An advertisement putting strategy optimization method based on topology data analysis

Abstract

The application discloses an advertisement putting strategy optimization method based on topological data analysis, relates to the technical field of data analysis, and comprises the following steps: constructing a topological characteristic prior distribution model, calculating the structural correlation of ring topological characteristics in the dynamic topological network and the prior distribution model, adaptively weighting the ring topological characteristics according to the structural correlation, generating a weighted topological characteristic graph, performing weight distribution optimization on the weighted topological characteristic graph according to the ring topological characteristics, and generating a cross-channel putting sequence through topological centrality constraint of nodes in the dynamic topological network to obtain an initial adaptive putting strategy. The application generates a cross-channel putting sequence through weight distribution optimization of ring topological characteristics and topological centrality constraint to form an initial adaptive putting strategy, and realizes dynamic balance of resource allocation proportion and channel cooperation priority.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method for optimizing advertising delivery strategies based on topological data analysis. Background Technology

[0002] Advertising strategy optimization is a core research area in digital marketing, and it has developed rapidly in recent years due to the increasing complexity of the multi-channel advertising ecosystem. Traditional methods mainly rely on statistical analysis and machine learning techniques, such as linear regression, decision trees, and neural network models based on click-through rate, conversion rate, or ROI, to predict advertising effectiveness and optimize resource allocation. In recent years, the introduction of deep learning technology has further improved feature extraction and prediction accuracy, using convolutional neural networks or recurrent neural networks to process user behavior sequences and capture cross-channel interaction patterns. In addition, graph neural networks are used to model the correlation between advertising channels, analyzing channel synergy effects through node embedding and edge weights.

[0003] However, existing advertising optimization methods still have room for improvement. For example, existing methods lack a dynamic adjustment mechanism based on topological centrality and circular features when generating delivery sequences, making it difficult to balance resource allocation ratios with channel collaboration priorities. In addition, traditional statistical and machine learning methods usually assume that the interactions between channels are linear or low-order relationships, making it difficult to capture high-order topological features (such as circular paths), resulting in insufficient representation of complex collaboration patterns and limiting the accuracy of resource allocation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an advertising placement strategy optimization method based on topology data analysis to solve the problem of balancing resource allocation ratio and channel collaboration priority.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for optimizing advertising delivery strategies based on topology data analysis, comprising:

[0008] Real-time acquisition of advertising data from multiple advertising channels, construction of advertising feature vectors, and mapping to obtain dynamic point cloud data;

[0009] The dynamic point cloud data is structurally decomposed using a nonlinear topology dimensionality reduction method to obtain an intermediate topology structure. A low-dimensional topology representation is generated through a topology filtering mechanism, and loop topology features are extracted to obtain a dynamic topology network.

[0010] A prior distribution model of topological features is constructed, the structural correlation between the ring topological features in the dynamic topological network and the prior distribution model is calculated, and the ring topological features are adaptively weighted according to the structural correlation to generate a weighted topological feature map.

[0011] Based on the ring topology features, the weighted topology feature map is optimized by weight allocation, and a cross-channel delivery sequence is generated by the topological centrality constraint of the nodes in the dynamic topology network to obtain the initial adaptive delivery strategy.

[0012] A channel collaboration behavior model is established and used as an optimization criterion to dynamically adjust the weight allocation and delivery sequence of the initial adaptive delivery strategy. When the circular topology feature is detected to deviate from the optimization criterion, the final adaptive delivery strategy is generated.

[0013] As a preferred embodiment of the advertising placement strategy optimization method based on topological data analysis described in this invention, the method involves: acquiring placement data from multiple advertising channels in real time, constructing advertising feature vectors, and mapping them to obtain dynamic point cloud data. Specifically:

[0014] Collect user exposure logs, cross-channel click sequences, and conversion event records from multiple advertising channels to generate the raw campaign dataset;

[0015] The original delivery dataset is timestamped, divided into time windows, and the channel interaction data within each time window is statistically analyzed.

[0016] Based on channel interaction data, the feature values ​​of each advertising channel are calculated, advertising feature vectors are constructed, and mapped to a high-dimensional topological space to generate dynamic point cloud data.

[0017] As a preferred embodiment of the advertising placement strategy optimization method based on topology data analysis described in this invention, the step of using a nonlinear topology dimensionality reduction method to perform structured decomposition on the dynamic point cloud data to obtain an intermediate topology structure specifically involves:

[0018] Local density analysis of dynamic point cloud data is used to determine the neighborhood distribution characteristics of each data point.

[0019] Based on the neighborhood distribution characteristics, a nonlinear dimensionality reduction cover set is constructed to divide the dynamic point cloud data into multiple overlapping subsets. Clustering operations are performed on each overlapping subset to generate local topological centers, which are then combined into an intermediate topological structure.

