Advertisement putting method and system based on multi-source data analysis

Abstract

The application belongs to the field of advertisement delivery, and provides an advertisement delivery method and system based on multi-source data analysis, which comprises collecting multi-source original data related to users; identifying multiple cognitive state nodes based on click behavior data and transaction path data; constructing a cognitive behavior causal graph based on the multiple cognitive state nodes, wherein the nodes of the causal graph represent user cognitive states; combining historical behavior samples and path structures in the causal graph, using a Bayesian inference model to predict advertisement response probability and conversion probability of target users, and estimating state transition probability of users from a current state node to a target state node based on different advertisement intervention contents; determining an optimal advertisement intervention path of users from a current cognitive state to an expected conversion state based on the state transition probability and the causal graph structure, and constructing a corresponding advertisement delivery sequence based on the path.

Description

Technical Field

[0001] This invention belongs to the field of advertising placement, and specifically relates to an advertising placement method and system based on multi-source data analysis. Background Technology

[0002] With the development of mobile internet and the widespread adoption of smart devices, advertising technology has gradually evolved from traditional static display methods to intelligent recommendation models based on user profiles and behavioral predictions. In existing technologies, advertising systems generally use user click behavior data, browsing paths, and historical purchase records as the basis for modeling, constructing user interest models through deep learning or logistic regression to achieve content matching and delivery to the target audience.

[0003] However, existing advertising recommendation methods have the following shortcomings:

[0004] First, existing methods typically extract features based on a single dimension or superficial behavior, which cannot fully depict the user's behavioral evolution path and make it difficult to accurately grasp their true cognitive state and potential conversion intentions.

[0005] Secondly, most methods only infer the connection between advertising content and user behavior through statistical correlation or machine learning prediction, lacking modeling of the causal mechanism of whether advertising intervention truly causes cognitive change. This results in the inability of the placement strategy to effectively intervene in the user's psychological path, thus affecting the overall conversion efficiency.

[0006] Furthermore, traditional advertising systems lack path-based strategic planning capabilities and are unable to formulate coherent and effective multi-stage intervention strategies from a holistic cognitive evolution perspective. This results in fragmented, repetitive, and stage-adaptive advertising content delivery, leading to wasted budgets and a decline in user experience. Summary of the Invention

[0007] To address the problems in the existing technology, this invention provides an advertising delivery method based on multi-source data analysis, comprising the following steps:

[0008] Collect multi-source raw data related to users, including at least user click behavior data and transaction path data, to characterize user behavior trajectory and purchase intention;

[0009] Based on the click behavior data and transaction path data, multiple cognitive state nodes are identified. These cognitive state nodes are used to characterize the user's psychological state or behavioral stage during the advertising interaction process.

[0010] A cognitive behavior causal graph is constructed based on the multiple cognitive state nodes. The nodes of the causal graph represent the user's cognitive state, and the directed edges of the causal graph represent the causal transfer relationship between different advertising contents and the user's cognitive state.

[0011] By combining historical behavior samples with the path structure in the causal graph, a Bayesian inference model is used to predict the advertising response probability and conversion probability of the target user, and the state transition probability of the user from the current state node to the target state node is estimated based on different advertising intervention content.

[0012] Based on the state transition probability and causal graph structure, the optimal advertising intervention path for the user to move from the current cognitive state to the expected conversion state is determined, and a corresponding advertising delivery sequence is constructed based on this path.

[0013] Furthermore, the steps for identifying multiple cognitive state nodes include:

[0014] During the system initialization phase, a tag system containing seven user cognitive states is predefined, namely initial exposure, potential interest, clear interest, observation and comparison, conversion preparation, actual conversion, and post-conversion diffusion.

[0015] Construct a user behavior sequence, which is based on click behavior data and transaction path data, and includes characteristics such as behavior event type, behavior content tag, entry location and operation duration.

[0016] The behavior sequence is input into a multi-layer Transformer encoder containing a self-attention mechanism for modeling, and the context aggregation representation of each behavior node is output.

[0017] The aggregated representation is classified using a Softmax classifier to generate a corresponding cognitive state label sequence. Consistency constraints and post-processing are then applied to the label sequence to obtain the user's cognitive state path.

[0018] Furthermore, the steps for constructing a cognitive-behavioral causal map include:

[0019] Based on the user's cognitive state label sequence, a triplet of user state transition samples under different ad delivery is constructed. The triplet includes a state pair, the delivered ad, and the transition result.

[0020] Construct a multi-type directed graph containing state nodes and advertising intervention edges, wherein the edges carry state transition conditional probabilities;

[0021] The state transition causal effect under advertising intervention is calculated using the Bayesian causal inference method. If the causal effect exceeds a set threshold, the edge is retained.

[0022] The training samples and causal edge weights are updated periodically based on a sliding time window. If the advertising content changes, the subgraph is reconstructed and the intervention capability of the new advertisement is estimated.

[0023] Furthermore, the prediction steps for the state transition probability include:

[0024] Construct a Bayesian network that jointly models the user's current state, advertising content, behavioral characteristics, and predicted state;

[0025] The graph structure of the Bayesian network is determined by a structure learning algorithm, and the conditional probability table is established by maximum likelihood estimation or Bayesian estimation.

[0026] Input the current state, the advertisement to be delivered, and user behavior characteristics, and use variational inference or forward sampling methods to calculate the conditional probability distribution of the transition from the current state to each candidate state;

[0027] For multiple candidate ads, output their corresponding ad response probability, state transition probability, and path utility score, which are then used for subsequent path planning.

[0028] Furthermore, the steps for constructing the ad delivery sequence include:

[0029] The state nodes and advertising intervention edges are constructed into a weighted directed graph, where the weight of the edge is a weighted function of the state transition probability and the advertising budget cost.

[0030] The product maximum path algorithm or the shortest path algorithm is used to search for the optimal path between the current state and the expected transformation state;

[0031] The corresponding advertising delivery sequence is determined according to the optimal path, and the advertising content, display timing, channel location and budget control parameters for each stage are specified.

[0032] The generated ad delivery sequence is stored in the user's personalized ad plan cache, and the delivery execution engine triggers the delivery in real time based on the user's status.

[0033] In another aspect, the present invention provides an advertising delivery system based on multi-source data analysis, comprising the following modules:

[0034] The data acquisition module is used to collect multi-source raw data related to users. The raw data includes at least user click behavior data and transaction path data, which are used to characterize user behavior trajectory and purchase intention.

