical association features of the category path and the user's historical preference matching degree, and generating the advertising content association feature set.

[0030] In this example, using the advertisement "AD001" as an example, the advertisement knowledge base is used to match the product category tree corresponding to the advertisement content identifier. The advertisement knowledge base stores product category information corresponding to various advertisements. Through matching, the product category tree corresponding to the mobile phone advertisement "AD001" can be found. This product category tree may include categories at different levels, such as "electronic products," "mobile phones," and "smartphones."

[0031] Step S1251: parse the parent-child node relationship of the product category tree to generate category graph structure data including a hierarchical connection relationship.

[0032] In this example, the parent-child node relationships of the product category tree are parsed. For example, "electronic products" is the parent node of "mobile phone," and "mobile phone" is the parent node of "smartphone." These node relationships are graphically represented to form category graph structure data. In this category graph structure data, nodes represent categories, and edges represent parent-child relationships. This method clearly demonstrates the hierarchical connection relationships between categories, providing a foundation for subsequent extraction of hierarchical association features.

[0033] Step S1252: Input the category graph structure data into the graph neural network, and generate a hierarchical perception embedding vector for each category node through feature propagation of adjacent nodes.

[0034] In this embodiment, the generated category graph structure data is input into the graph neural network. The graph neural network can learn the characteristics of each category node by propagating the features of adjacent nodes. For example, the "smartphone" node will obtain feature information from its parent node "mobile phone" and its other adjacent "smartphone" subcategory nodes. After calculation by the graph neural network, a hierarchical-aware embedding vector is generated for each category node. The above-mentioned hierarchical-aware embedding vector contains the hierarchical information of the category node in the category tree and the association information with other nodes, which can better represent the characteristics of the category node.

[0035] Step S1253: Count the historical exposure times and click-through conversion rate of each category node in the target user's historical interaction data set to generate a user category preference distribution vector.

[0036] In this embodiment, the historical exposure count and click-through conversion rate of each category node in the historical interaction data set of the target user (such as User A) are calculated. For example, in User A's historical interaction data, the "Electronic Products" category was exposed 10 times, with 3 clicks and a click-through conversion rate of 30%; the "Mobile Phones" category was exposed 8 times, with 2 clicks and a click-through conversion rate of 25%; and the "Smartphones" category was exposed 6 times, with 2 clicks and a click-through conversion rate of 33.3%. The above historical exposure counts and click-through conversion rates are sorted according to the order of the category nodes to generate a user category preference distribution vector, which reflects the user's preference for different category nodes.

[0037] Step S1254: performing a matrix product operation on the hierarchical perception embedding vector and the user category preference distribution vector to generate a category node matching weight.

[0038] In this embodiment, the previously generated hierarchical perception embedding vector and the user category preference distribution vector are subjected to a matrix product operation. Assume that the hierarchical perception embedding vector is a 3-dimensional vector (corresponding to the three category nodes of "electronic products", "mobile phones" and "smart phones"), and the user category preference distribution vector is also a 3-dimensional vector. Multiply the corresponding elements of these two vectors to obtain the matching weight of each category node. For example, the hierarchical perception embedding vector element of the "electronic products" category is 0.2, and the user category preference distribution vector element is 0.3, then the matching weight of the category is 0.2×0.3=0.06. In this way, the hierarchical information of the category nodes and the user's preference information are comprehensively considered to generate a more accurate category node matching weight.

[0039] Step S1255: concatenate the hierarchical-aware embedding vector and the category node matching weight to generate a dimension-aligned feature matrix of the advertising content-related feature set.

[0040] In this embodiment, the hierarchical perception embedding vector and the category node matching weight are spliced ​​together. For example, the hierarchical perception embedding vector is (0.2, 0.3, 0.4), and the category node matching weight is (0.06, 0.08, 0.1). They are spliced ​​together in order to obtain a new vector (0.2, 0.3, 0.4, 0.06, 0.08, 0.1). This vector is part of the dimension-aligned feature matrix of the advertising content-related feature set. All relevant hierarchical perception embedding vectors and category node matching weights are spliced ​​together in this way, and finally a dimension-aligned feature matrix of the advertising content-related feature set is generated. This dimension-aligned feature matrix contains the association information between the advertising content and the product category tree and the user's preference information for the above categories.

[0041] Step S130: calling a preset deep prediction model to perform temporal dependency modeling on the combined behavioral feature set to generate an advertising content preference prediction result, wherein the advertising content preference prediction result includes the user's click probability distribution for the candidate advertising content and the interaction intention classification label.

[0042] In this embodiment, after generating a combined behavioral feature set, a preset deep prediction model is called to perform temporal dependency modeling on it to predict the user's advertising content preference. Continuing with the online shopping platform as an example, the combined behavioral feature set of user A is input into the deep prediction model.

[0043] Step S131: inputting the combined behavior feature set into the temporal convolutional network of the depth prediction model to extract local patterns and generate primary temporal features of each behavior segment unit.

[0044] In this embodiment, the temporal convolutional network of the deep prediction model is used to extract local patterns in the combined behavior feature set. Taking the combined behavior feature set of user A as an example, the combined behavior feature set contains the features of multiple behavior fragment units. The temporal convolutional network processes the above features through convolution operations. For example, the convolution kernel slides on the combined behavior feature set and performs convolution calculations on the features of each behavior fragment unit. Assuming that the combined behavior feature set is a multidimensional vector, the convolution kernel size of the temporal convolutional network is 3×3, and the convolution kernel slides on the vector, selecting 3 consecutive feature elements each time for convolution operations. After the convolution operation, the primary temporal features of each behavior fragment unit are generated, and the primary temporal features reflect the local temporal patterns within each behavior fragment unit.

