An individualized precision marketing pushing method based on an internet of things

Abstract

The application discloses a kind of personalized precision marketing push methods based on Internet of Things, belong to Internet of Things technical field, which includes the following steps: step one, real-time collection comes from the serialization state data and context data of user associated Internet of Things terminal equipment, generates user short-term intention portrait;Step two, based on user short-term intention portrait, generate current active business journey instance;Step three, for each active node in current active business journey instance, based on user short-term intention portrait and real-time context data, generate push instruction sequence;Step four, execute push instruction sequence, and reach user through the preset communication channel with personalized marketing content.The application realizes the evolution of marketing interaction from isolated reach to context coherent multi-node journey by constructing the continuity business journey framework based on user short-term intention portrait, and according to real-time context dynamically generates current active business journey instance.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to a personalized and precise marketing push method based on IoT. Background Technology

[0002] Personalized marketing using IoT devices is a technology that collects user data through IoT terminal devices, analyzes their behavioral patterns and potential needs, and then delivers precise and contextualized marketing information, aiming to improve marketing conversion efficiency and user experience.

[0003] The working principle of existing technologies usually relies on isolated data analysis of a single device or a single interaction scenario. For example, pushing related product discounts by parsing the inventory logs of a smart refrigerator, or triggering general promotional broadcasts based on user location information captured by store sensors. Its core logic is to respond based on instantaneous, single-point data snapshots.

[0004] However, such methods have significant drawbacks: their marketing pushes are isolated, one-off message deliveries, lacking the ability to construct a context-coherent personalized business journey based on continuous, multi-dimensional IoT data. For example, a system might push online discounts after detecting a milk shortage in a user's smart refrigerator, but it fails to connect the user's subsequent experience of visiting a physical supermarket (sensed by in-vehicle devices) with their previous interests. This prevents the provision of continuous decision support such as parking guidance or shelf navigation, leading to interruptions in the user's decision-making process and inefficient conversion funnel. Existing technologies lack an intelligent mechanism capable of connecting fragmented user interests and achieving continuous interaction across different IoT scenarios. Therefore, there is an urgent need to provide a personalized and precise marketing push method based on the Internet of Things to address these issues. Summary of the Invention

[0005] The technical problem this invention aims to solve is to overcome the shortcomings of existing technologies, which exhibit isolated, one-off message delivery in their marketing pushes, lacking the ability to construct a context-coherent personalized business journey based on continuous, multi-dimensional IoT data. For example, a system might push online discounts after recognizing a milk shortage in a user's smart refrigerator, but it fails to connect the user's subsequent experience of visiting a physical supermarket (sensed by in-vehicle devices) with their previous interests. This prevents the provision of continuous decision support such as parking guidance or shelf navigation, leading to interruptions in the user's decision-making process and inefficient business conversion funnel. Existing technologies lack an intelligent mechanism capable of connecting fragmented user interests and achieving continuous interaction across different IoT scenarios. This invention provides a personalized and precise marketing push method based on the Internet of Things.

[0006] To solve the above-mentioned technical problems, one technical solution adopted by the present invention is to provide a personalized and precise marketing push method based on the Internet of Things, comprising the following steps: Step 1: Collect serialized status data and contextual data from user-associated IoT terminal devices in real time, and perform fusion analysis on the collected serialized status data and contextual data to generate a short-term user intent profile. Step 2: Based on the user's short-term intent profile and combined with preset business goals, construct a continuous business journey framework that includes expected interaction nodes; based on the real-time acquired contextual data, dynamically activate or adjust specific business journey nodes in the continuous business journey framework that match the current user's physical and digital context, and generate currently active business journey instances. Step 3: For each active node in the currently active business journey instance, based on the user's short-term intent profile and the real-time contextual data, generate personalized marketing content related to the context of the active node, calculate the push priority and expected utility value of each personalized marketing content in the current context, and generate a push instruction sequence based on the push priority and expected utility value through a preset push decision engine. Step 4: Execute the push instruction sequence to deliver the personalized marketing content to the user through a preset communication channel.

[0007] The present invention is further configured such that: the IoT terminal device in step one includes smart home appliances and wearable devices for collecting serialized state data, and vehicle-mounted devices and user mobile devices for collecting contextual data; The serialized status data includes smart home device operation logs and records of changes in product inventory status; The contextual data includes the real-time geographic location of the user's mobile device and environmental parameters.