[0020] As a preferred embodiment of the advertising placement strategy optimization method based on topology data analysis described in this invention, the step of generating a low-dimensional topology representation through a topology filtering mechanism and simultaneously extracting ring-shaped topology features to obtain a dynamic topology network specifically involves:

[0021] Based on the intermediate topology, a topology filtering function is constructed, the filtering value of each local topology center is calculated, and the intermediate topology is mapped to a low-dimensional space according to the filtering value to generate a low-dimensional topology representation.

[0022] Perform homology analysis on the low-dimensional topological representation to extract ring topological features;

[0023] Based on the characteristics of ring topology and low-dimensional topology representation, a dynamic topology network is constructed.

[0024] As a preferred embodiment of the advertising placement strategy optimization method based on topology data analysis described in this invention, the step of constructing a prior distribution model of topology features and calculating the structural correlation between the ring topology features in the dynamic topology network and the prior distribution model specifically involves:

[0025] Collect historical advertising channel collaboration data, extract typical ring-shaped topological features, and construct a prior distribution model of topological features;

[0026] Feature encoding is performed on the ring topological features in the dynamic topological network to generate topological feature vectors. The similarity between the topological feature vectors and the prior distribution model of topological features is calculated to obtain the structural correlation coefficient.

[0027] Based on the structural correlation coefficient, the distribution parameters of the ring topological features are adjusted to generate the adjusted ring topological features.

[0028] As a preferred embodiment of the advertising placement strategy optimization method based on topological data analysis described in this invention, the step of adaptively weighting the annular topological features according to the structural correlation to generate a weighted topological feature map specifically includes:

[0029] Based on the structural correlation coefficient and the adjusted ring topological features, the weighting coefficient of each ring topological feature is calculated, and the weighting coefficient is applied to the ring topological features to redistribute the feature weights.

[0030] Based on the redistributed ring topology features, update the node and edge attributes of the dynamic topology network, and convert the updated dynamic topology network into a weighted topology feature graph.

[0031] As a preferred embodiment of the advertising placement strategy optimization method based on topology data analysis described in this invention, the step of performing weight allocation optimization on the weighted topology feature map according to the ring-shaped topology features specifically includes:

[0032] Stability parameters of annular topological features are extracted from the weighted topological feature map, and the channel collaborative contribution value corresponding to each annular topological feature is calculated.

[0033] Based on the channel collaboration contribution value, adjust the weight allocation of nodes in the weighted topology feature graph;

[0034] The adjusted weight allocation is normalized to generate an optimized weight allocation scheme.

[0035] As a preferred embodiment of the advertising delivery strategy optimization method based on topology data analysis described in this invention, the step of generating a cross-channel delivery sequence by means of the topological centrality constraints of nodes in a dynamic topology network specifically includes:

[0036] The resource allocation ratio of nodes in the weighted topology feature graph is determined based on the optimized weight allocation scheme.

[0037] Calculate the topological centrality index for nodes in the dynamic topology feature graph, determine the channel importance of each node, sort the nodes based on channel importance and resource allocation ratio, and generate an initial deployment priority sequence.

[0038] By adjusting the path length of the ring topology feature, the node order in the initial delivery priority sequence is adjusted to obtain the cross-channel delivery sequence.

[0039] As a preferred embodiment of the advertising placement strategy optimization method based on topology data analysis described in this invention, the step of establishing a channel collaboration behavior model and using the channel collaboration behavior model as an optimization criterion to dynamically correct the weight allocation and placement sequence of the initial adaptive placement strategy specifically involves:

[0040] Based on historical delivery data and ring topology characteristics, a channel collaborative behavior model is constructed to generate optimization criteria;

[0041] The optimized weight allocation scheme and cross-channel delivery sequence in the initial adaptive delivery strategy are used for feature mapping to obtain the strategy feature vector;

[0042] The optimization criteria of the comparative strategy feature vector and the channel collaborative behavior model are used to detect the degree of deviation of the ring topology features;

[0043] Based on the degree of deviation, the optimized weight allocation scheme and cross-channel delivery sequence are adjusted to generate a revised delivery strategy.

[0044] As a preferred embodiment of the advertising delivery strategy optimization method based on topology data analysis described in this invention, the generation of the final adaptive delivery strategy specifically includes:

[0045] The revised delivery strategy is validated for consistency, and the coordination consistency index of weight allocation and delivery sequence is calculated.

[0046] Based on the consistency index, the channel weight parameters in the revised delivery strategy are adjusted and applied to the delivery sequence to obtain the optimized delivery sequence configuration and generate the final adaptive delivery strategy.