[0045] Step S132: Execute a multi-head self-attention mechanism on the primary temporal features to capture the long-range dependencies between different behavior segment units and generate enhanced temporal features.

[0046] In this embodiment, a multi-head self-attention mechanism is performed on the generated primary time series features. The multi-head self-attention mechanism allows the model to pay attention to the long-range dependencies between different behavior fragment units. Taking the primary time series features of user A as an example, the feature contains feature information of multiple behavior fragment units. The multi-head self-attention mechanism divides the primary time series features into multiple heads, and each head independently calculates the attention score. For example, assuming there are 3 heads, each head performs a different linear transformation on the primary time series features, and then calculates the attention score between each behavior fragment unit and other behavior fragment units. The attention score represents the degree of association between different behavior fragment units. The attention score is combined with the primary time series features by weighted summation to obtain enhanced time series features. The enhanced time series features can better reflect the long-range dependencies between different behavior fragment units and improve the model's ability to understand user behavior patterns.

[0047] Step S133: Input the enhanced temporal features into the gated recurrent unit network of the deep prediction model for state update to generate a dynamic hidden state sequence of user behavior evolution.

[0048] In this embodiment, the enhanced temporal features are input into the gated recurrent unit network of the deep prediction model. The gated recurrent unit network can perform state updates based on the input enhanced temporal features to generate a dynamic hidden state sequence of the user behavior evolution. Taking the enhanced temporal features of user A as an example, the gated recurrent unit network includes an input gate, a forget gate, and an output gate. The input gate controls the extent to which new input information enters the unit, the forget gate controls the extent to which old information is retained in the unit, and the output gate controls the extent to which the unit outputs information. At each time step, the gated recurrent unit network updates the current hidden state based on the current enhanced temporal features and the hidden state of the previous time step through calculations of the input gate, forget gate, and output gate. After calculations for multiple time steps, a dynamic hidden state sequence of the user behavior evolution is generated, which records the changes in the user's behavioral state at different time points.

[0049] Step S134: performing maximum pooling processing on the dynamic hidden state sequence to extract a global behavior representation vector, and performing feature cross calculation on the global behavior representation vector and the advertising content-related feature set to obtain a feature cross result.

[0050] In this embodiment, maximum pooling is performed on the dynamic hidden state sequence to extract a global behavior representation vector.

[0051] Step S1341: Divide the dynamic hidden state sequence into sliding windows along the time dimension to generate multiple local state subsequences.

[0052] In this embodiment, using user A's dynamic hidden state sequence as an example, a sliding window partitioning is performed along the time dimension. Assuming the sliding window size is 3 time steps, and the window slides 1 time step at a time, the dynamic hidden state sequence is then divided into multiple local state subsequences. For example, if the dynamic hidden state sequence contains hidden states for 10 time steps, the sliding window partitioning will result in 8 local state subsequences. Each local state subsequence contains hidden states for 3 consecutive time steps. These local state subsequences provide the basis for the subsequent max pooling operation.

[0053] Step S1342: Perform element-by-element maximum selection on the hidden state vector in each local state subsequence to generate local pooling features.

[0054] In this embodiment, for each local state subsequence, the hidden state vector therein is subjected to element-by-element maximum selection. For example, a local state subsequence contains hidden state vectors of three time steps, each of which is a multidimensional vector. The corresponding elements of these three vectors are compared, and the maximum value is selected as the corresponding element of the local pooling feature. Assuming that the dimension of the hidden state vector is 5-dimensional, then after element-by-element maximum selection, a 5-dimensional local pooling feature vector is obtained. This operation is performed on each local state subsequence to generate multiple local pooling features.

[0055] Step S1343: All local pooling features are concatenated and compressed to a preset dimension through a fully connected layer to generate an intermediate aggregation vector.

[0056] In this embodiment, all the generated local pooling features are connected in series. Assume that 8 local pooling feature vectors are generated, each with a dimension of 5. They are connected in sequence to obtain a 40-dimensional vector, which is then input into a fully connected layer. The function of the fully connected layer is to compress the high-dimensional vector to a preset dimension. For example, the preset dimension is 10 dimensions. The fully connected layer compresses the 40-dimensional vector into a 10-dimensional intermediate aggregate vector through linear transformation and activation function calculation. The intermediate aggregate vector contains the global information of the dynamic hidden state sequence.

[0057] Step S1344: Perform a residual connection on the intermediate aggregation vector and the terminal temporal features output by the temporal convolutional network, and perform layer normalization on the residual connection result to generate the final global behavior representation vector for downstream prediction tasks.

[0058] In this embodiment, the intermediate aggregation vector is residually connected with the terminal time series features output by the temporal convolutional network. Residual connection can help the model better learn feature information at different levels. Assuming that the dimension of the intermediate aggregation vector is 10 dimensions and the dimension of the terminal time series features output by the temporal convolutional network is also 10 dimensions, their corresponding elements are added together to obtain a new vector. The new vector is then layer-normalized, and layer normalization can make the element distribution of the vector more stable. After layer normalization, the final global behavior representation vector is generated. The final global behavior representation vector contains the global feature information of the user behavior and is used for subsequent downstream prediction tasks.