[0008] The present invention is further configured such that the steps for generating the user's short-term intent profile in step one are as follows: S1. Real-time collection of the serialized status data and context data from different IoT terminal devices; timestamp alignment, format conversion and outlier cleaning of the collected smart home device operation logs, product inventory status change records, real-time geographical location and environmental parameters of user mobile devices; and generation of a standardized data set. S2. The standardized data set is parsed to extract key features reflecting user behavior patterns, device interaction status and environmental context, forming a multi-dimensional feature vector. Based on the multi-dimensional feature vector, combined with the user's past behavior habits and preference records retrieved from the preset behavior database, a primary interest profile representing the user's current interest tendency is calculated and generated. The primary interest profile is compared and completed with the preset basic user profile template to output a structured interest profile. S3. Receive the structured interest profile and simultaneously acquire the latest real-time contextual data. Calculate the correlation strength between the real-time contextual data and each interest dimension in the structured interest profile. Assign a preset confidence coefficient to each interest dimension based on the correlation strength. Input the structured interest profile with the preset confidence coefficient into a preset intent inference model and output a short-term user intent profile.

[0009] The present invention is further configured such that the method for generating the continuous business journey framework in step two is as follows: Q1. Based on the preset business goal, create an initial journey framework that includes the starting state and the target state but with no interaction nodes. From the user short-term intent profile, select the core user intents that belong to the same category or have a strong correlation with the business goal under the preset business classification system, and use them as key anchor intents. Inject the key anchor intents into the initial journey framework. Q2. Based on the preset conversion strategy, convert each of the selected core user intentions into sub-business objectives; Q3. Based on the target type, required resources, and estimated time of each sub-business objective, plan expected interaction nodes within a preset time period in the future. Each expected interaction node defines the specific context, content format, and triggering conditions for interaction with the user. According to the natural evolution of user behavior over time and space, arrange all the generated expected interaction nodes into a node sequence according to their contextual logic and triggering conditions. Place the node sequence into and fill the blank interaction node part in the initial journey framework to form a complete path that starts from the current context, runs through multiple sub-business objectives, and ultimately leads to the preset business objective. The complete path is the continuous business journey framework.

[0010] The present invention is further configured such that the method for generating the currently active business journey instance in step two is as follows: Q4. Acquire the latest contextual data in real time, calculate the matching degree between it and the specific context defined by all the expected interaction nodes in the continuous business journey framework, retain only the candidate expected interaction nodes whose matching degree exceeds the preset activation threshold, and form a set of candidate expected interaction nodes to be activated. For each candidate expected interaction node in the set of candidate expected interaction nodes to be activated, simultaneously extract its predefined initial business value weight and urgency indicator in the continuous business journey framework, combine it with the intensity value of the corresponding intent dimension in the user short-term intent profile acquired in real time, correct the initial business value weight, generate a business value weight, and calculate the real-time urgency based on the urgency indicator and the time constraint strength of the current context; multiply the business value weight and the real-time urgency to obtain the comprehensive priority coefficient of each candidate expected interaction node, and sort all the candidate expected interaction nodes in the set of candidate expected interaction nodes to be activated in descending order according to the comprehensive priority coefficient to generate a dynamic priority sequence. Q5. Extract the top-ranked, preset number of candidate expected interactive nodes from the dynamic priority sequence as core activation nodes. Based on the preset node connection rule base in the continuous business journey framework, find and bind feasible predecessor and successor nodes for each core activation node to form a local behavior path segment with the core activation node as the hub. Integrate all the local behavior path segments, check and repair the scene coherence at the junction of the local behavior path segments, assemble them into an interactive path, and instantiate and encapsulate the definition context, content form, triggering conditions and execution order of the interactive path and all the core activation nodes to generate a currently active business journey instance.

[0011] The present invention is further configured such that the generation steps of the local behavior path fragment in step Q5 are as follows: Q501. Based on the predefined interaction scenario requirements of the core activation node in the continuous business journey framework and the real-time scenario data of the user, calculate the feasible interaction time window and content presentation mode boundary of the core activation node to form a basic interaction boundary. At the same time, combined with the user's current device status, network environment and attention level data, construct an interaction safety buffer outside the basic interaction boundary. The basic interaction boundary and the interaction safety buffer together constitute the complete interaction safety area of ​​the core activation node. Q502. Within the complete interactive safety area, taking the core activation node as a benchmark, explore forward a set of feasible preceding nodes that are logically consistent with its triggering conditions and contextually compatible, and explore backward a set of feasible succeeding nodes that are naturally connected with its execution results; based on the path dependency relationship and scenario coherence constraints defined in the preset node connection rule base, calculate the behavioral logic coherence score between each feasible preceding node and the core activation node, and between each feasible succeeding node and the core activation node; filter node pairs whose behavioral logic coherence scores are higher than a preset connectivity threshold, and connect the feasible preceding nodes and feasible succeeding nodes in the node pairs with the core activation node as the center to form multiple candidate local behavioral paths; Q503. Calculate the comprehensive efficiency index of each of the local behavioral paths, arrange all the local behavioral paths in descending order based on the comprehensive efficiency index, select the local behavioral path with the highest comprehensive efficiency index as the optimal local behavioral path, optimize the topology of the optimal local behavioral path, and generate local behavioral path segments.