[0047] The beneficial effects of this invention are as follows: By utilizing nonlinear topological dimensionality reduction methods and persistent homology analysis to extract cyclic topological features, a dynamic topological network is generated, achieving accurate characterization of complex collaborative patterns among advertising channels. This overcomes the shortcomings of traditional statistical and machine learning methods that assume linear or low-order relationships and struggle to capture high-order topological features, thus improving the accuracy of resource allocation. Furthermore, by optimizing weight allocation based on cyclic topological features and generating cross-channel delivery sequences through topological centrality constraints, an initial adaptive delivery strategy is formed, achieving a dynamic balance between resource allocation ratios and channel collaboration priorities. This addresses the limitations of existing advertising delivery methods that lack dynamic adjustment mechanisms for topological centrality and cyclic features, further enhancing the collaborative efficiency of delivery sequences and the adaptability of resource allocation, significantly improving the accuracy and effectiveness of multi-channel advertising delivery. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart of an advertising delivery strategy optimization method based on topology data analysis.

[0050] Figure 2 Generate flowcharts for dynamic topology networks.

[0051] Figure 3 Flowchart for generating weighted topological feature maps and initial strategies.

[0052] Figure 4 Flowcharts for dynamic correction and final strategy output. Detailed Implementation

[0053] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0054] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0056] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for optimizing advertising delivery strategies based on topology data analysis, including the following steps:

[0057] S1: Real-time acquisition of advertising data from multiple advertising channels, construction of advertising feature vectors, and mapping to obtain dynamic point cloud data;

[0058] User exposure logs, cross-channel click sequences, and conversion event records are obtained from server interfaces of multiple advertising channels and stored as the raw campaign dataset. Specifically, user exposure logs record the display time and user identifier of each ad on each channel, cross-channel click sequences record the user's click behavior path across different channels, and conversion events record the time and channel information when the user completes a purchase or registration or other target behavior.

[0059] The user exposure logs, cross-channel click sequences, and conversion event records in the original campaign dataset are timestamped and the time format is standardized to ensure the time sequence consistency of all data. The original campaign dataset after timestamping is divided into time windows of fixed length (each hour) to generate multiple time window datasets.

[0060] The fixed length of one hour is set based on the typical periodic changes in user behavior in the advertising scenario. One hour can balance data granularity and computing efficiency, capture real-time interactive dynamics, and avoid overly fine calculations that lead to complex calculations or overly coarse calculations that ignore short-term fluctuations.

[0061] For each time window dataset, the number of user impressions, clicks, conversions, and cross-channel click sequence jump frequency for each advertising channel are counted to generate channel interaction data. Based on the channel interaction data, feature values ​​for each advertising channel are extracted, including impression frequency, click-through rate, conversion rate, cross-channel jump probability, and user retention time.

[0062] The feature values ​​of each advertising channel are combined into a multi-dimensional vector to generate an advertising feature vector, where each dimension corresponds to a feature value. Each advertising feature vector is then normalized, adjusting the range of feature values ​​for each dimension to 0 to 1, generating a normalized advertising feature vector. Finally, the Euclidean distance function is applied to calculate the distance matrix between each advertising channel, as shown in the formula:

[0063] ;

[0064] in, Indicates the first The and the first The Euclidean distance between advertising channels, also known as the distance matrix, Indicates the first The advertising channel in the first Normalized values ​​on each feature Indicates the first The advertising channel in the first Normalized values ​​on each feature This represents the number of dimensions of the ad feature vector.

[0065] Based on the distance matrix, the normalized advertising feature vectors are projected onto a high-dimensional topological space to form a set of points representing the dynamic relationships between channels, which are then stored as dynamic point cloud data, recording the channel identifier and time window information corresponding to each point.

[0066] S2: The dynamic point cloud data is decomposed into a structured form using a nonlinear topology dimensionality reduction method to obtain an intermediate topology structure. A low-dimensional topology representation is generated through a topology filtering mechanism, and the loop topology features are extracted to obtain a dynamic topology network.