[0059] Next, a feature cross-calculation is performed on the global behavior representation vector and the feature set associated with the ad content. Taking the global behavior representation vector of user A and the feature set associated with the ad "AD001" as an example, feature cross-calculation can be achieved by multiplying the corresponding elements of the two vectors or performing other nonlinear transformations. Assuming the global behavior representation vector is (0.2, 0.3, 0.4) and the feature set associated with the ad content is (0.1, 0.2, 0.3), multiplying their corresponding elements yields a feature cross-calculation result of (0.02, 0.06, 0.12). This feature cross-calculation combines the global characteristics of user behavior with the associated characteristics of the ad content, providing richer information for subsequent predictions.

[0060] Step S135: Perform a nonlinear transformation on the feature cross-reaction result through the multi-layer perceptron network of the deep prediction model, and output the advertising content preference prediction result, wherein the click probability distribution is normalized by the Sigmoid function, and the interaction intention classification label is generated by the Softmax function.

[0061] In this embodiment, after the feature crossover result is obtained, it is input into the multilayer perceptron network of the deep prediction model for nonlinear transformation. The multilayer perceptron network consists of multiple fully connected layers, each of which contains multiple neurons, and the neurons are connected by weights. Taking the feature crossover result (0.02, 0.06, 0.12) of user A as an example, it is input into the input layer of the multilayer perceptron network. The neurons in the input layer receive the above feature values ​​and pass the results to the next layer through weighted summation and activation function calculation. Assume that the input layer has 3 neurons, corresponding to the 3 elements of the feature crossover result, the middle layer has 5 neurons, and the output layer has 2 neurons, corresponding to the click probability and interaction intent classification, respectively. The weighted summation calculation of each neuron can be expressed as multiplying the input value by the corresponding weight, then adding all the products, and adding a bias term. For example, the first neuron in the middle layer receives the outputs of three neurons in the input layer. Assuming the corresponding weights are 0.1, 0.2, and 0.3, respectively, and the bias is 0.05, then the input value of this neuron is 0.02 × 0.1 + 0.06 × 0.2 + 0.12 × 0.3 + 0.05 = 0.05 + 0.002 + 0.012 + 0.036 = 0.099. After processing by an activation function (such as the ReLU function), the neuron outputs a new value. Through layer-by-layer calculations in the multilayer perceptron network, the original prediction result is finally obtained at the output layer.

[0062] The click probability distribution is normalized using the Sigmoid function. The Sigmoid function maps the output value to the interval [0, 1], representing the probability of a click. Assuming the output value of the neuron corresponding to the click probability in the output layer is 0.8, after the Sigmoid function calculation, the click probability is 1 / (1 + e^(-0.8)) ≈ 0.69. The interaction intent classification label is generated using the Softmax function. Assuming the output values ​​of the neurons corresponding to the interaction intent classification in the output layer are 0.2 and 0.3, respectively, the Softmax function converts these values ​​into a probability distribution. First, the exponent values ​​are calculated: e^0.2 ≈ 1.22 and e^0.3 ≈ 1.35. Then, the probabilities are calculated: the probability of the first classification is 1.22 / (1.22 + 1.35) ≈ 0.47, and the probability of the second classification is 1.35 / (1.22 + 1.35) ≈ 0.53. The classification with the highest probability is the predicted interaction intent classification label. In this way, we obtain the advertising content preference prediction results including the user's click probability distribution for candidate advertising content and the interaction intention classification label.

[0063] Step S140: performing dynamic weight matching based on the advertising content preference prediction result and the advertising content associated feature set to determine a priority ranking list of candidate advertising contents.

[0064] In this embodiment, after obtaining the ad content preference prediction results and the ad content association feature set, dynamic weight matching is performed to determine a prioritized list of candidate ad content. Continuing with the example of an online shopping platform, the platform's ad content library contains a large number of ads, from which it is necessary to screen and prioritize ads suitable for user A.

[0065] Step S141: Filtering a candidate advertising content subset that matches the region and terminal type of the target user from the advertising content library.

[0066] In this embodiment, it is assumed that user A belongs to a certain urban area and uses a mobile phone as the terminal type. The advertising content library stores the geographical scope and terminal type information applicable to each advertisement. Therefore, this embodiment can filter out advertisements that are applicable to the region where user A is located and support mobile terminals from the advertising content library based on the above information. For example, there are 100 advertisements in the advertising content library, of which only 30 are applicable to the region where user A is located, and 20 of these 30 advertisements support mobile terminals, then these 20 advertisements constitute the candidate advertising content subset. The purpose of such screening is to ensure that the recommended advertisements match the actual situation of the user and to improve the relevance and effectiveness of the advertisements.

[0067] Step S142: Calculate the cosine similarity between each candidate advertisement content and the global behavior representation vector to generate a content relevance score.

[0068] In this embodiment, in order to measure the relevance between each candidate advertisement content and the user's global behavior representation vector, the cosine similarity between them is calculated.

[0069] Step S1421: extracting the embedding vector of the candidate advertising content from the advertising content library, wherein the embedding vector is generated by jointly encoding the advertising text and visual features using a pre-trained twin-tower model.