[0012] The present invention is further configured such that: the personalized marketing content in step three includes: based on a preset dynamic content template library, and according to the definition context of the core activation node and the user's real-time context state, dynamically generating multimodal marketing materials containing personalized product recommendations, contextualized promotional information and interactive guidance content.

[0013] The present invention is further configured such that: the calculation formula for the push priority and the expected utility value in step three is: based on the commercial value weight of the core activation node and the real-time urgency, the two are multiplied to obtain a basic priority score; at the same time, a preset adjustment factor is introduced, and the basic priority score is multiplied by the adjustment factor to obtain a comprehensive priority score; the comprehensive priority scores of all content to be pushed are normalized to obtain the push priority. The matching degree between the personalized marketing content and the user's real-time intent is calculated. The matching degree is obtained by analyzing the overlap between the content keywords and the intent dimensions in the user's short-term intent profile where the confidence coefficient exceeds a preset activation threshold. The real-time inventory satisfaction rate and expected conversion rate of the products or services recommended by the personalized marketing content are evaluated. The matching degree, the real-time inventory satisfaction rate, and the expected conversion rate are multiplied to obtain the original utility estimate. A channel efficiency coefficient based on the historical interaction success rate of the push channel is introduced. The original utility estimate is multiplied by the channel efficiency coefficient, and the result is standardized to generate the expected utility value.

[0014] The present invention is further configured such that the preset push decision engine in step three is specifically a hybrid decision system based on a multi-objective optimization algorithm and a preset real-time rule evaluation module.

[0015] The present invention is further configured such that: in step four, after the push instruction sequence is executed, the user's feedback behavior to the push content is monitored, and the feedback behavior is sent back to step one as new contextual data.

[0016] The beneficial effects of this invention are as follows: 1. This invention constructs a continuous business journey framework based on short-term user intent profiles and dynamically generates currently active business journey instances according to real-time contexts. This enables marketing interactions to evolve from isolated outreach to contextually coherent multi-node journeys, effectively connecting fragmented user interests and behaviors, and significantly enhancing the coherence of user decisions and business conversion efficiency. 2. This invention dynamically generates personalized marketing content by integrating real-time context and intent profiles at active business journey nodes, and makes intelligent decisions and pushes based on push priority and expected utility value. This achieves accurate matching between marketing content and the user's current physical and digital context, greatly improving the targeting of marketing and user experience. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the steps for generating local behavior path segments according to the present invention. Detailed Implementation

[0018] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0019] Please see Figure 1 - Figure 2 A personalized and precise marketing push method based on the Internet of Things includes the following steps: Step 1: Collect serialized status data and contextual data from user-associated IoT terminal devices in real time, fuse and analyze the collected serialized status data and contextual data, identify potential immediate needs and preferences of users, and generate short-term user intent profiles. Step 2: Based on the user's short-term intent profile and combined with the preset business goals, construct a continuous business journey framework that includes expected interaction nodes. The continuous business journey framework defines the key steps and context transition logic from initial contact to target conversion. Based on the context data acquired in real time, dynamically activate or adjust specific business journey nodes in the continuous business journey framework that match the current physical and digital context of the user, and generate currently active business journey instances. Step 3: For each active node in the current active business journey instance, based on the user's short-term intent profile and real-time contextual data, generate personalized marketing content related to the context of the active node, calculate the push priority and expected utility value of each personalized marketing content in the current context, and generate a push instruction sequence through a preset push decision engine based on the push priority and expected utility value. Step 4: Execute the push instruction sequence to deliver personalized marketing content to users through at least one preset communication channel.

[0020] This invention generates accurate short-term user intent profiles by collecting and analyzing multi-source IoT data in real time. Based on this, it dynamically constructs and activates context-matched continuous business journey instances, and intelligently generates matching personalized marketing content for each active node in the instance. By calculating push priority and expected utility value to drive decision-making, it ultimately achieves precise reach with high conversion rates, significantly improving the efficiency of marketing resource utilization and the consistency of user experience.