[0067] S2.1: For each data point in the dynamic point cloud data, count the number of neighboring points within a specified radius in the high-dimensional topological space, generating a local density value for each data point. The specified radius is set to 0.1 based on the standardized average of the Euclidean distances between all data points in the dynamic point cloud data. This ensures coverage of typical interaction points within the local neighborhood while avoiding excessive noise interference. If the radius is higher than 0.2, it may cover more sparse points, potentially introducing atypical interactions and causing the local density value to contain too much irrelevant information, reducing the accuracy of the topological structure. For the local density value of each data point, compare the distribution positions of its neighboring points in the high-dimensional topological space, calculate the spatial distribution variance of the neighboring points relative to the data point center, generate the uniformity of the neighboring point distribution, and determine the magnitude of change within adjacent time windows to obtain the density change rate. Record the uniformity of the neighboring point distribution and the density change rate of each data point as neighborhood distribution characteristics. Further explanation: The expression for calculating the uniformity of the neighboring point distribution by calculating the spatial distribution variance is:

[0068] ;

[0069] in, Indicates the first Uniformity of the distribution of neighboring points of each data point Indicates the first Local density values ​​of each data point Indicates the first The set of neighbor points of a data point Represents the set of neighboring points The Middle The coordinate vectors of each neighboring point in the high-dimensional topological space. Indicates the first Set of neighbor points of a data point The coordinate mean vector.

[0070] Based on the local density value in the neighborhood distribution characteristics, the coverage radius of each data point is set to the reciprocal of the local density value. This is set according to the principle of coverage set construction. If the coverage radius is higher than the local density value, the coverage radius decreases, generating a more refined topology suitable for dense areas. If the coverage radius is lower than the local density value, the coverage radius increases, generating a wide coverage structure suitable for sparse areas and avoiding fragmentation. At the same time, the overlap ratio is set to 0.3 times the density change rate. This is set according to the continuity requirement to ensure sufficient overlap in areas with large dynamic changes and to maintain the coherence of the topology.

[0071] Based on the set coverage radius and overlap ratio, a nonlinear dimensionality reduction coverage set is constructed to divide the dynamic point cloud data into multiple overlapping subsets. Each subset contains some data points and neighboring points. The K-means clustering method is applied to each overlapping subset. By iteratively calculating the distance from the data points in the overlapping subset to the initial random center, the data points are assigned to the nearest center, and the center position is updated until the final cluster center is determined, generating local topological centers. The center point coordinates and the set of data points to which each local topological center belongs are recorded. Based on the uniformity of the distribution of neighboring points in the neighborhood distribution characteristics, the center point coordinates of each local topological center are compared, and local topological centers with adjacent uniform distribution are connected to generate an intermediate topological structure.

[0072] S2.2: Based on the local topological centers of the intermediate topological structure, select a topological filtering function based on the density of the local topological center points. Set the input of the topological filtering function to the coordinates of the center point of each local topological center. Apply the topological filtering function to the coordinates of the center point of each local topological center to generate a filtered value for each local topological center, which is recorded as a scalar value. The expression is:

[0073] ;

[0074] in, Indicates the first The filtered value of a local topological center Indicates the first The weight of each local topological center is the reciprocal of the center point density value, which is obtained by counting the number of data points in the neighborhood. Indicates the first Coordinates of the center point of each local topology center The Euclidean norm.

[0075] Based on the filtered value of each local topological center, the coordinates of the center points of the intermediate topological structure are projected into a low-dimensional space to generate a low-dimensional topological representation. The adjacency relationships of each center are preserved and combined into an adjacency matrix. A persistent homology analysis method is applied to the adjacency matrix to extract the connection distances between each center, and a set of connection distances is generated based on Euclidean distance. Starting from the minimum distance, the filtering parameters are gradually increased, and the appearance and disappearance of connections in the adjacency matrix are recorded to form a topological evolution sequence that changes with the filtering parameters. One-dimensional homology groups are identified in the filtering sequence to determine the topological features that form closed paths and extract cyclic topological features. The extracted cyclic topological features are stored as a set of cyclic paths, and the center identifier and path length of each closed path are recorded.

[0076] Based on the set of loop paths and the adjacency relationships of the low-dimensional topology representation, local topology centers are mapped to nodes, loop paths are mapped to edges, a dynamic topology network is generated, and the topology attributes of nodes and edges are stored.

[0077] S3: Construct a prior distribution model of topological features, calculate the structural correlation between the ring topological features and the prior distribution model in the dynamic topological network, adaptively weight the ring topological features according to the structural correlation, and generate a weighted topological feature map;

[0078] S3.1: Obtain historical advertising channel collaboration data from the server interfaces of multiple advertising channels, including user click paths, conversion records, and exposure sequences across different channels, and store them as a historical dataset. For the click paths and conversion records in the historical dataset, construct an adjacency matrix based on channel connections to represent user jump relationships between channels. Apply persistent homology analysis to the adjacency matrix based on channel connections to generate a filtering sequence, tracking the topological evolution of channel connections as a function of a distance threshold. The distance threshold is set based on the standardized average of jump distances in the historical advertising channel collaboration data, with a value of 0.6. Values ​​higher than 0.6 will cover more sparse jump patterns and may introduce irrelevant or noisy connections. Identify one-dimensional homology groups in the filtering sequence to determine channel jump patterns forming closed paths, extract typical cyclic topological features, record them as a set of typical cyclic paths, and statistically analyze the frequency and length of each typical cyclic path to generate a prior distribution model of topological features, storing it as a probability distribution table.