[0070] In this embodiment, for each candidate ad content, its embedding vector is extracted from the ad content library. A pre-trained dual-tower model encodes the ad text and visual features separately. For example, for the candidate ad "AD002," its ad text is "Fashionable sports shoes, comfortable and breathable," and its visual feature is an image of sneakers. One tower of the dual-tower model processes the ad text, converting it into a vector representation through operations such as word embedding and convolutional neural networks. The other tower processes the visual features, extracting image features through a convolutional neural network, also converting them into a vector representation. These two vectors are then concatenated or otherwise fused to generate the embedding vector for the ad "AD002." Assume that this embedding vector is (0.1, 0.2, 0.3).

[0071] Step S1422: performing L2 normalization processing on the global behavior representation vector, and calculating the dot product between the normalized global behavior representation vector and each candidate advertising content embedding vector to generate an original similarity score.

[0072] In this embodiment, L2 normalization is performed on user A's global behavior representation vector. The purpose of L2 normalization is to normalize the length of the vector to 1, so that the vector only represents the direction. Assuming the global behavior representation vector is (0.2, 0.3, 0.4), its L2 norm = √(0.2^2+0.3^2+0.4^2)=√(0.04+0.09+0.16)=√0.29≈0.54. The normalized global behavior representation vector is (0.2 / 0.54, 0.3 / 0.54, 0.4 / 0.54)≈(0.37, 0.56, 0.74). Then, the dot product between the normalized global behavior representation vector and the embedding vector (0.1, 0.2, 0.3) of the candidate ad “AD002” is calculated. The original similarity score = 0.37×0.1+0.56×0.2+0.74×0.3=0.037+0.112+0.222=0.371.

[0073] Step S1423: Perform piecewise linear transformation on the original similarity scores, and map the score interval to the range of [0, 1] to generate a content relevance score.

[0074] In this embodiment, to convert the original similarity score into a more intuitive score, a piecewise linear transformation is performed. Assuming the original similarity score ranges from [-1, 1], this embodiment aims to map it to the range [0, 1]. This transformation can be performed using a linear function, for example, content relevance score = (original similarity score + 1) / 2. For an original similarity score of 0.371, the content relevance score = (0.371 + 1) / 2 = 0.6855.

[0075] Step S1424: De-weighting the content relevance score based on the historical exposure frequency of the candidate advertising content.

[0076] In this embodiment, considering that the historical exposure frequency of candidate ad content may reduce its appeal to users, the content relevance score is downgraded. Assuming that the candidate ad "AD002" has a historical exposure frequency of 5 times, according to the preset downgrading rules, the higher the exposure frequency, the greater the downgrading coefficient. Assuming the downgrading coefficient is 0.8, the downgraded content relevance score = 0.6855 × 0.8 = 0.5484. This downgrading process can avoid over-recommendation of ads that have been exposed multiple times and improve the diversity of ad recommendations.

[0077] Step S143: Matching the category path hierarchical association features in the advertisement content association feature set according to the interaction intention classification label to generate a category suitability score.

[0078] In this embodiment, the category path hierarchical association features in the advertising content association feature set are matched based on the interaction intention classification label in the advertising content preference prediction result. Assuming that the interaction intention classification label is "buy a mobile phone", for the candidate advertisement "AD001" (mobile phone advertisement), its advertising content association feature set contains category path hierarchical association features, such as "electronic products-mobile phones-smartphones". Then, a category adaptability score can be generated based on the degree of matching between the interaction intention classification label and the category path hierarchical association features. If the matching degree is high, for example, the interaction intention classification label completely matches a node in the category path, then the category adaptability score is high; if the matching degree is low, the score is low. Assuming the matching degree is 80%, the category adaptability score can be set to 0.8.

[0079] Step S144: Dynamically weight the click probability distribution, content relevance score, and category suitability score to generate a comprehensive recommendation score, wherein the weight coefficient of the dynamic weighted summation is dynamically adjusted according to the click rate in the real-time feedback data.

[0080] In this embodiment, the click probability distribution, content relevance score and category suitability score are dynamically weighted and summed. Assume that the click probability distribution is 0.69, the content relevance score (after de-weighting) is 0.5484, and the category suitability score is 0.8. The initial weight coefficient can be set based on experience. Assume that the weight of the click probability distribution is 0.4, the weight of the content relevance score is 0.3, and the weight of the category suitability score is 0.3. Then the comprehensive recommendation score = 0.69×0.4+0.5484×0.3+0.8×0.3=0.276+0.16452+0.24=0.68052. As the real-time feedback data is updated, the system will dynamically adjust the weight coefficient according to the click-through rate. For example, if it is found that the click-through rate is more correlated with the content relevance score, the weight of the content relevance score will be appropriately increased. Assuming that after adjustment, the weight of the click probability distribution becomes 0.3, the weight of the content relevance score becomes 0.4, and the weight of the category suitability score becomes 0.3, then the new comprehensive recommendation score = 0.69×0.3+0.5484×0.4+0.8×0.3=0.207+0.21936+0.24=0.66636.

[0081] Step S145: Arrange the candidate advertising content subsets in descending order according to the comprehensive recommendation scores, generate the priority ranking list, and filter out advertising content with scores lower than a preset threshold.

