A method and system for generating a digital marketing strategy by fusing multi-channel data

Abstract

The application provides a digital marketing strategy generation method and system fusing multi-channel data, relates to the technical field of digital marketing, and performs event segmentation processing on behavior data of a user to obtain heterogeneous behavior sequences of the user in different channels; performs semantic association mapping on consumption behavior segments of the user in different channels, identifies consumption stage association relationships between behaviors in different channels, and further constructs a unified semantic expression chain of consumption behaviors in different channels; identifies decision abnormal sections and behavior chain breakpoints in a consumption behavior evolution process of the user; dynamically completes and corrects a cross-channel behavior path of the user according to the decision abnormal sections and the behavior chain breakpoints to obtain a cross-channel decision track of the user; and generates a digital marketing strategy of the user in a current decision stage through the cross-channel decision track. The application can realize implicit semantic alignment of heterogeneous behaviors of multi-channel users, so that a complete decision link of the user across channels can be accurately restored.

Description

Technical Field

[0001] This application relates to the field of digital marketing technology, and more specifically, to a method and system for generating digital marketing strategies that integrates multi-channel data. Background Technology

[0002] With the rapid development of social media, e-commerce platforms, and private domain operation systems, enterprises have an increasing number of channels to obtain user behavior data. Digital marketing technology is gradually evolving from single-channel analysis to multi-channel collaborative analysis. By comprehensively analyzing users' browsing, interaction, transaction, and feedback behaviors across different digital channels, it is possible to more accurately identify users' consumption preferences, behavioral habits, and stage-specific decision-making tendencies. This improves the accuracy of marketing content delivery and the efficiency of marketing resource allocation, which is of great significance for improving user conversion rates, enhancing user stickiness, and optimizing enterprises' digital marketing capabilities.

[0003] However, in existing digital marketing technologies, the data structures, behavioral tagging systems, and event recording logic of different channels often differ significantly. The behavioral data of the same user across different channels is often stored and analyzed independently. For example, the data expression and behavioral semantics of a user's content browsing behavior on social media channels, product operation behavior on e-commerce transaction channels, and consultation and feedback behavior on private domain interaction channels are significantly different. This makes it difficult to establish a unified correlation between user behaviors across different channels. Most existing technologies can only make simple correlations based on user identifiers or surface behavioral fields, making it difficult to further identify the implicit consumption stage relationships and decision-making evolution logic between behaviors. This makes it easy for the user's behavioral trajectory across different channels to have semantic breaks, missing stages, and incomplete behavioral links. Consequently, the marketing strategy generation process may suffer from audience profiling bias, inaccurate judgment of marketing timing, and insufficient accuracy in content matching. Therefore, how to achieve implicit semantic alignment of heterogeneous user behaviors across multiple channels, thereby accurately reconstructing the complete decision-making link of users across channels, has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method and system for generating digital marketing strategies that integrates multi-channel data, which can achieve implicit semantic alignment of heterogeneous user behaviors across multiple channels, thereby accurately reconstructing the complete cross-channel decision-making process of users.

[0005] Firstly, this application provides a method for generating digital marketing strategies that integrates multi-channel data, the method comprising the following steps:

[0006] Event segmentation processing is performed on user behavior data in social content channels, e-commerce transaction channels, and private domain interaction channels to obtain heterogeneous behavior sequences of users in different channels;

[0007] Based on the behavioral context in each heterogeneous behavioral sequence, semantic association mapping is performed on the consumption behavior fragments of users in different channels to identify the consumption stage association relationship between behaviors in different channels, and a unified semantic expression chain of consumption behavior in different channels is constructed based on the consumption stage association relationship.

[0008] Based on the unified semantic expression chain, the stage continuity of the user's consumption behavior evolution process in different channels is verified, and the abnormal decision-making segments and behavior chain breakpoints of the user in the consumption behavior evolution process are identified.

[0009] Based on the decision-making anomaly segment and the behavior chain breakpoint, the user's cross-channel behavior path is dynamically completed and the path is corrected to obtain the user's cross-channel decision trajectory.

[0010] The cross-channel decision-making trajectory is used to generate digital marketing strategies for users at the current decision-making stage.

[0011] In this embodiment, user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels is segmented into events to obtain heterogeneous user behavior sequences across different channels, specifically including:

[0012] Acquire user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels;

[0013] The behavioral data within each channel is sorted chronologically to obtain the original behavioral time series for each channel.

[0014] The original behavior time series of each channel is segmented based on the time interval between adjacent behaviors and the changes in event type in each channel, resulting in multiple heterogeneous behavior segments within each channel.

[0015] The heterogeneous behavior fragments within each channel are organized in chronological order to form a heterogeneous behavior sequence for that channel, thereby obtaining the heterogeneous behavior sequence of users across different channels.

[0016] In this embodiment, semantic association mapping of user consumption behavior fragments in different channels is performed based on the behavioral context in each heterogeneous behavior sequence to identify the consumption stage association relationship between behaviors in different channels. Specifically, this includes:

[0017] Determine the behavioral context in each heterogeneous behavioral sequence, wherein the behavioral context includes a triggering intent identifier, a behavioral object identifier, and a behavioral result status identifier;

[0018] A semantic association graph between cross-channel behaviors is constructed based on the semantic homology coefficient and temporal adjacency coefficient between the contexts of each behavior.

[0019] For each association edge in the semantic association graph, calculate the consumption stage affiliation degree, and mark the association edges with affiliation degrees exceeding the stage association threshold as consumption stage association edges;

[0020] Based on the transitive and mutually exclusive constraints between all consumption stage association edges, the association consistency is verified to obtain the consumption stage association relationship between different channel behaviors.

[0021] In this embodiment, the association consistency is verified based on the transitive and mutually exclusive constraints between all consumption stage association edges. The specific consumption stage association relationships between different channel behaviors include:

[0022] Traverse all consumption stage association edges and extract the channel label and consumption stage label of each heterogeneous behavior fragment connected by each consumption stage association edge;

[0023] Based on the extracted channel tags and consumption stage tags, the transitive constraints between different channel tags under the same consumption stage tag and the mutual exclusion constraints between different consumption stage tags are identified.

[0024] Based on the transitive constraints, the connected components of the consumption stage association edges are aggregated, and the cross-stage conflict elimination is performed on the aggregated connected components based on the mutual exclusion constraints. The consumption stage association edges retained after eliminating conflicts are used as the consumption stage association relationships between different channel behaviors.

[0025] In this embodiment, constructing a unified semantic expression chain of consumption behavior across different channels based on the aforementioned consumption stage association specifically includes:

[0026] Based on the aforementioned consumption stage association, heterogeneous behavioral fragments from different channels are aggregated into consumption stage behavior clusters, with each consumption stage behavior cluster corresponding to a consumption stage state node.

[0027] For each consumption stage behavior cluster, the heterogeneous behavior fragments are sorted by the time of behavior occurrence, and stage state transition edges are inserted between adjacent behavior nodes to form a local semantic subchain for that consumption stage.

[0028] Extract the entry behavior semantic fingerprint and exit behavior semantic fingerprint of each local semantic subchain, and determine the succession relationship between different consumption stages based on the entry behavior semantic fingerprint and the exit behavior semantic fingerprint;

[0029] By globally connecting all consumption stages according to the aforementioned sequential relationship, a unified semantic expression chain of consumption behavior across different channels is constructed.

[0030] In this embodiment, based on the unified semantic expression chain, the stage continuity verification of the user's consumption behavior evolution process in different channels is performed, and the abnormal decision-making segments and behavioral chain breakpoints in the user's consumption behavior evolution process are identified, specifically including:

[0031] The intra-stage behavior density and inter-stage transition time interval of each consumption stage are extracted sequentially along the unified semantic expression chain and used as the benchmark feature sequence for stage continuity discrimination.

[0032] The baseline feature sequence is matched with a preset standard consumption stage evolution pattern sequence using a sliding window, and the continuity deviation index of each consumption stage is determined based on the matching deviation.

[0033] A segment consisting of multiple consecutive consumption stages whose continuity deviation index exceeds the tolerance threshold of consumption evolution is marked as an abnormal decision segment, and the consumption stage with the most severe deviation within the abnormal decision segment is located as the abnormal anchor point.