[0079] For the ring topological features in the dynamic topological network, extract the node sequence and length of each ring path and convert them into a topological feature vector of fixed dimension (e.g., 64). Compare the topological feature vector with the probability distribution table of the prior distribution model of the topological features. Based on the cosine similarity method, determine the similarity between the two, generate the structural correlation coefficient, and store it as a scalar value.

[0080] Based on the structural relevance coefficient, the path length weight of the loop topological features is adjusted. Specifically, for loop paths with a structural relevance coefficient higher than the relevance threshold, the path length weight is increased; for loop paths with a structural relevance coefficient lower than the relevance threshold, the path length weight is decreased, generating the adjusted loop topological features. The relevance threshold is set based on the average of the cosine similarity distribution between the topological feature vector and the prior distribution model of the topological features, and is set to 0.7. This is because 0.7 can effectively distinguish between highly and low-relevance loop paths, ensuring that the adjusted loop topological features focus on the core collaborative relationships of the advertising channels.

[0081] S3.2: For each ring-shaped topological feature's ring path, extract the structural correlation coefficient and the adjusted path length weight, and generate a weighted coefficient based on the product of the structural correlation coefficient and the path length weight; apply the weighted coefficient to each ring-shaped topological feature's ring path, adjust the eigenvalues ​​of the ring path, complete the feature weight redistribution, and generate the redistributed ring-shaped topological feature; based on the redistributed ring-shaped topological feature, update the weights of nodes in the dynamic topological network, setting the node weights to the sum of the weighted coefficients of the corresponding ring path, update the edge weights, setting the edge weights to the average weighted coefficients of the shared nodes between ring paths; for the updated dynamic topological network, extract the node weights and edge weights, construct a topological structure containing weighted nodes and weighted edges, generate a weighted topological feature graph, and store it as a set of node and edge weights.

[0082] Preferably, compared to traditional machine learning methods, this method constructs a prior distribution model of topological features through topological data analysis, extracts typical ring-shaped topological features using historical advertising channel collaboration data and persistent coherence analysis, and generates a probability distribution table to accurately capture channel collaboration patterns. Furthermore, it generates a weighted topological feature map by adaptively weighting the ring-shaped topological features using structural correlation coefficients, improving the representation accuracy of advertising channel collaboration relationships and the optimization effect of dynamic topological networks. This enhances the targeting of resource allocation and the collaborative efficiency of advertising delivery, making it suitable for complex multi-channel scenarios.

[0083] S4: Perform weight allocation optimization on the weighted topology feature graph based on the ring topology characteristics, and generate a cross-channel delivery sequence through the topological centrality constraint of nodes in the dynamic topology network to obtain the initial adaptive delivery strategy;

[0084] S4.1: Extract the ring path of the ring topology feature from the weighted topology feature map, count the occurrence frequency of each ring path in different time windows, generate the stability parameter of the ring path, and record it as a set of frequency values; for the stability parameter of each ring path, generate the corresponding channel collaborative contribution value based on the product of the frequency value and the node weight in the weighted topology feature map, and store it as a set of contribution values.

[0085] Based on the channel collaboration contribution value, the weights of nodes in the weighted topology feature graph are adjusted. Specifically, for nodes with contribution values ​​higher than the contribution threshold, the weight is increased and multiplied by an amplification factor; for nodes with contribution values ​​lower than the contribution threshold, the weight is decreased and multiplied by a shrinkage factor, thus generating an adjusted set of node weights.

[0086] To further explain, the contribution threshold is set to a standardized average of the channel synergy contribution values ​​in the weighted topological feature graph, with a value of 0.5, to effectively distinguish between high-contribution and low-contribution nodes and adapt to advertising scenarios; the amplification factor is set to 1.1 based on the typical enhancement ratio of advertising channel synergy, indicating that the weight of high-contribution nodes is moderately enhanced to maintain the stability of resource allocation; the reduction factor is set to 0.9 based on the typical decay ratio of advertising channel synergy, indicating that the weight of low-contribution nodes is moderately reduced to avoid excessively weakening effective channels.

[0087] The adjusted set of node weights is normalized to scale the weight values ​​of all nodes to the range of 0 to 1, generating an optimized weight allocation scheme, which is then stored as a standardized weight set.