[0082] In this embodiment, the comprehensive recommendation score of each advertisement in the candidate advertising content subset is sorted in descending order. Assume that the candidate advertising content subset contains 5 advertisements, and their comprehensive recommendation scores are 0.68052, 0.66636, 0.5, 0.3, and 0.2, respectively. After sorting in descending order, they are 0.68052, 0.66636, 0.5, 0.3, and 0.2. The preset threshold can be set according to actual conditions. Assuming the preset threshold is 0.4, then advertisements with scores below 0.4 are filtered out, and the remaining advertisements "AD001" (score 0.68052), "AD002" (score 0.66636), and "AD003" (score 0.5) form a priority ranking list. The advertisements in this priority ranking list are sorted from high to low according to the comprehensive recommendation score, and the higher the score, the more preferentially recommended to the user.

[0083] Step S150: Pushing an adapted advertising content set to target users based on the priority ranking list, and updating parameter configuration of the depth prediction model according to real-time feedback data.

[0084] In this embodiment, after obtaining the priority sorting list, an adapted advertising content set is pushed to the target user, and the parameter configuration of the deep prediction model is updated according to the real-time feedback data to improve the accuracy and effect of the advertising recommendation.

[0085] Step S151: Select the top N candidate advertisement contents from the priority sorting list to generate an initial push set, wherein N is dynamically set according to the screen size of the user terminal.

[0086] In this embodiment, the number of selected ads, N, is dynamically set based on the user's terminal screen size. Assuming User A uses a smaller phone screen, the system determines N to be 3 based on pre-set rules. The top three candidate ads from the priority list—"AD001" (score 0.68052), "AD002" (score 0.66636), and "AD003" (score 0.5)—are selected to form the initial push set. These ads are selected based on the comprehensive recommendation scores and are most likely to meet the user's preferences and needs.

[0087] Step S152: Monitor the user's real-time interactive behavior on the advertising content in the initial push set, record click events, exposure duration, and secondary forwarding operations to generate a real-time feedback data set.

[0088] In this embodiment, after pushing ads from the initial push set to user A, the system monitors the user's real-time interactive behavior. For example, after seeing ad "AD001," user A clicks on it, stays on the ad page for 2 minutes, and shares the ad with a friend. This interactive behavior is recorded, with the click event recorded as "clicked AD001," the exposure duration recorded as 2 minutes, and the secondary forwarding action recorded as "shared AD001 with a friend." User interactions are also recorded for ads "AD002" and "AD003." These records are integrated to generate a real-time feedback dataset containing the user's actual response to the pushed ads.

[0089] Step S153: Incrementally merge the real-time feedback data set and the historical interaction data set, reconstruct the training sample set and re-divide the validation set and test set.

[0090] In this embodiment, the generated real-time feedback data set is incrementally merged with the previously acquired historical interaction data set. Assuming that the historical interaction data set contains 1000 records and the real-time feedback data set contains 10 records, these 10 records are added to the historical interaction data set to obtain a new data set containing 1010 records. The training sample set is then reconstructed for the new data set, and the validation set and test set are repartitioned. For example, it can be divided in a ratio of 80%, 10%, and 10%, that is, 808 records are used as the training sample set, 101 records are used as the validation set, and 101 records are used as the test set. In this way, the training sample set is updated with the latest user interaction data, so that the model can learn the user's latest behavior patterns.

[0091] Step S154: Using a sliding window mechanism to fine-tune the parameters of the depth prediction model online, wherein the model loss function of the depth prediction model integrates the click-through rate prediction error and the intent classification cross entropy loss.

[0092] In this embodiment, a sliding window mechanism is used to perform online parameter fine-tuning on the depth prediction model.

[0093] Step S1541: setting a sliding window interval of a fixed time window length, and extracting the most recently generated interaction record within the sliding window interval from the real-time feedback dataset as the current fine-tuning sample.

[0094] In this example, the sliding window length is set to one day. Interaction records generated within the last day are extracted from the real-time feedback dataset as the current fine-tuning sample. Assuming the real-time feedback dataset records user interactions over a week, only the interaction records from the last day are extracted as the current fine-tuning sample. If there are five interaction records from the last day, these five records constitute the current fine-tuning sample.

[0095] Step S1542: input the current fine-tuning sample into the last two fully connected layers of the depth prediction model, unfreeze the weight parameters of the last two fully connected layers and freeze the parameters of the remaining layers.

[0096] In this embodiment, the current fine-tuning sample is input into the last two fully-connected layers of the depth prediction model. During fine-tuning, the weight parameters of the last two fully-connected layers are unfrozen so that they can be updated based on the current fine-tuning sample, while the parameters of the remaining layers are frozen to avoid excessive impact on the overall structure and parameters of the model. For example, a depth prediction model includes an input layer, an intermediate layer, and an output layer, and the intermediate layer has multiple fully-connected layers. Only the weight parameters of the last two fully-connected layers are unfrozen, while the parameters of the remaining layers remain unchanged.

[0097] Step S1543: Calculate the gradient direction according to the click-through rate prediction error and the intent classification cross entropy loss of the current fine-tuning sample, and use the adaptive momentum optimization algorithm to update the weight parameters of the last two fully connected layers along the gradient direction.

[0098] In this embodiment, the gradient direction is calculated based on the click-through rate prediction error of the current fine-tuning sample and the intent classification cross entropy loss. The click-through rate prediction error can be obtained by calculating the difference between the predicted click probability and the actual click situation. For example, if the predicted click probability is 0.6, but the actual user does not click, then the click-through rate prediction error is large. The intent classification cross entropy loss can measure the difference between the predicted interaction intent classification label and the actual interaction intention. The gradient direction is calculated by the above loss function, and the gradient direction indicates the direction in which the weight parameters need to be updated. An adaptive momentum optimization algorithm (such as the Adam algorithm) is used to update the weight parameters of the last two fully connected layers along the gradient direction. The Adam algorithm can dynamically adjust the learning rate based on the first-order moment estimation and second-order moment estimation of the gradient, making the update of the weight parameters more stable and efficient.