[0034] Starting from the abnormal anchor point, the relationship break position is searched forward and backward along the unified semantic expression chain. The interval between adjacent consumption stages where the inheritance relationship is missing at the break position is marked as the behavior chain break point.

[0035] In this embodiment, the user's cross-channel behavior path is dynamically completed and corrected based on the decision-making anomaly segment and the behavior chain breakpoint, resulting in the user's cross-channel decision trajectory, specifically including:

[0036] Based on the consumption stage range spanned by the decision anomaly segment, the upstream stable behavior sub-chain and the downstream recovery behavior sub-chain of the decision anomaly segment are extracted from the unified semantic expression chain.

[0037] Based on the stage interval type of the behavior chain breakpoint, a path completion template that connects with the exit semantic fingerprint of the upstream stable behavior sub-chain is matched from the preset consumer behavior path template library.

[0038] The completion behavior node sequence in the path completion template is injected into the decision anomaly segment, and the stage transition edges at the injection boundary are overlapped and rhythmically smoothed to obtain the corrected behavior subchain.

[0039] The modified behavior subchain is concatenated with the upstream stable behavior subchain and the downstream recovery behavior subchain to obtain the user's cross-channel decision trajectory.

[0040] In this embodiment, extracting the upstream stable behavior subchain and the downstream recovery behavior subchain of the decision-making anomaly segment from the unified semantic expression chain based on the consumption stage range spanned by the decision-making anomaly segment specifically includes:

[0041] Obtain the starting boundary consumption stage and the ending boundary consumption stage of the decision anomaly segment, and take the previous adjacent consumption stage of the starting boundary consumption stage as the upstream cutoff anchor point and the next adjacent consumption stage of the ending boundary consumption stage as the downstream cutoff anchor point.

[0042] Locate the termination position of the behavior node sequence corresponding to the upstream truncation anchor point in the unified semantic expression chain, and extract the behavior node sequence from the start node to the termination position in the unified semantic expression chain as an upstream stable behavior sub-chain;

[0043] Locate the starting position of the behavior node sequence corresponding to the downstream truncation anchor point in the unified semantic expression chain, and extract the behavior node sequence from the starting position to the end node in the unified semantic expression chain as the downstream recovery behavior sub-chain.

[0044] In this embodiment, generating a user's digital marketing strategy at the current decision-making stage through the cross-channel decision trajectory specifically includes:

[0045] Locate the user's current decision-making stage node from the cross-channel decision-making trajectory, and extract the stage attribute features and channel reach fingerprint of the decision-making stage node;

[0046] Based on the stage attribute characteristics, the marketing intervention target type corresponding to the current decision-making stage is determined. The marketing intervention target type includes cognitive arousal type, intention boosting type, or conversion facilitation type.

[0047] Based on the channel reach fingerprint and the marketing intervention target type, a combination of candidate strategy elements that is suitable for the user's current decision-making stage is matched from the channel-content joint strategy library;

[0048] The candidate strategy combinations are subjected to cross-channel outreach conflict elimination, duplicate outreach filtering, and outreach timing orchestration to generate a digital marketing strategy for the user at the current decision-making stage.

[0049] Secondly, this application provides a digital marketing strategy generation system that integrates multi-channel data, used to execute a digital marketing strategy generation method that integrates multi-channel data, the digital marketing strategy generation system comprising:

[0050] The event segmentation processing module is used to segment user behavior data in social content channels, e-commerce transaction channels and private domain interaction channels to obtain heterogeneous behavior sequences of users in different channels.

[0051] The semantic association mapping module is used to perform semantic association mapping on user consumption behavior fragments in different channels based on the behavioral context in each heterogeneous behavior sequence, identify the consumption stage association relationship between behaviors in different channels, and construct a unified semantic expression chain of consumption behavior in different channels based on the consumption stage association relationship.

[0052] The phase continuity verification module is used to perform phase continuity verification on the evolution of user consumption behavior in different channels based on the unified semantic expression chain, and to identify abnormal decision-making segments and behavioral chain breakpoints in the evolution of user consumption behavior.

[0053] The dynamic path completion and correction module is used to dynamically complete and correct the user's cross-channel behavior path based on the decision anomaly segment and the behavior chain breakpoint, so as to obtain the user's cross-channel decision trajectory.

[0054] The digital marketing strategy generation module is used to generate a digital marketing strategy for the user at the current decision-making stage based on the cross-channel decision trajectory.

[0055] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0056] User behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels is segmented into event data to obtain heterogeneous behavior sequences across different channels. Semantic association mapping is performed on user consumption behavior fragments across different channels based on the behavioral context of each heterogeneous behavior sequence, identifying the consumption stage relationships between behaviors across different channels. A unified semantic expression chain for consumption behavior across different channels is constructed based on these consumption stage relationships. The evolution of user consumption behavior across different channels is verified for stage continuity based on this unified semantic expression chain, identifying decision-making anomalies and behavioral chain breakpoints during the evolution of user consumption behavior. The user's cross-channel behavior path is dynamically completed and corrected based on these decision-making anomalies and behavioral chain breakpoints to obtain the user's cross-channel decision trajectory. A digital marketing strategy for the user at the current decision-making stage is generated using this cross-channel decision trajectory.

[0057] Therefore, this application demonstrates that a user's digital marketing strategy at the current decision-making stage can be generated through the cross-channel decision trajectory. Firstly, by segmenting user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels, heterogeneous behavior sequences from different channels are obtained. This allows for the structured division of discrete and disordered behavioral events across different channels according to the behavior occurrence process, improving the analyzability and stage identification capabilities of multi-channel behavioral data and avoiding the problem of difficulty in continuously identifying user behavior trajectories due to differences in behavioral record granularity. Secondly, based on the behavioral context in each heterogeneous behavior sequence, semantic association mapping is performed on consumption behavior fragments in different channels, and the consumption stage correlation between behaviors in different channels is identified. This constructs a unified semantic expression chain, overcoming the semantic isolation problem caused by differences in behavioral labeling systems and data expression methods between different channels, enabling users' browsing, transaction, and interaction behaviors across different channels to form a unified consumption stage. Semantic association enhances the implicit semantic alignment of user behavior across channels. Furthermore, based on a unified semantic expression chain, it verifies the continuity of user consumption behavior evolution at each stage and identifies decision-making anomalies and behavioral chain breakpoints. This allows for the detection of stage omissions, behavioral jumps, and semantic breaks in user cross-channel behavior trajectories, improving the continuity and completeness of user consumption decision-making process identification and reducing user profile bias and inaccurate marketing timing judgments caused by incomplete behavioral chains in existing technologies. Finally, it dynamically completes and corrects user cross-channel behavior paths based on decision-making anomalies and behavioral chain breakpoints, obtaining the user's cross-channel decision trajectory. Based on this trajectory, it generates a digital marketing strategy for the user's current decision-making stage, more accurately reconstructing the user's true decision-making evolution at different consumption stages. This improves the matching accuracy between marketing content and the user's current decision intent, thereby enhancing the stage adaptability and marketing precision in the digital marketing strategy generation process.

[0058] In summary, the technical solution adopted in this application can achieve implicit semantic alignment of heterogeneous user behaviors across multiple channels, thereby accurately restoring the complete decision-making process of users across channels. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this embodiment of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is an exemplary flowchart of a digital marketing strategy generation method that integrates multi-channel data according to the present application;

[0061] Figure 2 This is based on the phase continuity verification data analysis diagram provided in this application;

[0062] Figure 3 The comparison chart is generated based on the digital marketing strategy provided in this application.

[0063] Figure 4 This is a schematic diagram illustrating the application scenario of the digital marketing strategy generation system provided in this application;

[0064] Figure 5 This is a module structure diagram of a digital marketing strategy generation system that integrates multi-channel data, provided in this application. Detailed Implementation

[0065] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0066] This application provides a method and system for generating digital marketing strategies that integrates multi-channel data. Its core is to segment user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels to obtain heterogeneous behavior sequences across different channels. Based on the behavioral context of each heterogeneous behavior sequence, semantic association mapping is performed on the user's consumption behavior segments across different channels to identify the consumption stage relationships between behaviors on different channels. A unified semantic expression chain for consumption behavior across different channels is constructed based on these relationships. The evolution of user consumption behavior across different channels is verified for stage continuity based on this unified semantic expression chain, identifying decision-making anomalies and behavioral chain breakpoints. The user's cross-channel behavior path is dynamically completed and corrected based on these decision-making anomalies and behavioral chain breakpoints to obtain the user's cross-channel decision trajectory. Finally, a digital marketing strategy for the user at the current decision-making stage is generated using this cross-channel decision trajectory.