[0021] In step one, the IoT terminal devices include smart home appliances and wearable devices for collecting serialized status data, as well as vehicle-mounted devices and user mobile devices for collecting contextual data. Serialized status data includes smart home device operation logs and records of changes in product inventory status; Contextual data includes the real-time geographic location of the user's mobile device and environmental parameters.

[0022] The steps for generating the user's short-term intent profile in step one are as follows: S1. Real-time collection of serialized status data and context data from different IoT terminal devices; timestamp alignment, format conversion and outlier cleaning of the collected smart home device operation logs, product inventory status change records, real-time geographical location and environmental parameters of user mobile devices; generating a time-synchronized and structurally unified standardized data set. S2. Analyze the standardized dataset, extract key features that reflect user behavior patterns, device interaction status and environmental context, and form a multi-dimensional feature vector. Based on the multi-dimensional feature vector, combined with the user's past behavior habits and preference records retrieved from the preset behavior database, calculate and generate a primary interest profile that represents the user's current interest tendency. Compare and complete the primary interest profile with the preset basic user profile template, and output a structured interest profile. The pre-built behavior database is constructed by integrating historical user behavior data. This database stores serialized state data and contextual data collected from IoT terminal devices. After cleaning and classification, this serialized state data and contextual data form a set of user behavior records. The user behavior record set is indexed by time series and event type, supporting quick retrieval of users' past behavior habits and preferences. The database is updated regularly to include the latest behavior data, ensuring the timeliness and completeness of the behavior records.

[0023] The calculation method for the primary interest profile is as follows: Based on multi-dimensional feature vectors and user's past behavioral habits and preference records in a preset behavioral database; First, the multi-dimensional feature vectors are matched with the behavioral records for similarity to generate an interest matching score; then, the interest matching scores are weighted and summed, with the weights determined by the recentity and frequency of the behavioral records; the weighted summation result is normalized through a preset activation function to generate the intensity value of each interest dimension; finally, interest dimensions with intensity values ​​exceeding a threshold are aggregated to form the primary interest profile.

[0024] S3. Receive the structured interest profile and simultaneously acquire the latest real-time contextual data. Calculate the correlation strength between the real-time contextual data and each interest dimension in the structured interest profile. Assign a preset confidence coefficient to each interest dimension based on the correlation strength. Input the structured interest profile with the preset confidence coefficient into a preset intent inference model. The intent inference model combines the burst patterns of recent high-frequency behavior sequences to rank and quantify the urgency of each interest. Output a short-term user intent profile that integrates real-time context, historical habits, and behavioral bursts and reflects the user's current needs.

[0025] The correlation strength is calculated by comparing the feature vectors of each interest dimension in the real-time context data with those in the structured interest profile. First, the real-time context data is mapped to feature vectors, and the feature representation of each interest dimension in the structured interest profile is extracted. Then, the Euclidean distance between the two feature vectors is calculated, and the distance value is converted into a similarity score. The similarity score is then filtered by a moving average to eliminate the influence of instantaneous fluctuations. Finally, the correlation strength value is output to quantify the degree of matching between the real-time context and the interest dimension.

[0026] The preset confidence coefficient is dynamically allocated based on the association strength value. The association strength value is divided into multiple levels, and each level corresponds to a confidence coefficient range. The association strength value is mapped to the specific confidence coefficient through a lookup table method, with high strength corresponding to high confidence. The confidence coefficient is further adjusted in combination with the historical stability of the interest dimension. The higher the stability, the more reliable the coefficient. The adjusted confidence coefficient is used in the weighted intent inference process.

[0027] The pre-defined intent inference model adopts a graph neural network architecture. The input is a structured interest profile with confidence coefficients. The intent inference model uses interest dimensions as nodes to construct an interest relationship graph, and the edge weights are initialized by the confidence coefficients. The node information is aggregated through multi-layer graph convolution operations to capture the implicit relationships between interests. The output layer uses an attention mechanism to highlight key interest nodes and generate an intent probability distribution. Finally, the intent is determined by the highest value of the probability distribution and fed back to the business journey framework.