[0099] Step S1544: performing exponential moving average processing on the updated weight parameters of the last two fully connected layers and the corresponding historical weight parameters before the update to generate smoothed new weight parameters.

[0100] In this embodiment, in order to make the update of the weight parameters smoother, the weight parameters of the last two fully connected layers after the update are subjected to exponential moving average processing with the corresponding historical weight parameters before the update. Assume that the historical weight parameters before the update are W_old, the weight parameters after the update are W_new, and the decay coefficient of the exponential moving average is 0.9. Then the new smoothed weight parameter W_smooth = 0.9 × W_old + 0.1 × W_new. In this way, the stability of the historical weight parameters is taken into account, and new update information is introduced, making the update of the weight parameters more reasonable.

[0101] Step S1545: Load the smoothed new weight parameters into the online instance of the depth prediction model, and add the current fine-tuning sample to the validation set to calculate the click-through rate prediction accuracy.

[0102] In this embodiment, the smoothed new weight parameters are loaded into the online instance of the deep prediction model. The current fine-tuned sample is then added to the validation set, and the validation set is used to calculate the click-through rate prediction accuracy. For example, if there are 100 records in the validation set and the model predicts that users will click on 60 of them, but users actually click on 55 of them, then the click-through rate prediction accuracy = 55 / 60≈0.917 (rounded to three decimal places). By calculating the click-through rate prediction accuracy, we can evaluate the performance of the model on new data and understand whether the model's prediction effect has improved.

[0103] Step S1546: When the click rate prediction accuracy is higher than a preset threshold of the current online version, a hot update operation is triggered to synchronize the smoothed new weight parameters to the online advertising push service.

[0104] In this embodiment, a threshold for the click-through rate prediction accuracy is pre-set, for example, set to 0.9. The currently calculated click-through rate prediction accuracy is compared with the threshold and the accuracy of the current online version. If the current click-through rate prediction accuracy is higher than the preset threshold and better than the accuracy of the current online version, it means that the online parameter fine-tuning through the sliding window mechanism has achieved good results and the performance of the model has been improved. At this time, the hot update operation is triggered to synchronize the smoothed new weight parameters to the online advertising push service. In this way, the online advertising recommendation system can use the updated model parameters to recommend advertisements, thereby providing users with advertising content that is more in line with their preferences, improving the click-through rate of advertisements and the user's interactive experience. For example, after updating the weight parameters, the online advertising push service may make the originally less accurate advertising recommendations more in line with user needs, and users are more likely to click on the recommended ads, thereby improving the conversion rate of ads and the revenue of the platform.

[0105] Step S155: When the average click rate fluctuation in the real-time feedback data set exceeds a preset tolerance threshold, the model structure optimization operation is triggered to adjust the convolution kernel size of the temporal convolutional network and the hidden layer dimension of the gated recurrent unit network of the depth prediction model.

[0106] In this embodiment, the fluctuation of the average click-through rate in the real-time feedback data set is continuously monitored. A tolerance threshold is preset, for example, set to 10%. By calculating the average click-through rate of different time periods in the real-time feedback data set, and comparing the change in the average click-through rate of adjacent time periods. Assuming that the average click-through rate of the previous time period is 0.15, and the average click-through rate of the current time period is 0.12, the click-through rate has dropped by (0.15-0.12) / 0.15=0.2, or 20%, which exceeds the preset tolerance threshold of 10%. This indicates that the performance of the model may have fluctuated greatly, and the original model structure may not be able to adapt well to the new user behavior pattern.

[0107] At this point, the model structure optimization operation is triggered. First, the convolution kernel size of the temporal convolutional network of the depth prediction model is adjusted. The convolution kernel size affects the model's ability to extract local features. If the current convolution kernel size is too small, it may not be able to capture sufficiently long time series features. If the convolution kernel size is too large, it may result in excessive computation and easy overfitting. For example, the original convolution kernel size of the temporal convolutional network is 3×3. After analysis and experimentation, it can be tried to adjust it to 5×5 to enhance the model's ability to capture longer time series features.

[0108] At the same time, the hidden layer dimension of the gated recurrent unit network is adjusted. The hidden layer dimension determines the complexity of the information that the gated recurrent unit network can learn and represent. If the hidden layer dimension is too low, the model may not be able to learn sufficiently complex user behavior evolution patterns; if the hidden layer dimension is too high, it may lead to overfitting and waste of computing resources. Assuming that the hidden layer dimension of the original gated recurrent unit network is 64, after evaluation and experimentation, it can be adjusted to 128, allowing the deep prediction model to learn more complex changes in user behavior states. Through this model structure optimization operation, the deep prediction model can better adapt to new user data and behavior patterns, improving the accuracy and stability of advertising content preference prediction.

[0109] Based on the above steps, by acquiring a set of historical interaction data from target users, cross-dimensional feature fusion processing is performed, and a deep prediction model is used to perform temporal dependency modeling to generate advertising content preference predictions. A prioritized list of candidate advertising content is then determined through dynamic weight matching. Finally, a set of adapted advertising content is pushed to the target users, and the deep prediction model parameters are updated and the structure optimized based on real-time feedback data. This entire process forms a closed-loop advertising recommendation system that continuously adjusts recommendation strategies based on users' latest behavioral data, improving the accuracy and effectiveness of advertising recommendations, providing users with more personalized advertising content that better meets their needs, and also generating higher conversion rates and revenue for advertisers.