[0067] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1 As shown in the figure, this is an exemplary flowchart of a digital marketing strategy generation method that integrates multi-channel data according to this embodiment of the application. The digital marketing strategy generation method includes the following steps:

[0068] In step S1, the user's behavioral data in social content channels, e-commerce transaction channels, and private domain interaction channels are processed by event segmentation to obtain heterogeneous behavioral sequences of the user in different channels.

[0069] In this embodiment, the heterogeneous behavior sequences of users across social content channels, e-commerce transaction channels, and private domain interaction channels are segmented into events to obtain the following steps:

[0070] Acquire user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels;

[0071] The behavioral data within each channel is sorted chronologically to obtain the original behavioral time series for each channel.

[0072] The original behavior time series of each channel is segmented based on the time interval between adjacent behaviors and the changes in event type in each channel, resulting in multiple heterogeneous behavior segments within each channel.

[0073] The heterogeneous behavior fragments within each channel are organized in chronological order to form a heterogeneous behavior sequence for that channel, thereby obtaining the heterogeneous behavior sequence of users across different channels.

[0074] It should be noted that obtaining user behavior data requires the user's consent, and the data collection, retrieval, and processing processes must comply with relevant data security, personal information protection, and platform authorization rules.

[0075] It should also be noted that the behavioral data mentioned in this application refers to user activity records such as browsing, clicking, favoriting, adding to cart, placing orders, commenting, consulting, sharing, and conversational interaction generated in social content channels, e-commerce transaction channels, and private domain interaction channels; the original behavioral time series refers to a set of behavioral records formed by arranging behavioral data within the same channel in chronological order of the occurrence of the behavior; the event type change refers to the change in event categories such as browsing, interaction, transaction, and private domain communication between adjacent behaviors; the heterogeneous behavioral fragments refer to behavioral data fragments with the same or similar consumption intentions within a continuous time range within the same channel; and the heterogeneous behavioral sequence refers to a sequence of behavioral fragments formed by arranging multiple heterogeneous behavioral fragments within the same channel in chronological order.

[0076] In specific implementation, firstly, the system reads fields such as user identifier, channel source, behavior occurrence time, event type, behavior object, and behavior result status through the social content platform interface, e-commerce transaction platform order interface, and private domain interaction system log interface, respectively. Records missing user identifiers or behavior occurrence times are removed, and the remaining structured records are used as user behavior data in social content channels, e-commerce transaction channels, and private domain interaction channels. Secondly, the behavior data is grouped according to the channel source field, and within each channel group, the behavior data is arranged in ascending order using timestamps. The sorted set of behavior records is then used as the original behavior time series for each channel. Next, adjacent behavior records are read one by one along the original behavior time series of each channel, and segmentation is performed using a fixed time threshold segmentation method and an event type change segmentation method. The process involves several segments, where a fixed time threshold can be set based on the historical average of the time intervals between adjacent user behaviors within the channel. An event type change segmentation method is used to set segment boundaries when adjacent behaviors change from content browsing to product clicking, from product clicking to adding to cart and placing an order, or from private domain inquiries to transaction feedback. The continuous set of behavior records between two adjacent segment boundaries is then segmented as heterogeneous behavior fragments within the channel. Finally, the multiple heterogeneous behavior fragments obtained from each channel are arranged in ascending order according to their start time, and each heterogeneous behavior fragment is written with a channel tag, fragment start and end times, event type set, and behavior object set. The arranged set of heterogeneous behavior fragments is then organized into a heterogeneous behavior sequence for the channel. Furthermore, the heterogeneous behavior sequences formed by social content channels, e-commerce transaction channels, and private domain interaction channels are used as heterogeneous behavior sequences for users on different channels.

[0077] In step S2, semantic association mapping is performed on the user's consumption behavior segments in different channels based on the behavioral context in each heterogeneous behavior sequence, the consumption stage association relationship between different channel behaviors is identified, and a unified semantic expression chain of consumption behavior in different channels is constructed based on the consumption stage association relationship.

[0078] In this embodiment, the semantic association mapping of user consumption behavior segments in different channels based on the behavioral context in each heterogeneous behavior sequence, and the identification of the consumption stage association relationship between behaviors in different channels, can be achieved through the following steps:

[0079] Determine the behavioral context in each heterogeneous behavioral sequence, wherein the behavioral context includes a triggering intent identifier, a behavioral object identifier, and a behavioral result status identifier;

[0080] A semantic association graph between cross-channel behaviors is constructed based on the semantic homology coefficient and temporal adjacency coefficient between the contexts of each behavior.

[0081] For each association edge in the semantic association graph, calculate the consumption stage affiliation degree, and mark the association edges with affiliation degrees exceeding the stage association threshold as consumption stage association edges;

[0082] Based on the transitive and mutually exclusive constraints between all consumption stage association edges, the association consistency is verified to obtain the consumption stage association relationship between different channel behaviors.

[0083] It should be noted that, in this application, the behavioral context refers to the contextual information used to characterize the cause, target, and execution result of a consumer behavior segment; the triggering intent identifier indicates the user intent category corresponding to the consumer behavior segment; the behavioral object identifier indicates the product, content, store, or service object to which the consumer behavior segment targets; the behavioral result status identifier indicates the status formed after the completion of the consumer behavior segment, such as browsing, favorites, adding to cart, placing an order, inquiring, or not converting; the semantic homology coefficient indicates the degree of consistency between consumer behavior segments from different channels in terms of behavioral object and triggering intent; the temporal adjacency coefficient indicates the degree of proximity of consumer behavior segments from different channels in terms of occurrence time; the semantic association graph represents a graph structure with consumer behavior segments as nodes and cross-channel semantic association relationships as edges; and the consumption stage association edges represent association edges that can reflect that behaviors from different channels are in the same or adjacent consumption stages.

[0084] In specific implementation, firstly, consumption behavior segments are read one by one from each heterogeneous behavior sequence. Triggering intent identifiers are obtained by mapping the event type field in the consumption behavior segment. Behavior object identifiers are extracted based on product number, content number, store number, or session topic number. Behavior result status identifiers are extracted based on the behavior completion status field. The triggering intent identifiers, behavior object identifiers, and behavior result status identifiers are combined to form the behavior context in each heterogeneous behavior sequence. Secondly, a text vector matching method is used to vectorize the triggering intent identifiers and behavior object identifiers in different behavior contexts. The vector similarity result is used as a semantic homology coefficient, and corresponding adjacency is assigned according to the time interval between the occurrence times of two consumption behavior segments. The adjacency level is used as the temporal adjacency coefficient, and the consumer behavior segments are used as graph nodes. Linking edges are established between cross-channel behavior segments whose semantic homology coefficient and temporal adjacency coefficient both satisfy preset connection conditions. The graph structure after establishing these links is used as the semantic association graph between cross-channel behaviors. Next, each link in the semantic association graph is assigned a category according to a consumption stage rule table. This consumption stage rule table records the correspondence between event types such as browsing, favorites, adding to cart, inquiries, placing orders, and repeat purchases, and the cognitive stage, intention stage, conversion stage, and repeat purchase stage. The consumption stage affiliation degree of the link is calculated by combining the semantic homology coefficient, temporal adjacency coefficient, and behavior result status matching. The consumption stage affiliation degree can be determined using the following formula: ,in, Indicates the first The first consumer behavior segment and the first The degree of consumption stage affiliation between individual consumption behavior segments is used to determine whether two consumption behavior segments can be classified into the same consumption stage or adjacent consumption stages. Indicates the first The first consumer behavior segment and the first The semantic homology coefficient between consumer behavior segments is used to characterize the degree of consistency between two consumer behavior segments in terms of triggering intent identifier and behavior object identifier; Indicates the first The first consumer behavior segment and the first The temporal adjacency coefficient between consumer behavior segments is used to characterize the proximity of two consumer behavior segments in terms of their occurrence time. Indicates the first The first consumer behavior segment and the first The matching coefficient of behavioral outcome status between consumer behavior segments is used to characterize the degree of consistency in the stage progression of two consumer behavior segments in terms of behavioral outcome status such as browsing, favorites, adding to cart, consultation, and placing an order. , , represent the weight parameters corresponding to the semantic homology coefficient, temporal adjacency coefficient, and behavior result state matching coefficient, respectively. After calculating the consumption stage attribution degree, the association edges whose consumption stage attribution degree exceeds the stage association threshold are marked as consumption stage association edges. The stage association threshold is set using the statistical quantile value of the consumption stage attribution degree in the historical consumption behavior samples. Finally, the association consistency is checked based on the transitive and mutually exclusive constraints between all consumption stage association edges to obtain the consumption stage association relationship between different channel behaviors.