[0028] The method for generating the continuous business journey framework in step two is as follows: Q1. Based on the preset business objectives, create an initial journey framework that includes the starting state and the target state but with no interaction nodes. From the user's short-term intent profile, select the core user intents that belong to the same category or have a strong correlation with the business objectives under the preset business classification system, and use them as key anchor intents. Inject the key anchor intents into the initial journey framework. Q2. Based on the preset conversion strategy, each selected core user intent is converted into one or more specific and actionable sub-business objectives. The conversion strategy analyzes the semantic features, intensity, and historical realization path of the intent, and combines the currently available marketing resources and service capabilities to match and deduce in the preset business logic rule base, thereby generating a set of logically coherent and clearly defined sub-business objectives. These sub-business objectives together constitute a feasible business path to achieve the core user intent. Q3. Based on the target type, required resources, and estimated time for each sub-business objective, plan one or more expected interaction nodes within a pre-set time period in the future. Each expected interaction node defines the specific context, content format, and triggering conditions for interacting with users. According to the natural evolution of user behavior in time and space, arrange all generated expected interaction nodes into a spatiotemporally evolving node sequence according to their contextual logic and triggering conditions. Place the node sequence into and fill the blank interaction node part in the initial journey framework to form a complete path that starts from the current context, runs through multiple sub-business objectives, and ultimately leads to the pre-set business objective. The complete path is the continuous business journey framework.

[0029] The optimal value of the future preset time period is calculated based on the periodicity of user behavior and the urgency of business objectives. First, analyze the historical user interaction data, calculate the average interaction interval and variance, and determine the length of the baseline time period. Then, introduce the time decay factor of business objectives, and multiply the baseline time period by the decay factor to obtain the dynamic time period. The dynamic time period is then fine-tuned according to real-time situational changes, such as shortening the time period when user activity suddenly increases. Finally, the time period ensures that it covers the typical user behavior window while meeting the marketing timeliness requirements.

[0030] The method for generating currently active business journey instances in step two is as follows: Q4. Acquire the latest contextual data in real time and calculate the matching degree between it and the specific context defined by all expected interaction nodes in the continuous business journey framework. Only retain candidate expected interaction nodes with a matching degree exceeding the preset activation threshold to form a set of candidate expected interaction nodes to be activated. For each candidate expected interaction node in the set of candidate expected interaction nodes to be activated, simultaneously extract its predefined initial business value weight and urgency indicator in the continuous business journey framework. Combine the intensity value of the corresponding intent dimension in the real-time acquired user short-term intent profile to correct the initial business value weight and generate a business value weight. Calculate the real-time urgency based on the urgency indicator and the time constraint strength of the current context. Multiply the business value weight and the real-time urgency to obtain the comprehensive priority coefficient of each candidate expected interaction node. Sort all candidate expected interaction nodes in the set of candidate expected interaction nodes to be activated in descending order according to the comprehensive priority coefficient to generate a dynamic priority sequence. The specific steps for calculating the matching degree between contextual data and the specific contexts defined for all expected interaction nodes in the continuous business journey framework are as follows: First, align the real-time contextual data with the contextual parameters defined for the expected interaction nodes; extract the spatial, temporal, and device status features from the real-time context and compare them item by item with the node contextual features; calculate the similarity of each feature pair using a predefined metric function, such as inverse spatial distance or temporal overlap; weightedly fuse all feature similarities, with the weights preset by the feature importance; and output the matching degree value after normalization of the fusion result for node activation decision.

[0031] The optimal value of the preset activation threshold is dynamically optimized through a machine learning model; historical matching data and user response rates after activation at corresponding nodes are collected, and a logistic regression model is trained to predict the response probability; the matching value at the inflection point of the response probability curve is set as the baseline threshold; the baseline threshold is then multiplied by a risk coefficient based on the current marketing campaign type, and the threshold is lowered for high-risk campaigns to expand coverage; the threshold is recalibrated every period to adapt to changes in data distribution.

[0032] The steps for extracting the predefined initial business value weights and urgency indicators of candidate expected interactive nodes in the continuous business journey framework are as follows: the initial business value weights and urgency indicators are extracted from the node attribute library of the continuous business journey framework; the storage location of the target node is located by parsing the framework's metadata file; the weight field and the indicator field in the node attributes are read, with the weight field being numeric and the indicator field being enumerated; the data integrity is verified during the extraction process, and missing values ​​are filled with framework-level default values; the extraction results are cached to accelerate subsequent real-time processing.

[0033] The specific steps for adjusting the initial commercial value weights by combining the intensity values ​​of the corresponding intent dimensions in the real-time user short-term intent profile are as follows: The adjustment process uses the intensity value of the corresponding intent dimension in the user short-term intent profile as the adjustment factor; after normalizing the intensity value, it is multiplied by the initial commercial value weight to achieve linear adjustment; if multiple intent dimensions are related, the maximum intensity value is taken as the adjustment factor to avoid overweighting; the adjusted weights are then combined with the historical performance data of the nodes, and a performance decay factor is introduced for fine-tuning; the final weights reflect the real-time interest intensity of users and improve marketing accuracy.