[0110] Figure 2 The hardware structure of the advertising service system 100 provided in the embodiment of the present application for implementing the above-mentioned intelligent advertising content recommendation method using the application prediction model is shown as follows. Figure 2 As shown, the advertising service system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0111] In one possible design, the advertising serving system 100 may be a single server or a server group. The server group may be centralized or distributed (for example, the advertising serving system 100 may be a distributed system). In some embodiments, the advertising serving system 100 may be local or remote. For example, the advertising serving system 100 may access information and / or data stored in the machine-readable storage medium 120 via a network. For another example, the advertising serving system 100 may be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the advertising serving system 100 may be implemented on an advertising serving system. By way of example only, the advertising serving system may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, or the like, or any aggregation thereof.

[0112] Machine-readable storage medium 120 may store data and / or instructions. In some embodiments, machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, machine-readable storage medium 120 may store data and / or instructions that advertising serving system 100 may execute or use to implement the exemplary methods described herein.

[0113] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the intelligent advertising content recommendation method based on the application prediction model of the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0114] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned advertising service system 100. The implementation principles and technical effects are similar and will not be repeated here in this embodiment.

[0115] In addition, an embodiment of the present application further provides a readable storage medium, in which computer-executable instructions are set. When a processor runs the computer-executable instructions, the intelligent advertising content recommendation method using the above-mentioned application prediction model is implemented.

[0116] It should be noted that, in order to simplify the description of the present disclosure and thus facilitate understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present disclosure, multiple features may sometimes be combined into one embodiment, figure, or description thereof. Similarly, it should be noted that, in order to simplify the description of the present disclosure and thus facilitate understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present disclosure, multiple features may sometimes be combined into one embodiment, figure, or description thereof.

Claims

1. An intelligent advertising content recommendation method using a prediction model, characterized in that: The method comprises: Acquire a historical interaction data set of a target user in an advertising interaction scenario, wherein the historical interaction data set includes multiple advertising interaction sequences, each of which is composed of a user identifier, an advertising content identifier, an interaction behavior type, and an interaction timestamp; Performing cross-dimensional feature fusion processing on the historical interaction data set to generate a combined behavior feature set and an advertisement content-related feature set for each advertisement interaction sequence; Calling a preset deep prediction model to perform temporal dependency modeling on the combined behavioral feature set to generate an advertising content preference prediction result, wherein the advertising content preference prediction result includes a user's click probability distribution for candidate advertising content and an interaction intention classification label; performing dynamic weight matching based on the advertising content preference prediction result and the advertising content associated feature set to determine a priority ranking list of candidate advertising contents; Pushing a set of adapted advertising content to target users based on the priority sorted list, and updating parameter configuration of the depth prediction model according to real-time feedback data; The performing cross-dimensional feature fusion processing on the historical interaction data set to generate a combined behavior feature set and an advertisement content-related feature set for each advertisement interaction sequence includes: Performing time window segmentation processing on the interactive behavior types in the advertisement interactive sequence to generate a plurality of behavior segment units, each behavior segment unit including continuous interactive behaviors within a preset time length; Extracting the behavior density feature and interaction intensity feature of each behavior segment unit, wherein the behavior density feature is determined by the interaction frequency per unit time, and the interaction intensity feature is determined by the weighted value of the interaction duration and the operation depth; Performing semantic parsing on the advertising material corresponding to the advertising content identifier to extract semantic topic distribution features of the advertising text and attention weight features of the visual elements; Perform feature splicing on the behavior density features, interaction intensity features, semantic topic distribution features, and attention weight features corresponding to the same advertising interaction sequence to generate the combined behavior feature set; Calling the advertising knowledge base to match the product category tree corresponding to the advertising content identifier, extracting the hierarchical association features of the category path and the user's historical preference matching degree, and generating the advertising content association feature set; The step of extracting the hierarchical association features of the category path and the matching degree of the user's historical preferences to generate the advertisement content association feature set includes: Parsing the parent-child node relationship of the product category tree to generate category graph structure data including hierarchical connection relationships; Inputting the category graph structure data into a graph neural network, and generating a hierarchical perception embedding vector for each category node by propagating features of adjacent nodes; Counting the historical exposure times and click-through conversion rate of each category node in the target user's historical interaction data set to generate a user category preference distribution vector; Performing a matrix product operation on the hierarchical perception embedding vector and the user category preference distribution vector to generate a category node matching weight; The hierarchical perception embedding vector and the category node matching weight are concatenated to generate a dimension-aligned feature matrix of the advertising content-related feature set.

2. The intelligent advertising content recommendation method using a prediction model according to claim 1, characterized in that: The calling of a preset deep prediction model to perform time-series dependency modeling on the combined behavioral feature set to generate an advertising content preference prediction result includes: Inputting the combined behavior feature set into the temporal convolutional network of the deep prediction model to extract local patterns and generate primary temporal features for each behavior segment unit; Performing a multi-head self-attention mechanism on the primary temporal features to capture the long-range dependencies between different behavior fragment units and generate enhanced temporal features; Inputting the enhanced temporal features into the gated recurrent unit network of the deep prediction model for state update to generate a dynamic hidden state sequence of user behavior evolution; Performing maximum pooling processing on the dynamic hidden state sequence to extract a global behavior representation vector, and performing feature cross calculation on the global behavior representation vector and a feature set associated with the advertising content to obtain a feature cross calculation result; The feature cross-reaction result is nonlinearly transformed through the multi-layer perceptron network of the deep prediction model to output the advertising content preference prediction result, wherein the click probability distribution is normalized by the Sigmoid function and the interaction intention classification label is generated by the Softmax function.