[0085] In this embodiment, the association consistency verification based on the transitive and mutually exclusive constraints between all consumption stage association edges can be achieved by the following steps to obtain the consumption stage association relationships between different channel behaviors:

[0086] Traverse all consumption stage association edges and extract the channel label and consumption stage label of each heterogeneous behavior fragment connected by each consumption stage association edge;

[0087] Based on the extracted channel tags and consumption stage tags, the transitive constraints between different channel tags under the same consumption stage tag and the mutual exclusion constraints between different consumption stage tags are identified.

[0088] Based on the transitive constraints, the connected components of the consumption stage association edges are aggregated, and the cross-stage conflict elimination is performed on the aggregated connected components based on the mutual exclusion constraints. The consumption stage association edges retained after eliminating conflicts are used as the consumption stage association relationships between different channel behaviors.

[0089] It should be noted that, in this application, the channel label represents the data source channel to which the heterogeneous behavior segment belongs; the consumption stage label represents the consumption stage category corresponding to the heterogeneous behavior segment; the transitive constraint relationship represents the continuous merging connection relationship between different channel behavior segments under the same consumption stage label; the mutual exclusion constraint relationship represents the exclusion relationship that different consumption stage labels cannot simultaneously belong to the same consumption stage association result; the connected component represents the set of consumption stage association edges formed by the transitive constraint relationship; the cross-stage conflict represents the conflict state where the same connected component simultaneously contains mutually exclusive consumption stage labels; and the consumption stage association relationship represents the association relationship connecting different channel consumption behavior stages.

[0090] In specific implementation, firstly, all consumption stage-related edges are read one by one, and the channel source field and consumption stage tag field are read from the heterogeneous behavior segments at both ends of each consumption stage-related edge. The read channel source field is used as the channel tag, and the read consumption stage tag field is used as the consumption stage tag. Secondly, the consumption stage-related edges are grouped according to the consumption stage tags. Within the group corresponding to the same consumption stage tag, the connection relationship between different channel tags is read using a graph traversal method. Cross-channel connection relationships that can be continuously reached through one or more consumption stage-related edges are used as transitive constraints. At the same time, the order relationship between different consumption stage tags is read according to a preset consumption stage order table. Stage relationships that do not conform to the consumption stage order table but are connected to the same behavior segment are used as mutual exclusion constraints. Finally, a connected component search algorithm is used to aggregate the consumption stage-related edges that satisfy the transitive constraints. The aggregated edge set is used as a connected component, and the mutual exclusion constraints are compared one by one in each connected component. Consumption stage-related edges with cross-stage conflicts are removed, and the consumption stage-related edges retained after removing conflicts are used as consumption stage-related relationships between different channel behaviors.

[0091] In this embodiment, constructing a unified semantic expression chain of consumption behavior across different channels based on the aforementioned consumption stage association can be achieved through the following steps:

[0092] Based on the aforementioned consumption stage association, heterogeneous behavioral fragments from different channels are aggregated into consumption stage behavior clusters, with each consumption stage behavior cluster corresponding to a consumption stage state node.

[0093] For each consumption stage behavior cluster, the heterogeneous behavior fragments are sorted by the time of behavior occurrence, and stage state transition edges are inserted between adjacent behavior nodes to form a local semantic subchain for that consumption stage.

[0094] Extract the entry behavior semantic fingerprint and exit behavior semantic fingerprint of each local semantic subchain, and determine the succession relationship between different consumption stages based on the entry behavior semantic fingerprint and the exit behavior semantic fingerprint;

[0095] By globally connecting all consumption stages according to the aforementioned sequential relationship, a unified semantic expression chain of consumption behavior across different channels is constructed.

[0096] It should be noted that, in this application, the consumption stage behavior cluster refers to a set of multiple heterogeneous behavior fragments connected by the same consumption stage association; the consumption stage state node refers to a stage node used to carry cross-channel consumption behavior within the same consumption stage; the stage state transition edge refers to a connection edge used to connect adjacent behavior nodes and characterize the direction of consumption behavior advancement; the local semantic sub-chain refers to a behavior chain organized according to the time of behavior occurrence within the same consumption stage; the entry behavior semantic fingerprint refers to the semantic identifier of the starting behavior node of the local semantic sub-chain; the exit behavior semantic fingerprint refers to the semantic identifier of the ending behavior node of the local semantic sub-chain; the preceding and following inheritance relationship refers to the stage connection relationship formed between different consumption stages according to the semantic connection order; and the unified semantic expression chain refers to a chain structure used to uniformly express the evolution order of consumption behavior across different channels.

[0097] In specific implementation, firstly, using the consumption stage tags in the consumption stage association relationship as the aggregation basis, heterogeneous behavior fragments with the same consumption stage tag and connected through the consumption stage association relationship are grouped into the same group using key-value grouping. The set of heterogeneous behavior fragments in each group is taken as a consumption stage behavior cluster, and the corresponding consumption stage tag, channel tag set, and behavior fragment set are written into a node record, which is then used as a consumption stage state node. Secondly, within each consumption stage behavior cluster, the occurrence time of each heterogeneous behavior fragment is read, and the heterogeneous behavior fragments are sorted in ascending order using timestamps. Each sorted heterogeneous behavior fragment is taken as a behavior node, and stage state transition edges are written between adjacent behavior nodes. The ordered chain structure composed of behavior nodes and stage state transition edges is taken as the local semantic sub-chain of that consumption stage. Then... The process involves reading the first and last behavior nodes in each local semantic subchain, combining the trigger intent identifier, behavior object identifier, and behavior result status identifier in the first behavior node to form an entry behavior semantic fingerprint, and combining the trigger intent identifier, behavior object identifier, and behavior result status identifier in the last behavior node to form an exit behavior semantic fingerprint. A fingerprint field matching method is used to compare the exit behavior semantic fingerprints and entry behavior semantic fingerprints of different local semantic subchains. Two consumption stages that can connect the exit behavior semantic fingerprint to the entry behavior semantic fingerprint are recorded as a before-after inheritance relationship. Finally, all consumption stage status nodes are sorted according to the before-after inheritance relationship, and inter-stage connection edges are written between adjacent consumption stage status nodes with before-after inheritance relationships. The chain structure formed by connecting all local semantic subchains and their inter-stage connection edges serves as a unified semantic expression chain for consumption behavior in different channels.

[0098] In step S3, the continuity of the user's consumption behavior evolution process in different channels is verified based on the unified semantic expression chain, and the abnormal decision-making segments and behavior chain breakpoints in the user's consumption behavior evolution process are identified.

[0099] In this embodiment, the stage continuity verification of the user's consumption behavior evolution process in different channels based on the unified semantic expression chain, and the identification of abnormal decision-making segments and behavioral chain breakpoints in the user's consumption behavior evolution process, can be achieved through the following steps:

[0100] The intra-stage behavior density and inter-stage transition time interval of each consumption stage are extracted sequentially along the unified semantic expression chain and used as the benchmark feature sequence for stage continuity discrimination.

[0101] The baseline feature sequence is matched with a preset standard consumption stage evolution pattern sequence using a sliding window, and the continuity deviation index of each consumption stage is determined based on the matching deviation.

[0102] A segment consisting of multiple consecutive consumption stages whose continuity deviation index exceeds the tolerance threshold of consumption evolution is marked as an abnormal decision segment, and the consumption stage with the most severe deviation within the abnormal decision segment is located as the abnormal anchor point.