[0035] The state recognition module is used to identify multiple cognitive state nodes based on the click behavior data and transaction path data. The cognitive state nodes are used to characterize the user's psychological state or behavioral stage during the advertising interaction process.

[0036] The causal modeling module is used to construct a causal graph of cognitive behavior based on the multiple cognitive state nodes. The nodes of the causal graph represent the user's cognitive state, and the directed edges of the causal graph represent the causal transfer relationship between different advertising content and the user's cognitive state.

[0037] The probability reasoning module is used to combine historical behavior samples with the path structure in the causal graph, use a Bayesian reasoning model to predict the advertising response probability and conversion probability of the target user, and estimate the state transition probability of the user from the current state node to the target state node based on different advertising intervention content.

[0038] The path planning module is used to determine the optimal advertising intervention path for the user from the current cognitive state to the expected conversion state based on the state transition probability and causal graph structure, and to construct the corresponding advertising delivery sequence based on the path.

[0039] Furthermore, the state recognition module includes:

[0040] The tag system initialization unit is used to predefine a tag system containing seven user cognitive states during the system initialization phase, namely initial exposure, potential interest, clear interest, observation and comparison, conversion preparation, actual conversion, and post-conversion diffusion.

[0041] The behavior sequence construction unit is used to construct a user's behavior sequence based on click behavior data and transaction path data. The behavior sequence includes behavior event type, behavior content tag, entry location and operation duration characteristics.

[0042] The sequence modeling unit is used to model the behavior sequence input into a multi-layer Transformer encoder containing a self-attention mechanism, and output the context aggregation representation of each behavior node.

[0043] The state classification unit is used to classify the context aggregation representation to generate a corresponding cognitive state label sequence, and to perform consistency constraints and post-processing on the label sequence to obtain the user's cognitive state path.

[0044] Furthermore, the causal modeling module includes:

[0045] The triplet generation unit is used to construct a state transition sample triplet containing state pairs, advertising delivery, and transition results based on the user's cognitive state label sequence.

[0046] A causal graph structure construction unit is used to construct a multi-type directed graph containing state nodes and advertising intervention edges based on the triples, wherein the edges carry state transition conditional probabilities.

[0047] The causal effect calculation unit is used to calculate the state transition causal effect under advertising intervention using the Bayesian causal inference method, and retain the corresponding edge when the causal effect exceeds a set threshold.

[0048] The graph maintenance unit is used to periodically update training samples and causal edge weights based on a sliding time window, and to reconstruct the causal subgraph and estimate the intervention capability of the new advertisement when the advertisement content changes.

[0049] Furthermore, the probabilistic reasoning module includes:

[0050] The Bayesian network building unit is used to construct a Bayesian network that jointly models the user's current state, advertising content, behavioral features, and predicted state. The graph structure of the network is determined by a structure learning algorithm, and the conditional probability table is established by maximum likelihood estimation or Bayesian estimation.

[0051] The state probability inference unit is used to calculate the conditional probability distribution of transitions from the current state to each candidate state by using variational inference or forward sampling methods, given the current state, the advertisement to be displayed, and user behavior characteristics.

[0052] The response scoring unit is used to output the corresponding ad response probability, state transition probability and path utility score for multiple candidate ads, and is used by the path planning module to select the optimal ad.

[0053] Furthermore, the path planning module includes:

[0054] The graph construction and weighting unit is used to construct a weighted directed graph from state nodes and advertising intervention edges, wherein the weight of the edge is a weighted function of the state transition probability and the advertising budget cost.

[0055] The path search unit is used to search for the optimal path from the current state to the expected transformation state based on the product maximum path algorithm or the shortest path algorithm.

[0056] The ad delivery sequence generation unit is used to determine the corresponding ad delivery sequence according to the optimal path, and to set the ad content, display timing, channel location and budget control parameters for each stage of the delivery.

[0057] The caching and scheduling unit is used to store the ad delivery sequence into the user's personalized ad plan cache, and to trigger the presentation of ad content in real time by the ad delivery execution engine according to changes in user status.

[0058] This invention provides an advertising delivery method and system based on multi-source data analysis, which has the following significant advantages:

[0059] By collecting user click behavior data and transaction path data, and combining them with the Transformer model to model behavioral sequences, the system identifies the evolution trajectory of users' cognitive states during the advertising interaction process, and depicts the psychological path from exposure to conversion, effectively breaking through the limitations of traditional user profiling methods in terms of time dimension and psychological modeling.

[0060] By constructing a cognitive-behavioral causal map, we can identify the causal effects of different advertising content on the transition of users' cognitive states. This differs from traditional relevance ranking and enables accurate judgment of advertising intervention behavior, thereby avoiding ineffective intervention and waste of resources.

[0061] By leveraging a Bayesian network structure, user characteristics, advertising attributes, and state variables are integrated to achieve joint modeling of advertising response probability and state transition probability, and to support online updates and real-time prediction, thereby improving the accuracy and robustness of strategy decision-making.

[0062] Based on causal graphs and probabilistic reasoning, graph search algorithms and Markov path planning methods are further introduced to achieve the optimal advertising delivery path design from the current state to the target state, thereby improving the strategic nature and stage adaptability of the advertising system. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a flowchart of the method of the present invention;

[0065] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0066] The invention will now be described in preferred form with reference to the accompanying drawings and specific embodiments.

[0067] This embodiment solves the above problems through the following steps:

[0068] In one embodiment, reference Figure 1 This invention provides an advertising delivery method based on multi-source data analysis. By collecting and integrating user-related data from multiple heterogeneous sources, and combining user behavior trajectories and cognitive responses in different business systems, a unified data expression model is constructed. Furthermore, cognitive behavior analysis and causal reasoning algorithms are introduced to identify changes in users' psychological state and behavioral conversion paths during the advertising exposure process. Ultimately, personalized and dynamic advertising content generation and delivery strategy planning are achieved, thereby significantly improving advertising reach efficiency and conversion effect.

[0069] Step S1: Collect multi-source raw data related to users. The raw data includes at least user click behavior data and transaction path data, which are used to characterize user behavior trajectory and purchase intention.