[0034] Q5. Extract the top-ranked, predetermined number of candidate expected interaction nodes from the dynamic priority sequence as core activation nodes. Based on the pre-defined node connection rule base in the continuous business journey framework, find and bind feasible predecessor and successor nodes for each core activation node to form local behavior path segments with the core activation nodes as hubs. Integrate all local behavior path segments, check and repair the scene coherence at the junction of local behavior path segments to ensure the natural transition of user behavior logic, and finally assemble them into a complete and executable interaction path. Instantiate and encapsulate the definition context, content form, triggering conditions and execution order of the interaction path and all its core activation nodes to generate a currently active business journey instance that can directly drive marketing push.

[0035] The optimal value of the preset quantity is determined based on cognitive load theory and channel capacity balance; the average attention span is obtained through user surveys and used as the base quantity; the base quantity is compared with the throughput of the currently available communication channels, and the smaller value is taken to prevent information overload; at the same time, the complexity of business journey instances is considered, and the quantity is dynamically adjusted to maintain user engagement; the quantity is recalculated every week to adapt to environmental changes.

[0036] The node connection rule base defines feasible connection relationships between different types of candidate expected interactive nodes based on customer behavior path dependency and scenario coherence.

[0037] The steps for generating the local behavior path fragment in step Q5 are as follows: Q501. Based on the predefined interaction scenario requirements of the core activation node in the continuous business journey framework and the user's real-time scenario data, calculate the feasible interaction time window and content presentation mode boundary of the core activation node to form the basic interaction boundary. At the same time, combined with the user's current device status, network environment and attention level data, construct an interaction safety buffer outside the basic interaction boundary. The interaction safety buffer is used to accommodate the interaction delay or content adaptation deviation that may be caused by environmental fluctuations. The basic interaction boundary and the interaction safety buffer together constitute the complete interaction safety area of ​​the core activation node. Q502. Within the complete interactive safety area, taking the core activation node as the benchmark, explore forward the set of feasible preceding nodes that are logically consistent with its triggering conditions and contextually compatible, and explore backward the set of feasible successor nodes that are naturally connected with its execution results; based on the path dependency relationship and scenario coherence constraints defined in the preset node connection rule base, calculate the behavioral logic coherence score between each feasible preceding node and the core activation node, and between each feasible successor node and the core activation node; filter node pairs with behavioral logic coherence scores higher than the preset connectivity threshold, and connect the feasible preceding nodes and feasible successor nodes in the node pairs with the core activation node as the center to form multiple candidate local behavioral paths; The calculation method for behavioral logic coherence score is as follows: by evaluating the naturalness of behavioral transitions between nodes; first, a historical probability model of node behavior sequence is constructed to calculate the transition probability from the preceding node to the core activation node; the transition probability is weighted with the node context consistency score, which is generated by a semantic matching algorithm; the weighted result is then multiplied by the user behavior habit conformity, which is based on the frequency of historical paths; finally, the score is normalized to the range of 0-1, with a high score indicating smooth path.

[0038] Q503. Calculate the comprehensive efficiency index of each local behavior path. Based on the comprehensive efficiency index, sort all local behavior paths in descending order. Select the local behavior path with the highest comprehensive efficiency index as the optimal local behavior path. Optimize the topology of the optimal local behavior path to ensure smooth transitions between nodes without conflicts. Finally, generate local behavior path fragments that can be directly used for instantiation and encapsulation. The calculation steps for the comprehensive performance index of local behavioral pathways are as follows: The comprehensive performance index calculation integrates multiple performance indicators of local behavioral pathways; the indicators include pathway execution time, resource consumption rate, and user satisfaction prediction value; each indicator is normalized and then weighted and summed, with the weights allocated by the priority of business objectives; the summation result is then subtracted from the pathway risk coefficient, which is derived from the node failure history; the final index is used for pathway ranking, with high performance indicating high cost-effectiveness.