[0103] Starting from the abnormal anchor point, the relationship break position is searched forward and backward along the unified semantic expression chain. The interval between adjacent consumption stages where the inheritance relationship is missing at the break position is marked as the behavior chain break point.

[0104] It should be noted that, in this application, the intra-stage behavior density refers to the concentration of behavior nodes within the same consumption stage within a unit time range; the inter-stage transition time interval refers to the time span between adjacent consumption stages from the end behavior node of the previous consumption stage to the entry behavior node of the next consumption stage; the benchmark feature sequence refers to the combined sequence of intra-stage behavior density and inter-stage transition time interval arranged in the order of consumption stages; the standard consumption stage evolution pattern sequence refers to the stage advancement reference sequence pre-organized from normal consumption behavior samples; the continuity deviation index refers to the degree of deviation of the current consumption stage from the standard consumption stage evolution pattern; the decision anomaly segment refers to the continuous consumption stage segment where the continuity deviation index exceeds the consumption evolution tolerance threshold; the anomaly anchor point refers to the consumption stage with the largest deviation in the decision anomaly segment; and the behavior chain breakpoint refers to the break position where the inheritance relationship between adjacent consumption stages is missing in the unified semantic expression chain.

[0105] In specific implementation, firstly, the consumption stage state nodes are read sequentially along the consumption stage order in the unified semantic expression chain. The number of behavior nodes contained in each consumption stage state node and the duration from the entry behavior node to the exit behavior node of that consumption stage are counted. The density formed by the number of behavior nodes and the duration is taken as the behavior density within the stage. Simultaneously, the stage state transition edges between two adjacent consumption stages are read. The time span between the exit behavior node of the previous consumption stage and the entry behavior node of the next consumption stage is taken as the inter-stage transition time interval. Finally, the intra-stage behavior density and inter-stage transition time interval corresponding to each consumption stage are arranged in the order of the consumption stages. The sequence is used as a benchmark feature sequence for determining the continuity of consumption stages. Next, a sliding window matching method is used to match the benchmark feature sequence with a preset standard consumption stage evolution pattern sequence segment by segment. The sliding window length is set according to three consecutive consumption stages to cover the smallest continuous evolutionary segment between cognition, intention, and conversion. The behavior density level within each stage and the transition time interval level between stages are compared window by window. After obtaining the window matching results, the continuity deviation index of each consumption stage can be determined by combining the deviation of behavior density within each stage, the deviation of transition time between stages, and the absence of stage inheritance relationships. The continuity deviation index can be determined using the following formula: ,in, Indicates the first The continuity deviation index of each consumption stage is used to characterize the degree of abnormal deviation of the current consumption stage from the standard consumption stage evolution pattern. Indicates the first Behavioral density within each consumption stage; This indicates the first stage of the standard consumption evolution model. Reference behavior density corresponding to each consumption stage; Indicates the first The inter-stage transition time between a consumption stage and its adjacent preceding consumption stage; This indicates the first stage of the standard consumption evolution model. Reference transition time intervals corresponding to each consumption stage; Indicates the first The stage transition relationship of the consumption stage is missing a marker value when the first consumption stage is missing a marker value. The value is 1 when there is no inheritance relationship between the first consumption stage and its adjacent consumption stages; the value is 1 when the first consumption stage is missing an inheritance relationship. The value is 0 when there is a successor-inheritance relationship between a consumption stage and an adjacent consumption stage. , , These represent the weight parameters corresponding to behavioral density deviation, transition time deviation, and missing stage relationship, respectively. This represents a smoothing parameter used to avoid a denominator of zero. Next, a threshold marking method is used to read the continuity deviation index one by one. Consumption stages exceeding the consumption evolution tolerance threshold are marked as abnormal stages, and adjacent and consecutive abnormal stages are merged into the same segment. The merged segment is used as the decision abnormal segment. The consumption evolution tolerance threshold can be set using the statistical quantile value of the continuity deviation index in historical normal consumption behavior samples. Then, within the decision abnormal segment, the continuity deviation index corresponding to each abnormal stage is read, and the consumption stage with the largest continuity deviation index is used as the abnormal anchor point. Finally, starting from the abnormal anchor point, the stage state transition edges and previous / next inheritance relationship records between adjacent consumption stages are read forward and backward along the unified semantic expression chain. An edge existence check method is used to check whether adjacent consumption stages simultaneously have stage state transition edges and previous / next inheritance relationship records. The interval between adjacent consumption stages lacking previous / next inheritance relationship records is marked as a behavior chain breakpoint.

[0106] In this embodiment, to verify the reliability of the stage continuity check and behavior chain breakpoint location, historical cross-channel consumption behavior samples that have been anonymized can be used for data verification. The verification samples are divided into four groups according to the number of user paths: 200, 500, 1000, and 2000. Each user path includes behavior records from social content channels, e-commerce transaction channels, and private domain interaction channels, and the results of manual review are used as reference annotations for stage association, abnormal segments, and behavior chain breakpoints.

[0107]

[0108] As shown in Table 1, as the number of user paths increased from 200 to 2000, the accuracy of stage association improved from 88.6% to 94.1%, the recall rate of continuity verification improved from 84.8% to 92.4%, and the accuracy of behavior chain breakpoint location improved from 82.5% to 91.1%. This indicates that by jointly verifying the behavior density within a stage, the transition time interval between stages, and the standard consumption stage evolution pattern sequence, abnormal progress states in the consumption evolution process can be identified relatively stably. The average processing time increased from 41ms to 132ms, with the increase being basically consistent with the sample size, indicating that the method can still maintain controllable data processing overhead when multi-channel behavior records increase, making it suitable for use as a path quality verification step before digital marketing strategy generation. (Reference) Figure 2As shown in the figure, this figure is a data analysis diagram of stage continuity verification provided in this embodiment of the present application. In this figure, the stage association accuracy, continuity verification recall rate and behavior chain breakpoint location accuracy all show a steady upward trend with the increase of the verification sample size, and the three indicators tend to stabilize after 1000 user paths. This indicates that the unified semantic expression chain can reduce the isolated judgment between behavior segments of different channels, so that the abnormal segment identification results do not depend on the behavior intensity of a single channel, thereby improving the data credibility of subsequent path completion and path correction.

[0109] In step S4, the user's cross-channel behavior path is dynamically completed and corrected based on the decision anomaly segment and the behavior chain breakpoint to obtain the user's cross-channel decision trajectory.

[0110] In this embodiment, the user's cross-channel behavior path is dynamically completed and corrected based on the decision anomaly segment and the behavior chain breakpoint to obtain the user's cross-channel decision trajectory. This can be achieved through the following steps:

[0111] Based on the consumption stage range spanned by the decision anomaly segment, the upstream stable behavior sub-chain and the downstream recovery behavior sub-chain of the decision anomaly segment are extracted from the unified semantic expression chain.

[0112] Based on the stage interval type of the behavior chain breakpoint, a path completion template that connects with the exit semantic fingerprint of the upstream stable behavior sub-chain is matched from the preset consumer behavior path template library.

[0113] The completion behavior node sequence in the path completion template is injected into the decision anomaly segment, and the stage transition edges at the injection boundary are overlapped and rhythmically smoothed to obtain the corrected behavior subchain.

[0114] The modified behavior subchain is concatenated with the upstream stable behavior subchain and the downstream recovery behavior subchain to obtain the user's cross-channel decision trajectory.

[0115] It should be noted that, in this application, the stage interval type refers to the stage missing type between adjacent consumption stages on both sides of the behavior chain breakpoint; the consumption behavior path template library refers to the template set of pre-stored completion paths corresponding to different stage interval types; the path completion template refers to the reference path structure used to fill in the missing consumption behavior nodes at the behavior chain breakpoint; the completion behavior node sequence refers to the set of behavior nodes provided by the path completion template and used to inject decision abnormal segments; the corrected behavior sub-chain refers to the abnormal segment replacement chain formed after completion, deduplication and smoothing; and the cross-channel decision trajectory refers to the complete behavior path used to characterize the user's cross-channel consumption decision advancement process.