[0070] In advertising strategies, users' true interests and conversion intentions are often hidden within their multi-dimensional behavioral data. A single data dimension cannot fully reflect users' interest motivations and decision-making processes. Therefore, effectively integrating user behavior data from multiple sources is a prerequisite for building an accurate ad delivery mechanism. The core of this step lies in dynamically characterizing users' behavioral trajectories by integrating click behavior data and transaction path data from multiple data sources, thereby inferring their potential reactions to advertising content. Click behavior reveals users' explicit preferences, while the transaction path reveals the entire process users go through to complete their intent conversion. Combining the two can more completely represent users' cognitive evolution process and true purchase intentions.

[0071] Select data systems that include at least the following sources as the targets for information collection:

[0072] User behavior log system for web or app;

[0073] Sales management systems (such as PIC systems, smart employee badges);

[0074] Contract and commission management systems (such as channel distribution systems and nationwide marketing systems);

[0075] User reporting and transaction confirmation system.

[0076] Record every click event by a user at digital advertising touchpoints (such as mini-programs, promotional links, and in-app ads); each click event should include information such as timestamp, click location, ad content ID, ad type, user ID, and page; if there is page dwell time, swipe path, or redirection behavior after clicking, collect it as well.

[0077] The transaction path is defined as the complete process from a user's first contact to the final order (subscription, signing, or successful referral); key behavioral nodes are extracted, such as registering an account, filling in customer information, reporting customers, viewing property details, receiving commissions, and signing contracts; a behavioral sequence is constructed for each user, and it is marked whether a final conversion has been achieved (i.e., whether a certain transaction loop behavior has been completed); if there are multiple sources, data fusion is achieved through a unified user identifier (mobile phone number, device ID, account system).

[0078] By collecting and integrating click behavior data and transaction path data through the above steps, we can not only refine the entire process from user interest to action, but also capture the implicit fluctuations in their intentions. Unlike traditional static profiling, this method can characterize the cognitive dynamics and path selection of users throughout the entire advertising exposure process, thus providing a solid data foundation for subsequent cognitive state modeling, causal inference, and advertising intervention path generation.

[0079] For example, a user clicks on a "Referral Rewards" advertisement card while browsing property information on a real estate app. The system records the following sequence of actions:

[0080] 2025-03-02 09:15, clicked the ad "Recommend a friend to earn commission";

[0081] At 9:18 AM on March 2, 2025, you were redirected to the project details page and browsed the "XX Garden" property for 3 minutes.

[0082] March 2, 2025, 9:25 AM: Registered as a freelance broker and linked mobile phone number;

[0083] 2025-03-03 11:02, Submit the information of the first recommended customer, click Submit;

[0084] At 10:00 on March 10, 2025, the customer signed the subscription agreement and a commission confirmation slip was generated.

[0085] The above process constitutes a complete transaction path, and each behavioral node will serve as a cognitive state node input into the subsequent modeling process. Based on this, the system can identify that Xiao Li is sensitive to "commission incentives" and appreciates the "cost-effectiveness of the property," thus identifying him as a high-intent customer and targeting him with "more commission strategy ads" in the future to further activate his dissemination behavior.

[0086] Step S2: Based on the click behavior data and transaction path data, identify multiple cognitive state nodes. The cognitive state nodes are used to characterize the user's psychological state or behavioral stage during the advertising interaction process.

[0087] Throughout the entire advertising process, users don't exist as static "interested" or "uninterested" labels, but rather exhibit a dynamic evolution of cognitive processes at different stages, such as from "initial attention" to "hesitation and comparison," then to "expressing purchase intent," and even "actual conversion." While these psychological changes cannot be directly observed, they can be indirectly modeled through user behavior sequences. The purpose of identifying these cognitive state nodes is to abstract user behavior into a quantifiable cognitive path, enabling subsequent advertising interventions to accurately pinpoint "which psychological stage the user is experiencing" and match appropriate advertising content and delivery timing accordingly, thereby effectively improving the intervention effect and conversion success rate.

[0088] Traditional advertising systems often simply summarize user behavior, such as "clicked the ad 3 times," ignoring the underlying motivations behind those clicks. However, by constructing a cognitive state node system, each action can be given a stage-specific meaning, transforming the advertising system from a "reactive" to an "understanding" system.

[0089] The specific implementation process of step S2 includes:

[0090] Step S21, Define the cognitive state labeling system

[0091] First, during the system initialization phase, a fixed sequence of user cognitive state tags is predefined. This sequence forms a complete link from the first ad exposure to the final conversion, specifically including the following seven state tags:

[0092] State S1: Initial Exposure

[0093] State S2: Latent Interest

[0094] State S3: Clear Interest

[0095] State S4: Observation and Comparison

[0096] State S5: Transformation Preparation

[0097] State S6: Actual Transformation

[0098] State S7: Diffusion after transformation

[0099] This status label system is used to represent the psychological evolution stages of users during the advertising interaction process and serves as the classification target for subsequent status recognition models.

[0100] Step S2, construct user behavior sequence

[0101] The user click behavior data and transaction path data collected in step S1 must first be constructed into a unified, ordered time series input. The specific steps are as follows:

[0102] All user behavior events are sorted in ascending order by timestamp to form an event sequence;

[0103] Each event is represented as a 5-tuple:

[0104] e i =(t i ,a i ,c i ,p i ,s i )in:

[0105] t i Indicates the time when the event occurred;

[0106] a i The behavior type (clicking an ad, staying on a page, registering, submitting a recommendation, signing a contract, etc., is mapped to discrete encoding);

[0107] c i Tags representing behavioral content (such as property name, advertisement ID, commission policy ID, etc.);

[0108] pi Indicates the location or entry point of the action (mini program, app homepage, event page);

[0109] s i Indicates dwell time or operation duration (unit: seconds).

[0110] One-Hot encoding is used for all discrete attributes (such as behavior type and entry location); a pre-trained BERT model is used to encode the behavior content labels, converting the behavior content text into a vector of dimension T; dwell time and behavior time interval are standardized; the above vectors are concatenated to form a behavior feature vector sequence X = {x1, x2, ..., x...} of dimension T. T}

[0111] Step S23: Model user behavior sequences using Transformer

[0112] The user's behavioral feature sequence is modeled using the standard Transformer Encoder architecture; each input vector x i Add position encoding PE i The input is embedded; the Transformer contains 4 encoder layers, each with self-attention and a feedforward network, connected to residuals using LayerNorm; the output is a set of context-enhanced sequence representations {h1, h2, ..., h...} T}, each h t This represents the aggregated state of the user's behavior context at time t.