3. The intelligent advertising content recommendation method using a prediction model according to claim 2, characterized in that: The performing dynamic weight matching based on the advertising content preference prediction result and the advertising content associated feature set to determine a priority ranking list of candidate advertising contents includes: Filtering a subset of candidate advertising content from the advertising content library that matches the target user's region and terminal type; Calculating the cosine similarity between each candidate advertisement content and the global behavior representation vector to generate a content relevance score; Matching the category path hierarchical association features in the advertisement content association feature set according to the interaction intention classification label to generate a category suitability score; Performing a dynamic weighted summation of the click probability distribution, content relevance score, and category suitability score to generate a comprehensive recommendation score, wherein the weight coefficient of the dynamic weighted summation is dynamically adjusted according to the click rate in the real-time feedback data; The candidate advertising content subsets are arranged in descending order according to the comprehensive recommendation scores, the priority ranking list is generated, and advertising content with scores below a preset threshold is filtered out.

4. The intelligent advertising content recommendation method using a prediction model according to claim 1, characterized in that: Pushing an adapted advertising content set to a target user based on the priority sorted list and updating parameter configuration of the depth prediction model according to real-time feedback data includes: Selecting the top N candidate advertisement contents from the priority sorted list to generate an initial push set, wherein N is dynamically set according to the screen size of the user terminal; Monitor users' real-time interaction with the advertising content in the initial push set, record click events, exposure duration, and secondary forwarding operations to generate a real-time feedback data set; Incrementally merge the real-time feedback dataset with the historical interaction dataset, reconstruct the training sample set, and re-divide the validation set and test set; A sliding window mechanism is used to fine-tune the parameters of the deep prediction model online, where the model loss function of the deep prediction model integrates the click-through rate prediction error and the intent classification cross entropy loss; When the average click-through rate fluctuation in the real-time feedback dataset exceeds a preset tolerance threshold, a model structure optimization operation is triggered to adjust the convolution kernel size of the temporal convolutional network and the hidden layer dimension of the gated recurrent unit network of the depth prediction model.

5. The intelligent advertising content recommendation method using a prediction model according to claim 4, characterized in that: The online parameter fine-tuning of the depth prediction model using a sliding window mechanism includes: Setting a sliding window interval with a fixed time window length, and extracting the most recently generated interaction record within the sliding window interval from the real-time feedback dataset as a current fine-tuning sample; Input the current fine-tuning sample into the last two fully connected layers of the depth prediction model, unfreeze the weight parameters of the last two fully connected layers and freeze the parameters of the remaining layers; Calculating a gradient direction based on the click-through rate prediction error and the intent classification cross entropy loss of the current fine-tuning sample, and updating the weight parameters of the last two fully connected layers along the gradient direction using an adaptive momentum optimization algorithm; Performing exponential moving average processing on the updated weight parameters of the last two fully connected layers and the corresponding historical weight parameters before the update to generate smoothed new weight parameters; Loading the smoothed new weight parameters into the online instance of the depth prediction model, and adding the current fine-tuning sample to the validation set to calculate the click-through rate prediction accuracy; When the click-through rate prediction accuracy is higher than a preset threshold of the current online version, a hot update operation is triggered to synchronize the smoothed new weight parameters to the online advertising push service.

6. The intelligent advertising content recommendation method using a prediction model according to claim 3, characterized in that: Calculating the cosine similarity between each candidate advertisement content and the global behavior representation vector to generate a content relevance score includes: Extracting embedding vectors of candidate advertising content from an advertising content library, where the embedding vectors are generated by jointly encoding the advertising text and visual features using a pre-trained dual-tower model; Perform L2 normalization on the global behavior representation vector and calculate the dot product between the normalized global behavior representation vector and each candidate ad content embedding vector to generate the raw similarity score; Perform piecewise linear transformation on the original similarity scores and map the score interval to the range of [0, 1] to generate content relevance scores; In addition, the content relevance score of the candidate advertising content is downgraded based on the historical exposure frequency of the candidate advertising content.

7. The intelligent advertising content recommendation method using a prediction model according to claim 2, characterized in that: The performing maximum pooling processing on the dynamic hidden state sequence to extract a global behavior representation vector includes: Divide the dynamic hidden state sequence into sliding windows along the time dimension to generate multiple local state subsequences; Perform element-by-element maximum selection on the hidden state vector in each local state subsequence to generate local pooling features; All local pooling features are concatenated and compressed to the preset dimension through a fully connected layer to generate an intermediate aggregation vector; The intermediate aggregation vector is residually connected with the terminal temporal features output by the temporal convolutional network, and the residual connection results are layer-normalized to generate the final global behavior representation vector for downstream prediction tasks.

8. An advertising service system, characterized in that: The advertising service system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to run the programs, instructions or codes in the memory to implement the intelligent advertising content recommendation method using the application prediction model described in any one of claims 1-7.

Images