[0116] In specific implementation, firstly, based on the consumption stage range spanned by the decision anomaly segment, the upstream stable behavior sub-chain and the downstream recovery behavior sub-chain of the decision anomaly segment are extracted from the unified semantic expression chain; secondly, the consumption stage labels of adjacent consumption stages on both sides of the breakpoint of the behavior chain are read, and the combination result of the previous consumption stage label and the next consumption stage label is taken as the stage interval type. A path template corresponding to the stage interval type is retrieved from a preset consumption behavior path template library using label matching. Then, the exit behavior semantic fingerprint of the upstream stable behavior sub-chain is matched with the entry behavior semantic fingerprint of the path template, and the successfully matched path template is used as the path completion template connected to the exit semantic fingerprint of the upstream stable behavior sub-chain; next, the path completion template... The completed behavior node sequence is inserted into the corresponding position of the decision anomaly segment according to the node sequence order. The entry end of the completed behavior node sequence is connected to the exit end of the upstream stable behavior sub-chain, and the exit end of the completed behavior node sequence is connected to the entry end of the downstream recovery behavior sub-chain. Then, duplicate behavior nodes at the injection boundary are deleted by node identifier comparison. The excessive time span of the state transition edge between adjacent stages is adjusted by time interval limiting. The processed behavior chain is used as the corrected behavior sub-chain. Finally, the upstream stable behavior sub-chain, the corrected behavior sub-chain, and the downstream recovery behavior sub-chain are chained together in the order of the upstream stable behavior sub-chain, the corrected behavior sub-chain, and the downstream recovery behavior sub-chain. The channel label, consumption stage label, behavior occurrence time, and behavior object identifier of each behavior node are retained. The complete chained behavior path obtained by splicing is used as the user's cross-channel decision trajectory.

[0117] In this embodiment, extracting the upstream stable behavior subchain and the downstream recovery behavior subchain of the decision-making anomaly segment from the unified semantic expression chain based on the consumption stage range spanned by the decision-making anomaly segment can be achieved through the following steps:

[0118] Obtain the starting boundary consumption stage and the ending boundary consumption stage of the decision anomaly segment, and take the previous adjacent consumption stage of the starting boundary consumption stage as the upstream cutoff anchor point and the next adjacent consumption stage of the ending boundary consumption stage as the downstream cutoff anchor point.

[0119] Locate the termination position of the behavior node sequence corresponding to the upstream truncation anchor point in the unified semantic expression chain, and extract the behavior node sequence from the start node to the termination position in the unified semantic expression chain as an upstream stable behavior sub-chain;

[0120] Locate the starting position of the behavior node sequence corresponding to the downstream truncation anchor point in the unified semantic expression chain, and extract the behavior node sequence from the starting position to the end node in the unified semantic expression chain as the downstream recovery behavior sub-chain.

[0121] It should be noted that, in this application, the starting boundary consumption stage refers to the consumption stage that first becomes abnormal in the decision-abnormal segment; the ending boundary consumption stage refers to the last abnormal consumption stage in the decision-abnormal segment; the upstream truncation anchor point refers to the consumption stage used to limit the endpoint of the upstream stable behavior subchain; the downstream truncation anchor point refers to the consumption stage used to limit the starting point of the downstream recovery behavior subchain; the behavior node sequence refers to the set of behavior nodes arranged in the unified semantic expression chain according to the behavior occurrence time and stage inheritance relationship; the upstream stable behavior subchain refers to the behavior chain segment that maintains stage continuity before the decision-abnormal segment; and the downstream recovery behavior subchain refers to the behavior chain segment that restores stage continuity after the decision-abnormal segment.

[0122] In specific implementation, firstly, the consumption stage state nodes contained in the decision anomaly segment are read, and the earliest consumption stage state node is selected as the starting boundary consumption stage according to the consumption stage order in the unified semantic expression chain, and the latest consumption stage state node is selected as the ending boundary consumption stage. Then, the previous adjacent consumption stage of the starting boundary consumption stage is read forward along the unified semantic expression chain, and this previous adjacent consumption stage is used as the upstream truncation anchor point. Then, the next adjacent consumption stage of the ending boundary consumption stage is read backward along the unified semantic expression chain, and this next adjacent consumption stage is used as the downstream truncation anchor point. Secondly, the corresponding upstream truncation anchor point is retrieved in the unified semantic expression chain. The system retrieves the local semantic subchain, reads the position marker of the end behavior node in the local semantic subchain, uses this position marker as the termination position of the upstream stable behavior subchain, and extracts the behavior node sequence from the starting node of the unified semantic expression chain to the termination position according to the chain node order, using the extracted behavior node sequence as the upstream stable behavior subchain; finally, it retrieves the local semantic subchain corresponding to the downstream truncated anchor point in the unified semantic expression chain, reads the position marker of the entry behavior node in the local semantic subchain, uses this position marker as the starting position of the downstream recovery behavior subchain, and extracts the behavior node sequence from the starting position to the end node of the unified semantic expression chain according to the chain node order, using the extracted behavior node sequence as the downstream recovery behavior subchain.

[0123] In step S5, the user's digital marketing strategy at the current decision-making stage is generated through the cross-channel decision trajectory.

[0124] In this embodiment, generating a user's digital marketing strategy at the current decision-making stage through the cross-channel decision trajectory can be achieved through the following steps:

[0125] Locate the user's current decision-making stage node from the cross-channel decision-making trajectory, and extract the stage attribute features and channel reach fingerprint of the decision-making stage node;

[0126] Based on the stage attribute characteristics, the marketing intervention target type corresponding to the current decision-making stage is determined. The marketing intervention target type includes cognitive arousal type, intention boosting type, or conversion facilitation type.

[0127] Based on the channel reach fingerprint and the marketing intervention target type, a combination of candidate strategy elements that is suitable for the user's current decision-making stage is matched from the channel-content joint strategy library;

[0128] The candidate strategy combinations are subjected to cross-channel outreach conflict elimination, duplicate outreach filtering, and outreach timing orchestration to generate a digital marketing strategy for the user at the current decision-making stage.

[0129] It should be noted that, in this application, the decision-making stage node refers to a node in the cross-channel decision-making trajectory used to characterize the user's current consumption progress status; the stage attribute features refer to the consumption stage category, behavioral outcome status, and stage dwell status corresponding to the decision-making stage node; the channel reach fingerprint refers to the user's reach preference and recent reach status in different channels; the marketing intervention target type refers to the marketing guidance target to be achieved in the current decision-making stage; the channel-content joint strategy library refers to a set of strategies that are pre-configured to adapt to different channels, consumption stages, and marketing content; the candidate strategy element combination refers to a set of marketing strategy units to be screened obtained by matching from the channel-content joint strategy library; and the digital marketing strategy refers to a combination strategy of channel reach, content delivery, and reach timing generated for the user's current decision-making stage.

[0130] In specific implementation, firstly, the terminal behavior node and its corresponding consumption stage status node are read along the cross-channel decision-making trajectory. This consumption stage status node is taken as the user's current decision-making stage node, and the consumption stage tag, recent behavior result status, and stage dwell time are extracted from this decision-making stage node as stage attribute features. Simultaneously, the channel tag, recent touch time, and touch response status of each behavior node within this decision-making stage node are read, and the combined result of the channel tag, recent touch time, and touch response status is taken as the channel touch fingerprint. Secondly, the stage attribute features are read using a stage rule mapping table. Specifically, the stage attribute features of the cognitive stage lacking effective interaction results are mapped to the cognitive arousal type; the stage attribute features of the intention stage with results such as collection, consultation, and adding to cart are mapped to the intention boosting type; and the stage attribute features of the conversion stage with results such as pre-order dwell or payment interruption are mapped to the conversion type. The process involves several steps: First, a marketing intervention target type is mapped and used as the target type for the current decision-making stage. Second, a strategy unit matching the channel reach fingerprint and the marketing intervention target type is retrieved from the channel-content joint strategy library using field matching. This strategy unit includes the reach channel, content theme, reach time window, and reach frequency limit. The retrieved strategy unit set is used as a candidate strategy element combination adapted to the user's current decision-making stage. Finally, the candidate strategy element combination is processed by using a channel time window conflict check to delete strategy units that repeatedly reach the same user within the same time window, and by using a content theme identifier comparison to delete strategy units with duplicate content themes. The remaining strategy units are then arranged according to the user's most recent reach time, channel response priority, and reach frequency limit. The arranged strategy unit combination is then used as the user's digital marketing strategy for the current decision-making stage.