[0113] The representation h at each time step t Input Softmax classifier: The classifier outputs a 7-dimensional probability vector, corresponding to the aforementioned seven cognitive states; the model is trained using cross-entropy loss, and the labels are derived from expert annotations of the historical conversion process or automatically generated based on the post-transaction backtracking path.

[0114] Generate a corresponding state label sequence Y = {y1, y2, ..., y3} for each user's behavior sequence. T If there are consecutive identical labels, the segments are merged to obtain a sequence of cognitive state segments, which serves as the state path.

[0115] Step S24, State Consistency Constraints and Post-processing

[0116] To avoid invalid state jumps in model predictions, such as anomalous paths like "S1→S5→S2", the following constraint strategy is adopted:

[0117] The state labels have a partial order relationship, and the number of the next state must not be lower than that of the previous state; the maximum sub-segment method is used to correct short-term fluctuating states (such as a single S2 inserted in the middle); the final output is the state transition path.

[0118] This step addresses the challenges of interpreting and modeling user behavior by constructing a behavior sequence recognition system based on the Transformer model. This system transforms user behavior processes into cognitive state paths with clear semantics, resolving the difficulties in interpreting and modeling user behavior. Compared to rule-based or static classification approaches, this method offers the following advantages: it captures the temporal relationships between behaviors and distinguishes the impact of behavior sequence on cognitive evolution; it integrates the semantics of preceding and following behaviors through a self-attention mechanism, enhancing the accuracy of state recognition; the generated cognitive state labels provide precise anchors for advertising intervention and personalized content generation; and finally, the state label system can be continuously updated based on business needs, and the behavior feature vector structure can be seamlessly extended to other input dimensions.

[0119] For example, a user's recent behavior record is as follows:

[0120] 2025-04-10 10:01: Clicked on the "XX Garden New Project Launch" advertisement card, browsed for 2 minutes, and then returned;

[0121] 2025-04-10 10:05: Redirected to the floor plan page, stayed for 3 minutes;

[0122] 2025-04-10 10:12: View the "Referral Commission Earning Guide" and bind your agent identity;

[0123] 2025-04-10 11:30: Submitted the first recommended customer;

[0124] 2025-04-11 09:00: Visited the project details page again;

[0125] 2025-04-12 17:30: Click on the "Limited-Time Double Rebate" promotional advertisement to enter the subscription process.

[0126] Based on the above behavioral sequence, the system automatically identified the following evolution of Xiao Wang's cognitive state:

[0127] Initial exposure → Increased interest → Engagement and interaction → Increased intent → High conversion rate

[0128] The system will eventually mark them as "high-intent stage" users, and can then target them with ads comparing "commission payment speed" to further guide them to sign up. They will also be marked as "seed spreaders" and recommended to be targeted with ads for "invite friends to earn double commissions".

[0129] The identification of this cognitive state path enables advertising content to be highly matched with the user's state, achieving the dual purpose of precise targeting and psychological intervention.

[0130] Step S3: Construct a cognitive behavior causal graph based on the multiple cognitive state nodes. The nodes of the causal graph represent the user's cognitive state, and the directed edges of the causal graph represent the causal transfer relationship between different advertising content and the user's cognitive state.

[0131] In advertising delivery systems, understanding how user behavior changes due to advertising is more crucial than simply knowing what users did. Traditional recommendation systems only build behavioral relevance models, neglecting the actual causal impact of advertising behavior as an intervention variable on users' cognitive evolution paths. If we cannot clearly define the causal relationship between advertising and user state transitions, advertising intervention can only be blind, making strategic guidance difficult. Constructing a cognitive-behavioral causal map aims to capture the causal impact mechanism of advertising on users' psychological states and behavioral transition paths, thereby enabling intelligent intervention reasoning on what type of advertising can successfully guide users from a certain cognitive state to the next.

[0132] The nodes of the causal graph represent cognitive states, and the edges represent possible paths for state changes after an ad is placed. This structure can not only explain historical behavior paths, but also be used to predict which ads should be placed in the current state to drive conversions, providing a solid foundation for subsequent ad path optimization and content recommendation.

[0133] To achieve the above objectives, step 3 is implemented through the following steps.

[0134] Step S31, Prepare the input dataset

[0135] Using the user cognitive state sequence identified in step S2:

[0136] S u ={(t1,s1),(t2,s2),...,(t n ,s n )}

[0137] Among them, each s i Represents a cognitive state label, timestamp t i Indicates the time of the status change.

[0138] For every two adjacent state pairs (s) i ,s i+1 Check for any advertising interference within the specified time period:

[0139] If the user is in [t] i ,t i+1 [Clicked or received an advertisement between] kIf , then it is considered a candidate causal pair of "advertising intervention leads to state transition".

[0140] Constructing the training sample set:

[0141] D={(s i ,a k ,s i+1 )}

[0142] Indicates that in state s i Place an advertisement a k After the user enters state s i+1 The observed triplet.

[0143] Step S32, construct the causal graph structure

[0144] Define the cause-effect graph structure: Construct a multi-type directed graph G = (V, E), where:

[0145] The node set V = S, which represents all cognitive state labels;

[0146] Edge set E = {(s i ,s j ;a k )} represents advertisement a k The triggered state transition path;

[0147] Each edge carries a conditional probability weight P(s). j |s i ,a k ), indicating the state s i Advertisement a k Transfer to s j The estimated probability.

[0148] For each state transition pair and advertisement type, the transition strength is calculated using a Bayesian causal inference method enhanced by conditional mutual information:

[0149]

[0150] Wherein, P(s) j |do(a k ),s i ) represents the state transition probability after applying advertising intervention;

[0151] P(s j |s i ) represents the transition probability under natural conditions (without advertising intervention);

[0152] Specifically, the conditional causal effect can be estimated using the causal graph inference framework DoWhy+EconML.

[0153] Perform CE calculation on all triples. If CE(s)i ,a k ,s j If )>δ, then it is considered that advertisement a k In state s i The following is a transfer to s j If there is a significant causal effect, retain the edge; δ is the causal significance threshold (e.g., 0.2), which is set through cross-validation.