[0088] S4.2: Extract the standardized weight value of each node from the optimized weight allocation scheme, divide the standardized weight value of each node by the sum of the standardized weight values ​​of all nodes, generate the resource allocation ratio of the nodes in the weighted topology feature graph, and store it as a set of ratio values.

[0089] For each node in the dynamic topology network, the number of edge connections with other nodes in the dynamic topology network is counted, and the degree centrality index of the node is generated and normalized to between 0 and 1. The normalized degree centrality index is compared with a preset importance threshold. The importance threshold is set based on the average value of the normalized degree centrality values, which is 0.5. Nodes with a value higher than 0.5 are considered to have high importance, and nodes with a value lower than 0.5 are considered to have low importance. The importance results of each node are combined into an importance value set.

[0090] For each node, a weighting factor is set based on the weighted average of the resource allocation ratio and the channel importance value, combining the resource allocation ratio and the channel importance value. For example, the weighting factor of the resource allocation ratio is 0.6 and the channel importance value is 0.4. The higher weight of the resource allocation ratio is to optimize the efficiency of budget allocation, while the slightly lower weight of the channel importance value is to reflect the synergy in the dynamic topology network, thereby generating a comprehensive ranking score. The nodes are then sorted from high to low according to the comprehensive ranking score to generate an initial deployment priority sequence.

[0091] The path length of each ring path is extracted from the ring topology features. The ring path to which the node belongs in the initial deployment priority sequence is identified. For ring paths with a path length greater than 5, the priority order of the node is increased to generate an adjusted node order. 5 is determined based on the standardized average of the path lengths in the ring topology features. A length greater than 5 indicates strong channel synergy, while a length less than 5 indicates weak synergy.

[0092] Based on the adjusted node order, the nodes are mapped to the corresponding advertising channels, the advertising channels are arranged in order, a cross-channel delivery sequence is generated, and stored as a channel priority list.

[0093] By combining the standardized weight set in the optimized weight allocation scheme with the cross-channel delivery sequence, an initial adaptive delivery strategy is obtained, which includes a strategy configuration table containing the weight value and delivery priority of each advertising channel.

[0094] Preferably, compared to existing advertising strategies, this method optimizes advertising strategies through topological data analysis. It utilizes the characteristics of ring topology and dynamic topological networks to capture the complex collaborative relationships between channels and generate an initial adaptive advertising strategy. Compared to traditional statistical optimization methods, this method effectively improves the accuracy of cross-channel resource allocation and the synergy of the advertising sequence through nonlinear topological dimensionality reduction and topological centrality constraints, thereby enhancing the adaptability and stability of advertising effects and making it suitable for dynamic multi-channel advertising scenarios.

[0095] S5: Establish a channel collaboration behavior model, use the channel collaboration behavior model as the optimization criterion, dynamically correct the weight allocation and delivery sequence of the initial adaptive delivery strategy, and generate the final adaptive delivery strategy when the circular topology feature is detected to deviate from the optimization criterion.

[0096] S5.1: Extract click paths, conversion records, and exposure sequences of advertising channels from historical delivery data, combine them with the ring path set of ring topological features, generate a channel collaborative behavior model, and store it as a probability distribution table describing the probability of channel jumps; based on the probability distribution table in the channel collaborative behavior model, extract optimization indicators that reflect the channel collaboration effect, generate optimization criteria, and record them as a collaborative probability target set.

[0097] For the optimized weight allocation scheme in the initial adaptive delivery strategy, the standardized weight value of each advertising channel is extracted. For the cross-channel delivery sequence, the priority order of the channels is extracted. The standardized weight value and priority order are combined into a multi-dimensional vector to generate the strategy feature vector.

[0098] The standardized weight values ​​and priority order of each advertising channel are extracted from the strategy feature vector and combined into a weight value set and a priority order set. The collaborative probability target set is extracted from the optimization criteria, and the jump probability target value of each advertising channel is obtained and stored as a probability target value set. For each advertising channel, the standardized weight value in the weight value set is compared with the jump probability target value in the probability target value set, and a weight difference value is generated based on the absolute value of the difference between the two. At the same time, for each advertising channel, the ranking in the priority order set is compared with the jump probability target value in the probability target value set, and a priority difference value is generated based on the absolute value of the ranking difference.

[0099] The weight difference value and priority difference value of each advertising channel are weighted and averaged. For example, the weight factor of the weight difference value is set to 0.7 and the weight factor of the priority difference value is set to 0.3. This setting is based on the fact that the weight has a greater impact on budget allocation than the priority order in advertising placement, thereby generating the difference value of channel synergy effect and integrating it into a difference value set.