[0131] In this embodiment, to verify the application effect of the digital marketing strategy generation results, a comparative verification can be set up under the conditions of the same product category, the same campaign period, and the same reach frequency limit. The verification subjects are 1200 historically active users who have undergone anonymization processing. The single-channel rule strategy is used to generate marketing content based on the most recent behavior of a single channel, and the multi-channel uncorrected strategy is used to directly merge multi-channel behaviors to generate marketing content. The method of this application is used to generate a digital marketing strategy after completing cross-channel decision trajectory completion and reach conflict elimination.

[0132]

[0133] As shown in Table 2, compared with the single-channel rule strategy, the effective response rate of the proposed method increased from 10.8% to 18.9%, and the conversion completion rate increased from 3.7% to 6.8%. Compared with the multi-channel unmodified strategy, the duplicate reach rate of the proposed method decreased from 14.2% to 5.8%, and the reach conflict rate decreased from 10.7% to 4.6%. These results indicate that by using cross-channel decision trajectories to locate the user's current decision-making stage and combining this with channel reach fingerprints for candidate strategy element combination screening, the number of times the same user is repeatedly and conflictingly reached across different channels can be reduced, while simultaneously improving the fit between marketing content and the current consumption stage. Although the average generation time of the proposed method is 68ms, which is higher than the control strategy, this time is still within the acceptable range for online strategy generation and can meet the processing requirements for real-time generation of user-level digital marketing strategies. (Reference) Figure 3 As shown in the figure, this figure is a comparison of the generation effect of digital marketing strategies provided in this embodiment of the present application. In this figure, the method of the present application is higher than the control strategy in terms of effective response rate and conversion completion rate, and lower than the control strategy in terms of repeated reach rate and reach conflict rate. This shows that the strategy generation process formed by correcting decision abnormal segments, completing behavioral chain breakpoints, and eliminating cross-channel reach conflicts can improve user conversion effect while suppressing overreach problems, so that the generated digital marketing strategy has higher execution stability and business credibility.

[0134] Additionally, it should be noted that the reference Figure 4 As shown in the figure, this diagram illustrates an application scenario of the digital marketing strategy generation system provided in this embodiment of the application. In this diagram, users generate multi-channel user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels. This multi-channel user behavior data is collected and input into the digital marketing strategy generation system. The digital marketing strategy generation system is used to uniformly process user behavior data from different sources and of different types, and to generate digital marketing strategies adapted to the current decision-making stage based on the evolution of user behavior across different channels. The supporting data and knowledge base shown at the bottom of the diagram provide basic data, rule templates, and strategy matching criteria for system operation, enabling the system to establish stable data support relationships between multi-channel behavior data. After the generated digital marketing strategy is output through the marketing strategy application terminal, it can reach users across social content channels, e-commerce transaction channels, and private domain interaction channels, and the user's response to the reached content is fed back to the multi-channel user behavior data side. Therefore, Figure 4 The overall presentation demonstrates the closed-loop application of this application, encompassing multi-channel user behavior collection, digital marketing strategy generation, cross-channel outreach, and performance feedback. This illustrates the overall application relationship of this application in generating and continuously optimizing marketing strategies for multi-channel consumer behavior scenarios.

[0135] Therefore, this application demonstrates that a user's digital marketing strategy at the current decision-making stage can be generated through the cross-channel decision trajectory. Firstly, by segmenting user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels, heterogeneous behavior sequences from different channels are obtained. This allows for the structured division of discrete and disordered behavioral events across different channels according to the behavior occurrence process, improving the analyzability and stage identification capabilities of multi-channel behavioral data and avoiding the problem of difficulty in continuously identifying user behavior trajectories due to differences in behavioral record granularity. Secondly, based on the behavioral context in each heterogeneous behavior sequence, semantic association mapping is performed on consumption behavior fragments in different channels, and the consumption stage correlation between behaviors in different channels is identified. This constructs a unified semantic expression chain, overcoming the semantic isolation problem caused by differences in behavioral labeling systems and data expression methods between different channels, enabling users' browsing, transaction, and interaction behaviors across different channels to form a unified consumption stage. Semantic association enhances the implicit semantic alignment of user behavior across channels. Furthermore, based on a unified semantic expression chain, it verifies the continuity of user consumption behavior evolution at each stage and identifies decision-making anomalies and behavioral chain breakpoints. This allows for the detection of stage omissions, behavioral jumps, and semantic breaks in user cross-channel behavior trajectories, improving the continuity and completeness of user consumption decision-making process identification and reducing user profile bias and inaccurate marketing timing judgments caused by incomplete behavioral chains in existing technologies. Finally, it dynamically completes and corrects user cross-channel behavior paths based on decision-making anomalies and behavioral chain breakpoints, obtaining the user's cross-channel decision trajectory. Based on this trajectory, it generates a digital marketing strategy for the user's current decision-making stage, more accurately reconstructing the user's true decision-making evolution at different consumption stages. This improves the matching accuracy between marketing content and the user's current decision intent, thereby enhancing the stage adaptability and marketing precision in the digital marketing strategy generation process.

[0136] In summary, the technical solution adopted in this application can achieve implicit semantic alignment of heterogeneous user behaviors across multiple channels, thereby accurately restoring the complete decision-making process of users across channels.

[0137] Example 2: This application provides a digital marketing strategy generation system that integrates multi-channel data, referencing... Figure 5 As shown in the figure, this is a module structure diagram of a digital marketing strategy generation system that integrates multi-channel data according to this embodiment of the application. The digital marketing strategy generation system includes:

[0138] The event segmentation processing module 100 is used to segment user behavior data in social content channels, e-commerce transaction channels and private domain interaction channels to obtain heterogeneous behavior sequences of users in different channels.

[0139] The semantic association mapping module 200 is used to perform semantic association mapping on the consumption behavior fragments of users in different channels according to the behavioral context in each heterogeneous behavior sequence, identify the consumption stage association relationship between behaviors in different channels, and construct a unified semantic expression chain of consumption behavior in different channels according to the consumption stage association relationship.

[0140] The phase continuity verification module 300 is used to perform phase continuity verification on the evolution process of user consumption behavior in different channels based on the unified semantic expression chain, and to identify abnormal decision-making segments and behavioral chain breakpoints in the user's consumption behavior evolution process.

[0141] The dynamic path completion and correction module 400 is used to dynamically complete and correct the user's cross-channel behavior path based on the decision anomaly segment and the behavior chain breakpoint, so as to obtain the user's cross-channel decision trajectory.

[0142] The digital marketing strategy generation module 500 is used to generate a digital marketing strategy for the user at the current decision-making stage based on the cross-channel decision trajectory.

[0143] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0144] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0145] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A method for generating digital marketing strategies that integrates multi-channel data, characterized in that, The method for generating digital marketing strategies includes the following steps: Event segmentation processing is performed on user behavior data in social content channels, e-commerce transaction channels, and private domain interaction channels to obtain heterogeneous behavior sequences of users in different channels; Based on the behavioral context in each heterogeneous behavioral sequence, semantic association mapping is performed on the consumption behavior fragments of users in different channels to identify the consumption stage association relationship between behaviors in different channels, and a unified semantic expression chain of consumption behavior in different channels is constructed based on the consumption stage association relationship. Based on the unified semantic expression chain, the stage continuity of the user's consumption behavior evolution process in different channels is verified, and the abnormal decision-making segments and behavior chain breakpoints of the user in the consumption behavior evolution process are identified. Based on the decision-making anomaly segment and the behavior chain breakpoint, the user's cross-channel behavior path is dynamically completed and the path is corrected to obtain the user's cross-channel decision trajectory. The cross-channel decision-making trajectory is used to generate digital marketing strategies for users at the current decision-making stage.

2. The method for generating digital marketing strategies by integrating multi-channel data as described in claim 1, characterized in that, Event segmentation processing is performed on user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels to obtain heterogeneous user behavior sequences across different channels, specifically including: Acquire user behavior data across social content channels, e-commerce transaction channels, and private domain interaction channels; The behavioral data within each channel is sorted chronologically to obtain the original behavioral time series for each channel. The original behavior time series of each channel is segmented based on the time interval between adjacent behaviors and the changes in event type in each channel, resulting in multiple heterogeneous behavior segments within each channel. The heterogeneous behavior fragments within each channel are organized in chronological order to form a heterogeneous behavior sequence for that channel, thereby obtaining the heterogeneous behavior sequence of users across different channels.