[0154] Step S33, Atlas Update and Maintenance Mechanism

[0155] Advertising effectiveness changes over time, so the training sample set and transition statistics are updated periodically by rolling the window; when the advertising content changes, the subgraph is reconstructed and the causal edge weights are updated; short-cycle A / B testing is used to supplement the sample for new ads to quickly estimate their intervention capability under different states.

[0156] In this step, constructing a cognitive-behavioral causal map has the following significant advantages:

[0157] It can identify "which ads truly drive changes in cognitive state" rather than just ranking by click frequency; it can design the optimal delivery sequence based on different ad combination paths in the graph; it not only knows who the user is, but also "at what stage and what content they need to drive conversion"; and it is convenient for ad delivery management decisions and performance tracking.

[0158] For specific examples:

[0159] In cognitive state S2: potential interest, the probability of the "limited-time commission increase" advertisement leading to a transfer to S3: explicit interest is 0.58, the probability of a natural transfer is 0.23, and the causal effect is 0.35.

[0160] In cognitive state S4: observation and comparison, the advertisement "customer positive video recommendation" can guide the transition to S5: conversion preparation, with a CE of 0.41;

[0161] In the S3→S5 path, the CE of the "Frequently Asked Questions" advertisement is negative, indicating that the advertisement may cause cognitive degradation. The system will automatically avoid pushing the advertisement in the S3 state.

[0162] Based on the graph structure, when a user is currently in state S2, the system will prioritize displaying "limited-time commission increase" ads to maximize the probability of a successful state transition, accurately intervene in the psychological stage, and optimize the return on advertising budget.

[0163] Step S4: Combining historical behavior samples with the path structure in the causal graph, a Bayesian inference model is used to predict the advertising response probability and conversion probability of the target user, and the state transition probability of the user moving from the current state node to the target state node is estimated based on different advertising intervention content.

[0164] Advertising placement is essentially a strategy optimization problem with intervention intent. Different advertising content has varying intervention effects on different users in different cognitive states. In the previous step, we clarified the structure of "whether advertising is causal" by constructing a cognitive-behavioral causal graph. However, in actual placement decisions, it is necessary to further quantify the probability that each type of advertising content will successfully guide the user to complete a state transition in a specific state, thereby achieving optimal path planning and placement control. This problem is essentially a Bayesian conditional reasoning problem: we need to predict whether the user will transition to the next stage state given the user's cognitive state and the advertising intervention behavior.

[0165] The advantage of Bayesian inference lies in its ability to naturally integrate prior knowledge (historical behavioral samples) with structural information in causal graphs. It not only outputs transition probabilities but also continuously updates prediction accuracy based on observed data, achieving adaptive intelligent inference. Furthermore, modeling within the Bayesian framework is highly interpretable, making it suitable for applications in high-risk decision-making environments (such as advertising budget control).

[0166] To achieve the above objective, step S4 can be implemented through the following process:

[0167] Step S41, construct the state-advertisement-response triple dataset

[0168] Extract all edges from the causal graph (s i ,a k ,s j ), indicating that in state s i Place an ad below k Later transferred to s j The observed samples.

[0169] For each state edge, record the following variables:

[0170] Current state: s i

[0171] Ad content ID: a k

[0172] Advertising type characteristics (e.g., incentive-based, content-based, social): encoded as feature vectors;

[0173] User feature vector: output by the behavior representation model in step S2;

[0174] Final state: s j Or whether it has been transformed into a certain target state (e.g., transformed / not transformed).

[0175] Construct the training sample set: D = {(x u ,si ,a k ,s j )}

[0176] Where, x u It represents the user's historical behavior at the time of the campaign, s i This is the state before deployment, s j This is the subsequent status.

[0177] Step S42, construct the Bayesian network structure

[0178] Construct a joint probability distribution model:

[0179] P(S t+1 A t ,S t ,X u )

[0180] in:

[0181] S t Indicates the user's current state;

[0182] A t Indicates the content of the advertisement being placed;

[0183] X u This represents the user's current action.

[0184] S t+1 This indicates the predicted state.

[0185] Construct a directed graph model with the following dependencies:

[0186] X u →S t+1 This indicates that user characteristics affect the final state;

[0187] S t →S t+1 This indicates that the initial state has a decisive influence on the transition path;

[0188] A t →S t+1 This indicates that advertising intervention is the core causal variable;

[0189] S t →A t This indicates that the type of advertising is influenced by cognitive state (i.e., targeted advertising).

[0190] When training the model, a structural learning algorithm (such as greedy structure search or maximum expectation a posteriori algorithm) can be used to determine the graph structure; a conditional probability table (CPT) can be calculated using maximum likelihood estimation or Bayesian estimation; a Gaussian mixture Bayesian network can be used to model continuous variables (such as behavior vectors), or a variational Bayesian approximation method can be used to process high-dimensional variables. The specific training process is not the focus of this invention, and those skilled in the art can use any relevant techniques in the prior art to implement it.

[0191] Step S43, State transition probability prediction

[0192] During the prediction phase, the input is:

[0193] Current user state S t =s i ;

[0194] Current advertising campaign plan A t =a k

[0195] User behavior feature vector X u .

[0196] Inferring the probability distribution of the target state:

[0197] P(S t+1 =s j |S t =s i A t =a k ,X u )

[0198] Conditional probability inference is performed on the Bayesian network using variational inference or forward sampling methods to obtain the transition probability vector. That is, from state s i Through advertisement a k Transition to any state s j The probability distribution.

[0199] If the target state is set to s * The corresponding transformation probability is:

[0200] P convert =P(S) t+1 =s * |S t A t ,X u )

[0201] Step S44: Output response probability and path score

[0202] For multiple optional ads a k Calculate their corresponding values ​​respectively:

[0203] Ad response probability (e.g., click-through rate): P(click|A) t =a k ,S t ,X u );

[0204] State transition probability: P(S) t+1 =s * |S t A t =a k ,X u )

[0205] Path utility score: set weight w c w t Overall response rate and conversion rate:

[0206] Score(a k ) = w c ·P(click)+w t ·P (convert)

[0207] Select the ad content with the highest score as the optimal intervention content and proceed to the next step, the ad path planning module.

[0208] In this step, compared to traditional click-through rate (CTR) prediction models, Bayesian networks can integrate the three-dimensional causal relationship of "user state - advertising intervention - behavioral conversion," which is more in line with the actual advertising-driven mechanism. At the same time, Bayesian models naturally support incremental learning and prior updates, allowing predictions to be adjusted in real time based on new behavioral data. Each state transition prediction can be interpreted as a probabilistic decision under specific conditions, making it suitable for strategy formulation and campaign optimization. This model quantifies the transition probabilities of edges based on a causal graph structure, achieving integrated modeling of structural and quantitative information.