[0100] Based on the set of difference values, the standardized weight values ​​of each advertising channel in the optimized weight allocation scheme are adjusted, and the weight values ​​are appropriately increased or decreased to approach the synergy probability target; the priority order of channels in the cross-channel delivery sequence is adjusted to make the priority order more in line with the synergy probability target, and a revised delivery strategy is generated.

[0101] S5.2: Extract the weight value and delivery sequence of each advertising channel from the revised delivery strategy, and generate a set of weight values ​​and a list of delivery sequences.

[0102] For the set of weight values, compare the weight value of each advertising channel in the set with the conversion rate of the corresponding channel in the historical delivery data. Based on the average of the absolute values ​​of the differences between the two, generate a weight consistency score. For the delivery sequence list, compare the order of the advertising channels in the delivery sequence list with the order of the channel jump probability in the historical delivery data. Based on the average of the absolute values ​​of the ranking differences, generate a sequence consistency score. Merge the weight consistency score and the sequence consistency score to generate a collaborative consistency index, which is stored as a consistency score set.

[0103] Based on the consistency score set in the consistency index, advertising channels with lower scores are identified. For advertising channels with consistency scores below the consistency threshold, the weight values ​​are appropriately adjusted, increasing or decreasing the weights to more closely match the conversion rates of historical campaign data. For the order of channels with consistency scores below the consistency threshold, their positions in the campaign sequence are adjusted to better match the bounce probabilities of historical campaign data, generating adjusted channel weight parameters and campaign sequences. To further clarify, the consistency threshold is set based on the standardized average of the weight consistency score and the sequence consistency score. As mentioned earlier, if the score is below the consistency threshold, the weight values ​​and the position of the campaign sequence need to be adjusted. If the score is above the consistency threshold, it indicates that the resource allocation is reasonable and no adjustment is required.

[0104] The adjusted channel weight parameters are applied to the delivery sequence, updating the weight value and sequence position of each advertising channel to generate an optimized delivery sequence configuration, which includes a configuration table of weights and sequences. Based on the optimized delivery sequence configuration, the weight value and sequence position of each advertising channel are combined to generate the final adaptive delivery strategy, which is stored as a strategy table containing channel weights and priorities, and the generation timestamp is recorded.

[0105] This embodiment also provides a computer device applicable to the advertising placement strategy optimization method based on topology data analysis, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the advertising placement strategy optimization method based on topology data analysis proposed in the above embodiment.

[0106] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0107] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the advertising placement strategy optimization method based on topology data analysis as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0108] In summary, this invention: by utilizing nonlinear topological dimensionality reduction methods and persistent homology analysis to extract cyclic topological features, a dynamic topological network is generated, achieving accurate characterization of complex collaborative patterns among advertising channels. This overcomes the shortcomings of traditional statistical and machine learning methods that assume linear or low-order relationships and struggle to capture high-order topological features, thus improving the accuracy of resource allocation. Furthermore, by optimizing weight allocation based on cyclic topological features and generating cross-channel delivery sequences using topological centrality constraints, an initial adaptive delivery strategy is formed, achieving a dynamic balance between resource allocation ratios and channel collaboration priorities. This addresses the limitations of existing advertising delivery methods that lack dynamic adjustment mechanisms for topological centrality and cyclic features, further enhancing the collaborative efficiency of delivery sequences and the adaptability of resource allocation, significantly improving the accuracy and effectiveness of multi-channel advertising delivery.

[0109] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for optimizing advertising delivery strategies based on topology data analysis, characterized in that: include, Real-time acquisition of advertising data from multiple advertising channels, construction of advertising feature vectors, and mapping to obtain dynamic point cloud data; The dynamic point cloud data is structurally decomposed using a nonlinear topology dimensionality reduction method to obtain an intermediate topology structure. A low-dimensional topology representation is generated through a topology filtering mechanism, and loop topology features are extracted to obtain a dynamic topology network. A prior distribution model of topological features is constructed, the structural correlation between the ring topological features in the dynamic topological network and the prior distribution model is calculated, and the ring topological features are adaptively weighted according to the structural correlation to generate a weighted topological feature map. Based on the ring topology features, the weighted topology feature map is optimized by weight allocation, and a cross-channel delivery sequence is generated by the topological centrality constraint of the nodes in the dynamic topology network to obtain the initial adaptive delivery strategy. A channel collaboration behavior model is established and used as an optimization criterion to dynamically adjust the weight allocation and delivery sequence of the initial adaptive delivery strategy. When the circular topology feature is detected to deviate from the optimization criterion, the final adaptive delivery strategy is generated.

2. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: Real-time acquisition of ad delivery data from multiple advertising channels, construction of ad feature vectors, and mapping to obtain dynamic point cloud data, specifically: Collect user exposure logs, cross-channel click sequences, and conversion event records from multiple advertising channels to generate the raw campaign dataset; The original delivery dataset is timestamped, divided into time windows, and the channel interaction data within each time window is statistically analyzed. Based on channel interaction data, the feature values ​​of each advertising channel are calculated, advertising feature vectors are constructed, and mapped to a high-dimensional topological space to generate dynamic point cloud data.

3. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The method of using nonlinear topological dimensionality reduction to perform structured decomposition on the dynamic point cloud data to obtain an intermediate topological structure is as follows: Local density analysis of dynamic point cloud data is used to determine the neighborhood distribution characteristics of each data point. Based on the neighborhood distribution characteristics, a nonlinear dimensionality reduction cover set is constructed to divide the dynamic point cloud data into multiple overlapping subsets. Clustering operations are performed on each overlapping subset to generate local topological centers, which are then combined into an intermediate topological structure.

4. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The process of generating a low-dimensional topological representation through topological filtering and extracting ring-shaped topological features to obtain a dynamic topological network is as follows: Based on the intermediate topology, a topology filtering function is constructed, the filtering value of each local topology center is calculated, and the intermediate topology is mapped to a low-dimensional space according to the filtering value to generate a low-dimensional topology representation. Perform homology analysis on the low-dimensional topological representation to extract ring topological features; Based on the characteristics of ring topology and low-dimensional topology representation, a dynamic topology network is constructed.

5. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The construction of the prior distribution model of topological features, and the calculation of the structural correlation between the ring topological features in the dynamic topological network and the prior distribution model, specifically involves: Collect historical advertising channel collaboration data, extract typical ring-shaped topological features, and construct a prior distribution model of topological features; Feature encoding is performed on the ring topological features in the dynamic topological network to generate topological feature vectors. The similarity between the topological feature vectors and the prior distribution model of topological features is calculated to obtain the structural correlation coefficient. Based on the structural correlation coefficient, the distribution parameters of the ring topological features are adjusted to generate the adjusted ring topological features.

6. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The step of adaptively weighting the annular topological features based on the structural correlation to generate a weighted topological feature map is specifically as follows: Based on the structural correlation coefficient and the adjusted ring topological features, the weighting coefficient of each ring topological feature is calculated, and the weighting coefficient is applied to the ring topological features to redistribute the feature weights. Based on the redistributed ring topology features, update the node and edge attributes of the dynamic topology network, and convert the updated dynamic topology network into a weighted topology feature graph.

7. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The step of performing weight allocation optimization on the weighted topological feature map based on the ring-shaped topological features is specifically as follows: Stability parameters of annular topological features are extracted from the weighted topological feature map, and the channel collaborative contribution value corresponding to each annular topological feature is calculated. Based on the channel collaboration contribution value, adjust the weight allocation of nodes in the weighted topology feature graph; The adjusted weight allocation is normalized to generate an optimized weight allocation scheme.

8. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The process of generating cross-channel delivery sequences by leveraging the topological centrality constraints of nodes in a dynamic topology network specifically involves: The resource allocation ratio of nodes in the weighted topology feature graph is determined based on the optimized weight allocation scheme. Calculate the topological centrality index for nodes in the dynamic topology feature graph, determine the channel importance of each node, sort the nodes based on channel importance and resource allocation ratio, and generate an initial deployment priority sequence. By adjusting the path length of the ring topology feature, the node order in the initial delivery priority sequence is adjusted to obtain the cross-channel delivery sequence.

9. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The establishment of a channel collaboration behavior model, using this model as an optimization criterion, dynamically modifies the weight allocation and delivery sequence of the initial adaptive delivery strategy, specifically as follows: Based on historical delivery data and ring topology characteristics, a channel collaborative behavior model is constructed to generate optimization criteria; The optimized weight allocation scheme and cross-channel delivery sequence in the initial adaptive delivery strategy are used for feature mapping to obtain the strategy feature vector; The optimization criteria of the comparative strategy feature vector and the channel collaborative behavior model are used to detect the degree of deviation of the ring topology features; Based on the degree of deviation, the optimized weight allocation scheme and cross-channel delivery sequence are adjusted to generate a revised delivery strategy.

10. The advertising placement strategy optimization method based on topology data analysis as described in claim 1, characterized in that: The specific steps for generating the final adaptive delivery strategy are as follows: The revised delivery strategy is validated for consistency, and the coordination consistency index of weight allocation and delivery sequence is calculated. Based on the consistency index, the channel weight parameters in the revised delivery strategy are adjusted and applied to the delivery sequence to obtain the optimized delivery sequence configuration and generate the final adaptive delivery strategy.

Images