3. The method for generating digital marketing strategies by integrating multi-channel data as described in claim 1, characterized in that, Based on the behavioral context in each heterogeneous behavioral sequence, semantic association mapping is performed on user consumption behavior segments across different channels to identify the consumption stage relationships between behaviors across different channels. Specifically, this includes: Determine the behavioral context in each heterogeneous behavioral sequence, wherein the behavioral context includes a triggering intent identifier, a behavioral object identifier, and a behavioral result status identifier; A semantic association graph between cross-channel behaviors is constructed based on the semantic homology coefficient and temporal adjacency coefficient between the contexts of each behavior. For each association edge in the semantic association graph, calculate the consumption stage affiliation degree, and mark the association edges with affiliation degrees exceeding the stage association threshold as consumption stage association edges; Based on the transitive and mutually exclusive constraints between all consumption stage association edges, the association consistency is verified to obtain the consumption stage association relationship between different channel behaviors.

4. The method for generating digital marketing strategies by integrating multi-channel data as described in claim 3, characterized in that, Based on the transitive and mutual exclusion constraints between all consumption stage association edges, a consistency check is performed to obtain the specific consumption stage association relationships between different channel behaviors, including: Traverse all consumption stage association edges and extract the channel label and consumption stage label of each heterogeneous behavior fragment connected by each consumption stage association edge; Based on the extracted channel tags and consumption stage tags, the transitive constraints between different channel tags under the same consumption stage tag and the mutual exclusion constraints between different consumption stage tags are identified. Based on the transitive constraints, the connected components of the consumption stage association edges are aggregated, and the cross-stage conflict elimination is performed on the aggregated connected components based on the mutual exclusion constraints. The consumption stage association edges retained after eliminating conflicts are used as the consumption stage association relationships between different channel behaviors.

5. The method for generating a digital marketing strategy that integrates multi-channel data as described in claim 1, characterized in that, Constructing a unified semantic expression chain for consumption behavior across different channels based on the aforementioned consumption stage relationships specifically includes: Based on the aforementioned consumption stage association, heterogeneous behavioral fragments from different channels are aggregated into consumption stage behavior clusters, with each consumption stage behavior cluster corresponding to a consumption stage state node. For each consumption stage behavior cluster, the heterogeneous behavior fragments are sorted by the time of behavior occurrence, and stage state transition edges are inserted between adjacent behavior nodes to form a local semantic subchain for that consumption stage. Extract the entry behavior semantic fingerprint and exit behavior semantic fingerprint of each local semantic subchain, and determine the succession relationship between different consumption stages based on the entry behavior semantic fingerprint and the exit behavior semantic fingerprint; By globally connecting all consumption stages according to the aforementioned sequential relationship, a unified semantic expression chain of consumption behavior across different channels is constructed.

6. The method for generating a digital marketing strategy that integrates multi-channel data as described in claim 1, characterized in that, Based on the unified semantic expression chain, the evolution of user consumption behavior across different channels is continuously verified in stages, identifying anomalous decision-making segments and behavioral chain breakpoints during the evolution of user consumption behavior. Specifically, these include: The intra-stage behavior density and inter-stage transition time interval of each consumption stage are extracted sequentially along the unified semantic expression chain and used as the benchmark feature sequence for stage continuity discrimination. The baseline feature sequence is matched with a preset standard consumption stage evolution pattern sequence using a sliding window, and the continuity deviation index of each consumption stage is determined based on the matching deviation. A segment consisting of multiple consecutive consumption stages whose continuity deviation index exceeds the tolerance threshold of consumption evolution is marked as an abnormal decision segment, and the consumption stage with the most severe deviation within the abnormal decision segment is located as the abnormal anchor point. Starting from the abnormal anchor point, the relationship break position is searched forward and backward along the unified semantic expression chain. The interval between adjacent consumption stages where the inheritance relationship is missing at the break position is marked as the behavior chain break point.

7. The method for generating a digital marketing strategy that integrates multi-channel data as described in claim 1, characterized in that, Based on the decision-making anomaly segments and the behavioral chain breakpoints, the user's cross-channel behavioral path is dynamically completed and corrected to obtain the user's cross-channel decision-making trajectory, which specifically includes: Based on the consumption stage range spanned by the decision anomaly segment, the upstream stable behavior sub-chain and the downstream recovery behavior sub-chain of the decision anomaly segment are extracted from the unified semantic expression chain. Based on the stage interval type of the behavior chain breakpoint, a path completion template that connects with the exit semantic fingerprint of the upstream stable behavior sub-chain is matched from the preset consumer behavior path template library. The completion behavior node sequence in the path completion template is injected into the decision anomaly segment, and the stage transition edges at the injection boundary are overlapped and rhythmically smoothed to obtain the corrected behavior subchain. The modified behavior subchain is concatenated with the upstream stable behavior subchain and the downstream recovery behavior subchain to obtain the user's cross-channel decision trajectory.

8. The method for generating a digital marketing strategy that integrates multi-channel data as described in claim 7, characterized in that, Based on the consumption stage range spanned by the decision-making anomaly segment, the upstream stable behavior subchain and the downstream recovery behavior subchain of the decision-making anomaly segment are extracted from the unified semantic expression chain, specifically including: Obtain the starting boundary consumption stage and the ending boundary consumption stage of the decision anomaly segment, and take the previous adjacent consumption stage of the starting boundary consumption stage as the upstream cutoff anchor point and the next adjacent consumption stage of the ending boundary consumption stage as the downstream cutoff anchor point. Locate the termination position of the behavior node sequence corresponding to the upstream truncation anchor point in the unified semantic expression chain, and extract the behavior node sequence from the start node to the termination position in the unified semantic expression chain as an upstream stable behavior sub-chain; Locate the starting position of the behavior node sequence corresponding to the downstream truncation anchor point in the unified semantic expression chain, and extract the behavior node sequence from the starting position to the end node in the unified semantic expression chain as the downstream recovery behavior sub-chain.

9. The method for generating a digital marketing strategy that integrates multi-channel data as described in claim 1, characterized in that, Specifically, generating digital marketing strategies for users at the current decision-making stage based on the cross-channel decision trajectory includes: Locate the user's current decision-making stage node from the cross-channel decision-making trajectory, and extract the stage attribute features and channel reach fingerprint of the decision-making stage node; Based on the stage attribute characteristics, the marketing intervention target type corresponding to the current decision-making stage is determined. The marketing intervention target type includes cognitive arousal type, intention boosting type, or conversion facilitation type. Based on the channel reach fingerprint and the marketing intervention target type, a combination of candidate strategy elements that is suitable for the user's current decision-making stage is matched from the channel-content joint strategy library; The candidate strategy combinations are subjected to cross-channel outreach conflict elimination, duplicate outreach filtering, and outreach timing orchestration to generate a digital marketing strategy for the user at the current decision-making stage.

10. A digital marketing strategy generation system integrating multi-channel data, used to execute the digital marketing strategy generation method integrating multi-channel data as described in any one of claims 1 to 9, characterized in that, The digital marketing strategy generation system includes: The event segmentation processing module is used to segment user behavior data in social content channels, e-commerce transaction channels and private domain interaction channels to obtain heterogeneous behavior sequences of users in different channels. The semantic association mapping module is used to perform semantic association mapping on user consumption behavior fragments in different channels based on the behavioral context in each heterogeneous behavior sequence, identify the consumption stage association relationship between behaviors in different channels, and construct a unified semantic expression chain of consumption behavior in different channels based on the consumption stage association relationship. The phase continuity verification module is used to perform phase continuity verification on the evolution of user consumption behavior in different channels based on the unified semantic expression chain, and to identify abnormal decision-making segments and behavioral chain breakpoints in the evolution of user consumption behavior. The dynamic path completion and correction module is used to dynamically complete and correct the user's cross-channel behavior path based on the decision anomaly segment and the behavior chain breakpoint, so as to obtain the user's cross-channel decision trajectory. The digital marketing strategy generation module is used to generate a digital marketing strategy for the user at the current decision-making stage based on the cross-channel decision trajectory.

Images