[0209] For specific examples:

[0210] A user is currently in cognitive state S3: clearly defined interests. The system presents three alternative advertisements:

[0211] a1: Limited-time double rebate;

[0212] a2: Customer testimonial videos for the property development;

[0213] a3: List of recommended high-commission projects.

[0214] The system infers through a Bayesian network:

[0215] P(S4|S3,a1)=0.45,P click =0.32

[0216] P(S4|S3,a2)=0.61,P click =0.21

[0217] P(S4|S3,a3)=0.37,P click =0.40

[0218] After combining the path score calculation, the system selects ad a2 for push. Although its click-through rate is not the highest, its conversion jump probability is the highest, thus reflecting the advertising intervention strategy with the goal-driven approach as the core.

[0219] Step S5: Based on the state transition probability and causal graph structure, determine the optimal advertising intervention path for the user from the current cognitive state to the expected conversion state, and construct the corresponding advertising delivery sequence based on the path.

[0220] The ultimate goal of advertising is not "clicks" themselves, but to guide users through a cognitive evolution path, continuously transitioning from the current state to the target state (such as signing a contract, recommending a product, or converting again). However, the intervention effects of different ads at different cognitive stages are asymmetrical; some ads may be ineffective or even have negative impacts at specific stages. Therefore, selecting ads solely based on a single conversion probability is insufficient to guide the entire process of intervention. To achieve systematic, goal-oriented advertising, it is necessary to find a state transition path with the highest overall conversion probability or the lowest path cost, based on the state transition probability matrix and causal graph structure obtained in the previous step. This allows for the development of an ad sequence of "what to advertise first, what to advertise next," thus achieving advertising path planning. This strategy makes advertising both "process-oriented" and "stage-based," avoiding duplicate, conflicting, and inefficient interventions.

[0221] To achieve the above objective, step S5 may be specifically implemented as follows:

[0222] Step S51, Define the problem modeling objective

[0223] The current state is set to S0, and the target transformation state is set to S. * Build from S0 to S * A directed graph; each node in the graph represents the user's cognitive state, and each edge in the graph represents the user's cognitive state. This indicates that advertisement a is in state S. i The state transition triggered by the next step is S. j With transition probability

[0224] Each edge also has a cost function, specifically:

[0225] The cost of placement (such as budget consumption, opportunity cost of impressions) or the negative logarithmic probability, i.e.:

[0226]

[0227] The goal is to start from the current state, select a path that passes through several ad touchpoints, and guide the user to the target state S. * The probability of finding the path is the highest, or the total cost of the path is the lowest.

[0228] Step S52: Construct the graph structure and weight it.

[0229] Traverse the causal graph to construct a directed weighted graph G = (V, E), where:

[0230] V represents a node with a cognitive state;

[0231] E represents all possible intervention edges, in the form of triples (S i ,a,S j ).

[0232] Assign two attributes to each edge:

[0233] Transition probability weights Cost weighting: such as budget consumption cost(a), or overall score:

[0234]

[0235] Step S53, Path Search and Planning

[0236] If the goal is to maximize the transformation probability: use the product maximum path algorithm (which can be transformed into the shortest path through logarithmic transformation);

[0237] If the objective is to minimize overall cost: use Dijkstra's shortest path algorithm or A* heuristic search algorithm;

[0238] Furthermore, for situations where there is uncertainty and variability in the state transition, a Markov Decision Process (MDP) is introduced, and dynamic programming is used for strategy planning.

[0239] When executing path planning, the current state S0 and the target state S are input. * The transition probability and cost of advertising intervention edges; output a set of state node sequences {S0, S1, ..., S...} k =S *} and its corresponding advertising sequence {a0,a1,…,a k-1}

[0240] Step S54: Generate an ad delivery sequence

[0241] Each state transition node corresponds to an advertising intervention action; by pairing each state node in the path with an advertisement, the following delivery sequence is generated:

[0242] [(S0,a0),(S1,a1),…,(S k-1 ,a k-1 )]

[0243] Furthermore, for each advertisement:

[0244] Set the timing of the campaign (e.g., immediately, the next day, or when user behavior meets a certain condition); specify the ad placement and channels (e.g., mini-program recommendations, app push notifications, SMS notifications, etc.); set the campaign budget and display frequency.

[0245] Finally, the sequence is stored in the user's personalized advertising plan cache, waiting for the engine to be triggered.

[0246] This step constructs an evolutionary strategy for ad placement from "points" to "paths," moving away from isolated ad placement and instead guiding user behavior strategically, directionally, and purposefully. Through probability-weighted path selection, the highest conversion rate can be achieved at the lowest cost. Different ad types are matched to different cognitive states to achieve precise content matching.

[0247] Specific examples:

[0248] A user is currently in state S2: the potential interest stage, and the system expects them to transition to state S6: the actual conversion stage.

[0249] Based on the causal graph and the state transition probability estimation of the Bayesian model, the system constructs the following transition path:

[0250]

[0251] in:

[0252] a1: "Video introduction of popular real estate projects" advertisement, conversion probability 0.6;

[0253] a2: "Limited-time high commission referral incentive" ad, with a conversion probability of 0.5;

[0254] a3: "Instructions for Signing a Contract and Rebate Payment Process" advertisement, transfer probability 0.8.

[0255] The total probability of path transformation is approximately:

[0256] P total =0.6 × 0.5 × 0.8 = 0.24

[0257] The system thus generates an ad delivery sequence:

[0258] In the first phase, push a1; after the user enters S3, push a2 24 hours later; if the user does not sign up within 24 hours after entering S5, reassess the status and enter the dynamic planning resubmission sequence.

[0259] Through such path-based planning and intervention, advertising is no longer a single-point stimulus, but a cognitive behavior guidance tool driven by stages and oriented towards conversion.

[0260] See Figure 2 In another embodiment, the present invention also provides an advertising delivery system based on multi-source data analysis, comprising:

[0261] The data acquisition module is used to collect multi-source raw data related to users. The raw data includes at least user click behavior data and transaction path data, which are used to characterize user behavior trajectory and purchase intention.