[0039] The method for constructing the interactive security buffer in step Q501 is as follows: Q5011. Based on the predefined interaction scenario requirements of the core activation node, extract its key scenario parameter thresholds, and combine them with device performance indicators, network bandwidth fluctuation range, and historical interaction success rate data in the user's real-time scenario data to calculate and generate the basic interaction boundary. The basic interaction boundary defines the minimum resource guarantee range that can stably support the interaction of the core activation node under the current technical conditions. At the same time, analyze the user's real-time attention level data and its changing trend, calculate the attention stability index, and calculate the network quality attenuation factor based on the real-time latency and packet loss rate data of the network environment. Multiply the attention stability index and the network quality attenuation factor to generate a dynamic compensation factor. The steps for generating the basic interaction boundary are as follows: based on the minimum resource requirements of the core active node; the requirements are extracted from the node definition, such as computing power, storage space and network latency threshold; the requirements are converted into spatiotemporal constraints to determine the minimum area required for interaction; the area boundary is represented by a geometric model, such as a rectangle or a circle, with the radius derived from the requirement parameters; the generation process verifies the feasibility of the requirements to avoid unrealistic constraints.

[0040] The attention stability index is calculated as follows: The attention stability index is calculated by analyzing the variation characteristics of the user's attention time series; recent attention value series are collected, and their standard deviation and autocorrelation coefficient are calculated; the reciprocal of the standard deviation is taken and multiplied by the autocorrelation coefficient to obtain the original stability value; the original value is then compared with the user's baseline attention level, which is defined by the historical average; the comparison result is normalized to generate the final index, and a high index indicates concentrated attention.

[0041] Q5012. Based on the basic interaction boundary, the dynamic compensation factor is used as the boundary expansion coefficient to isotropically expand the basic interaction boundary. The expanded area is the initial interaction security buffer. Subsequently, the compliance of the initial interaction security buffer with the user's current device hardware limitations, privacy settings policies, and platform service terms is verified. If there is a conflict, the buffer boundary is shrunk according to preset rules until it is fully compliant. Finally, the interaction security buffer is output. The interaction security buffer serves as the outer protection zone of the complete interaction security area of ​​the core activation node, providing additional fault tolerance space and adaptive adjustment leeway for interaction.

[0042] Preset rules: Includes conflict resolution strategies to adjust the interaction security buffer; rules are categorized by conflict type, such as device resource conflicts or privacy policy conflicts; each rule includes condition judgment and action execution, triggering a shrinking action when the condition is met; the shrinking action reduces the buffer size proportionally or by absolute value to ensure compliance with constraints; rule priority is set by the severity of the conflict, with severe conflicts being handled first.

[0043] Example The system collects real-time serialized status data from the user's smart refrigerator operation log, showing a milk inventory remaining at 30%. Simultaneously, it acquires real-time contextual data from the user's mobile device, showing their proximity to the supermarket and environmental parameters at 5 PM. This data is then timestamped to generate a standardized dataset. After parsing, multi-dimensional feature vectors are extracted, and a similarity score of 0.85 is calculated with historical records in the behavioral database. This similarity is weighted to generate a preliminary interest profile, which is then completed into a structured interest profile. The correlation strength between real-time contextual data and interest dimensions is calculated to be 0.9, with a reliability coefficient of 0.95. This is input into a graph neural network intent inference model, which outputs a short-term user intent profile. Based on business objectives, key anchor intents are selected and converted into sub-business objectives. Expected interaction nodes, such as store entry reminders, are planned within a preset 2-hour timeframe. The real-time context and node matching degree is calculated as follows: spatial distance inverse ratio is 0.8, temporal overlap is 0.9, and the weighted fusion matching degree is 0.85, exceeding the expected... Activation threshold is set at 0.75, and node activation is performed. An initial business value weight of 0.6 is extracted, adjusted to 0.54 based on user intent strength of 0.9, and a real-time urgency of 0.8 is calculated, resulting in a comprehensive priority coefficient of 0.432. The top-ranked core activation node is selected to generate a basic interaction boundary. A dynamic compensation factor of 0.63 is obtained by combining an attention stability index of 0.7 and a network quality decay factor of 0.9, expanding it into an interaction safety buffer. A behavioral logic coherence score of 0.88 exceeds the connectivity threshold of 0.8, forming a local behavioral path. The comprehensive efficiency index of 0.8 ranks best, generating local behavioral path fragments and assembling them into currently active business journey instances. After generating personalized marketing content, the push priority is normalized to 0.9. The expected utility value is calculated to 0.5355 based on a matching degree of 0.85, an inventory fulfillment rate of 1.0, a conversion rate of 0.7, and a channel efficiency coefficient of 0.9. The push instruction sequence reaches users through the app.

[0044] In step three, personalized marketing content includes: based on a preset dynamic content template library, and according to the defined context of the core activation node and the user's real-time context, dynamically generating multimodal marketing materials that include personalized product recommendations, contextualized promotional information, and interactive guidance content.