[0262] The state recognition module is used to identify multiple cognitive state nodes based on the click behavior data and transaction path data. The cognitive state nodes are used to characterize the user's psychological state or behavioral stage during the advertising interaction process.

[0263] The causal modeling module is used to construct a causal graph of cognitive behavior based on the multiple cognitive state nodes. The nodes of the causal graph represent the user's cognitive state, and the directed edges of the causal graph represent the causal transfer relationship between different advertising content and the user's cognitive state.

[0264] The probability reasoning module is used to combine historical behavior samples with the path structure in the causal graph, use a Bayesian reasoning model to predict the advertising response probability and conversion probability of the target user, and estimate the state transition probability of the user from the current state node to the target state node based on different advertising intervention content.

[0265] The path planning module is used to determine the optimal advertising intervention path for the user from the current cognitive state to the expected conversion state based on the state transition probability and causal graph structure, and to construct the corresponding advertising delivery sequence based on the path.

[0266] In a further implementation, the state recognition module includes:

[0267] The tag system initialization unit is used to predefine a tag system containing seven user cognitive states during the system initialization phase, namely initial exposure, potential interest, clear interest, observation and comparison, conversion preparation, actual conversion, and post-conversion diffusion.

[0268] The behavior sequence construction unit is used to construct a user's behavior sequence based on click behavior data and transaction path data. The behavior sequence includes behavior event type, behavior content tag, entry location and operation duration characteristics.

[0269] The sequence modeling unit is used to model the behavior sequence input into a multi-layer Transformer encoder containing a self-attention mechanism, and output the context aggregation representation of each behavior node.

[0270] The state classification unit is used to classify the context aggregation representation to generate a corresponding cognitive state label sequence, and to perform consistency constraints and post-processing on the label sequence to obtain the user's cognitive state path.

[0271] In a further implementation, the causal modeling module includes:

[0272] The triplet generation unit is used to construct a state transition sample triplet containing state pairs, advertising delivery, and transition results based on the user's cognitive state label sequence.

[0273] A causal graph structure construction unit is used to construct a multi-type directed graph containing state nodes and advertising intervention edges based on the triples, wherein the edges carry state transition conditional probabilities.

[0274] The causal effect calculation unit is used to calculate the state transition causal effect under advertising intervention using the Bayesian causal inference method, and retain the corresponding edge when the causal effect exceeds a set threshold.

[0275] The graph maintenance unit is used to periodically update training samples and causal edge weights based on a sliding time window, and to reconstruct the causal subgraph and estimate the intervention capability of the new advertisement when the advertisement content changes.

[0276] In a further implementation, the probabilistic reasoning module includes:

[0277] The Bayesian network building unit is used to construct a Bayesian network that jointly models the user's current state, advertising content, behavioral features, and predicted state. The graph structure of the network is determined by a structure learning algorithm, and the conditional probability table is established by maximum likelihood estimation or Bayesian estimation.

[0278] The state probability inference unit is used to calculate the conditional probability distribution of transitions from the current state to each candidate state by using variational inference or forward sampling methods, given the current state, the advertisement to be displayed, and user behavior characteristics.

[0279] The response scoring unit is used to output the corresponding ad response probability, state transition probability and path utility score for multiple candidate ads, and is used by the path planning module to select the optimal ad.

[0280] In a further implementation, the path planning module includes:

[0281] The graph construction and weighting unit is used to construct a weighted directed graph from state nodes and advertising intervention edges, wherein the weight of the edge is a weighted function of the state transition probability and the advertising budget cost.

[0282] The path search unit is used to search for the optimal path from the current state to the expected transformation state based on the product maximum path algorithm or the shortest path algorithm.

[0283] The ad delivery sequence generation unit is used to determine the corresponding ad delivery sequence according to the optimal path, and to set the ad content, display timing, channel location and budget control parameters for each stage of the delivery.

[0284] The caching and scheduling unit is used to store the ad delivery sequence into the user's personalized ad plan cache, and to trigger the presentation of ad content in real time by the ad delivery execution engine according to changes in user status.

[0285] It should be noted that the explanations and descriptions of the aforementioned advertising placement method embodiments based on multi-source data analysis also apply to the apparatus of the embodiments of this application, and will not be repeated here.

[0286] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0287] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0288] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0289] The above description is merely a specific embodiment of this application. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application. For some module structures not specifically defined in this invention, the content described in the prior art shall prevail. The prior art mentioned in the foregoing background and specific embodiments sections can be considered as part of this invention and used to understand the meaning of some technical features or parameters.

Claims

1. An advertising placement method based on multi-source data analysis, characterized in that, The method includes the following steps: Collect multi-source raw data related to users, including at least user click behavior data and transaction path data, to characterize user behavior trajectory and purchase intention; Based on the click behavior data and transaction path data, multiple cognitive state nodes are identified. These cognitive state nodes are used to characterize the user's psychological state or behavioral stage during the advertising interaction process. A cognitive behavior causal graph is constructed based on the multiple cognitive state nodes. The nodes of the causal graph represent the user's cognitive state, and the directed edges of the causal graph represent the causal transfer relationship between different advertising contents and the user's cognitive state. By combining historical behavior samples with the path structure in the causal graph, a Bayesian inference model is used to predict the advertising response probability and conversion probability of the target user, and the state transition probability of the user from the current state node to the target state node is estimated based on different advertising intervention content. Based on the state transition probability and causal graph structure, the optimal advertising intervention path for the user to move from the current cognitive state to the expected conversion state is determined, and a corresponding advertising delivery sequence is constructed based on this path. The steps for identifying multiple cognitive state nodes include: During the system initialization phase, a tag system containing seven user cognitive states is predefined, namely initial exposure, potential interest, clear interest, observation and comparison, conversion preparation, actual conversion, and post-conversion diffusion; user behavior sequences are constructed, and the behavior sequences are input into a multi-layer Transformer encoder containing a self-attention mechanism for modeling, and the context aggregation representation of each behavior node is output; The aggregated representation is classified using a Softmax classifier to generate a corresponding cognitive state label sequence. Consistency constraints and post-processing are then applied to the label sequence to obtain the user's cognitive state path.

2. The advertising placement method based on multi-source data analysis according to claim 1, characterized in that, The construction of user behavior sequences includes the construction of behavior sequences based on click behavior data and transaction path data, and includes behavior event types, behavior content tags, entry locations and operation duration characteristics. All behavior events of each user are arranged in ascending order by timestamp to form an event sequence.

3. The advertising placement method based on multi-source data analysis according to claim 1, characterized in that, The steps for constructing a cognitive-behavioral causal map include: Based on the user's cognitive state label sequence, a triplet of user state transition samples under different ad delivery is constructed. The triplet includes a state pair, the delivered ad, and the transition result. Construct a multi-type directed graph containing state nodes and advertising intervention edges, wherein the edges carry state transition conditional probabilities; The state transition causal effect under advertising intervention is calculated using the Bayesian causal inference method. If the causal effect exceeds a set threshold, the edge is retained. The training samples and causal edge weights are updated periodically based on a sliding time window. If the advertising content changes, the subgraph is reconstructed and the intervention capability of the new advertisement is estimated.

4. The advertising placement method based on multi-source data analysis according to claim 1, characterized in that, The prediction steps for the state transition probability include: Construct a Bayesian network that jointly models the user's current state, advertising content, behavioral characteristics, and predicted state; The graph structure of the Bayesian network is determined by a structure learning algorithm, and the conditional probability table is established by maximum likelihood estimation or Bayesian estimation. Input the current state, the advertisement to be delivered, and user behavior characteristics, and use variational inference or forward sampling methods to calculate the conditional probability distribution of the transition from the current state to each candidate state; For multiple candidate ads, output their corresponding ad response probability, state transition probability, and path utility score, which are then used for subsequent path planning.

5. The advertising placement method based on multi-source data analysis according to claim 1, characterized in that, The steps for constructing the advertising delivery sequence include: The state nodes and advertising intervention edges are constructed into a weighted directed graph, where the weight of the edge is a weighted function of the state transition probability and the advertising budget cost. The product maximum path algorithm or the shortest path algorithm is used to search for the optimal path between the current state and the expected transformation state; The corresponding advertising delivery sequence is determined according to the optimal path, and the advertising content, display timing, channel location and budget control parameters for each stage are specified. The generated ad delivery sequence is stored in the user's personalized ad plan cache, and the delivery execution engine triggers the delivery in real time based on the user's status.

6. An advertising delivery system based on multi-source data analysis, characterized in that, The system includes the following modules: The data acquisition module is used to collect multi-source raw data related to users. The raw data includes at least user click behavior data and transaction path data, which are used to characterize user behavior trajectory and purchase intention. The state recognition module is used to identify multiple cognitive state nodes based on the click behavior data and transaction path data. The cognitive state nodes are used to characterize the user's psychological state or behavioral stage during the advertising interaction process. The causal modeling module is used to construct a causal graph of cognitive behavior based on the multiple cognitive state nodes. The nodes of the causal graph represent the user's cognitive state, and the directed edges of the causal graph represent the causal transfer relationship between different advertising content and the user's cognitive state. The probability reasoning module is used to combine historical behavior samples with the path structure in the causal graph, use a Bayesian reasoning model to predict the advertising response probability and conversion probability of the target user, and estimate the state transition probability of the user from the current state node to the target state node based on different advertising intervention content. The path planning module is used to determine the optimal advertising intervention path for the user from the current cognitive state to the expected conversion state based on the state transition probability and causal graph structure, and to construct the corresponding advertising delivery sequence based on the path.

7. The advertising delivery system based on multi-source data analysis according to claim 6, characterized in that, The status recognition module includes: The tag system initialization unit is used to predefine a tag system containing seven user cognitive states during the system initialization phase, namely initial exposure, potential interest, clear interest, observation and comparison, conversion preparation, actual conversion, and post-conversion diffusion. The behavior sequence construction unit is used to construct a user's behavior sequence based on click behavior data and transaction path data. The behavior sequence includes behavior event type, behavior content tag, entry location and operation duration characteristics. The sequence modeling unit is used to model the behavior sequence input into a multi-layer Transformer encoder containing a self-attention mechanism, and output the context aggregation representation of each behavior node. The state classification unit is used to classify the context aggregation representation to generate a corresponding cognitive state label sequence, and to perform consistency constraints and post-processing on the label sequence to obtain the user's cognitive state path.

8. The advertising delivery system based on multi-source data analysis according to claim 6, characterized in that, The causal modeling module includes: The triplet generation unit is used to construct a state transition sample triplet containing state pairs, advertising delivery, and transition results based on the user's cognitive state label sequence. A causal graph structure construction unit is used to construct a multi-type directed graph containing state nodes and advertising intervention edges based on the triples, wherein the edges carry state transition conditional probabilities. The causal effect calculation unit is used to calculate the state transition causal effect under advertising intervention using the Bayesian causal inference method, and retain the corresponding edge when the causal effect exceeds a set threshold. The graph maintenance unit is used to periodically update training samples and causal edge weights based on a sliding time window, and to reconstruct the causal subgraph and estimate the intervention capability of the new advertisement when the advertisement content changes.

9. The advertising delivery system based on multi-source data analysis according to claim 6, characterized in that, The probabilistic reasoning module includes: The Bayesian network building unit is used to construct a Bayesian network that jointly models the user's current state, advertising content, behavioral features, and predicted state. The graph structure of the network is determined by a structure learning algorithm, and the conditional probability table is established by maximum likelihood estimation or Bayesian estimation. The state probability inference unit is used to calculate the conditional probability distribution of transitions from the current state to each candidate state by using variational inference or forward sampling methods, given the current state, the advertisement to be displayed, and user behavior characteristics. The response scoring unit is used to output the corresponding ad response probability, state transition probability and path utility score for multiple candidate ads, and is used by the path planning module to select the optimal ad.

10. The advertising delivery system based on multi-source data analysis according to claim 6, characterized in that, The path planning module includes: The graph construction and weighting unit is used to construct a weighted directed graph from state nodes and advertising intervention edges, wherein the weight of the edge is a weighted function of the state transition probability and the advertising budget cost. The path search unit is used to search for the optimal path from the current state to the expected transformation state based on the product maximum path algorithm or the shortest path algorithm. The ad delivery sequence generation unit is used to determine the corresponding ad delivery sequence according to the optimal path, and to set the ad content, display timing, channel location and budget control parameters for each stage of the delivery. The caching and scheduling unit is used to store the ad delivery sequence into the user's personalized ad plan cache, and to trigger the presentation of ad content in real time by the ad delivery execution engine according to changes in user status.

Images