[0045] The pre-built dynamic content template library is a structured digital asset library that integrates successful historical marketing cases and multi-channel content paradigms. This library pre-configures multiple content templates for each type of core activation node. Each template defines the personalized product recommendation script structure, contextualized promotional information display elements, and interactive logic for engaging content. Key fields in the content templates are set as variables, dynamically bound to real-time user contextual state data (such as geographic location, device type, and time) and key dimensions in the user's short-term intent profile. When the system generates personalized marketing content, it retrieves matching content templates from the library based on the defined context of the core activation node and injects the real-time variable values ​​into the templates, thereby quickly generating multimodal marketing materials that fit the current scenario.

[0046] In step three, the formula for calculating the push priority and expected utility value is as follows: Based on the commercial value weight and real-time urgency of the core activation node, the two are multiplied to obtain the basic priority score; at the same time, a preset adjustment factor is introduced, and the basic priority score is multiplied by the adjustment factor to obtain the comprehensive priority score. The comprehensive priority scores of all content to be pushed are normalized to obtain the push priority. The current moment's user attention accessibility coefficient and the channel instantaneous reach rate coefficient are used as adjustment factors; The matching degree between personalized marketing content and users' real-time intent is calculated. The matching degree is obtained by analyzing the overlap between the content keywords and the intent dimensions in the user's short-term intent profile where the confidence coefficient exceeds the preset activation threshold. The real-time inventory satisfaction rate and expected conversion rate of the products or services recommended by the personalized marketing content are evaluated. The matching degree, real-time inventory satisfaction rate and expected conversion rate are multiplied to obtain the original utility estimate. A channel efficiency coefficient based on the historical interaction success rate of the push channel is introduced. The original utility estimate is multiplied by the channel efficiency coefficient, and the result is standardized to generate the expected utility value.

[0047] In step three, the preset push decision engine is a hybrid decision system based on a multi-objective optimization algorithm and a preset real-time rule evaluation module. The hybrid decision system receives the push priority and expected utility value, and generates the final executable push instruction sequence through dynamic weight allocation, conflict resolution and channel load balancing calculation.

[0048] In step four, after the push instruction sequence is executed, the user's feedback behavior to the push content is monitored, and the feedback behavior is sent back to step one as new contextual data. At the same time, the status of the currently active business journey instance is updated to trigger the adaptive adjustment of subsequent nodes or the initialization of a new journey.

[0049] Steps three and four utilize a pre-set dynamic content template library to quickly and accurately generate marketing content. Combined with multi-factor calculation of push priority and expected utility value, intelligent decision-making is achieved. Finally, relying on a hybrid push decision engine and feedback mechanism, the personalization of marketing content, the accuracy of push timing, and the overall cross-channel collaboration efficiency are significantly improved, thereby optimizing user experience and increasing marketing conversion rates.

[0050] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A personalized and precise marketing push method based on the Internet of Things, characterized in that: Includes the following steps: Step 1: Collect serialized status data and contextual data from user-associated IoT terminal devices in real time, and perform fusion analysis on the collected serialized status data and contextual data to generate a short-term user intent profile. Step 2: Based on the user's short-term intent profile and combined with preset business goals, construct a continuous business journey framework that includes expected interaction nodes; based on the real-time acquired contextual data, dynamically activate or adjust specific business journey nodes in the continuous business journey framework that match the current user's physical and digital context, and generate currently active business journey instances. Step 3: For each active node in the currently active business journey instance, based on the user's short-term intent profile and the real-time contextual data, generate personalized marketing content related to the context of the active node, calculate the push priority and expected utility value of each personalized marketing content in the current context, and generate a push instruction sequence based on the push priority and expected utility value through a preset push decision engine. Step 4: Execute the push instruction sequence to deliver the personalized marketing content to the user through a preset communication channel.

2. The personalized and precise marketing push method based on the Internet of Things according to claim 1, characterized in that: The IoT terminal devices in step one include smart home appliances and wearable devices for collecting serialized status data, as well as vehicle-mounted devices and user mobile devices for collecting contextual data. The serialized status data includes smart home device operation logs and records of changes in product inventory status; The contextual data includes the real-time geographic location of the user's mobile device and environmental parameters.

3. The personalized and precise marketing push method based on the Internet of Things according to claim 2, characterized in that: The steps for generating the user's short-term intent profile in step one are as follows: S1. Real-time collection of the serialized status data and context data from different IoT terminal devices; timestamp alignment, format conversion and outlier cleaning of the collected smart home device operation logs, product inventory status change records, real-time geographical location and environmental parameters of user mobile devices; and generation of a standardized data set. S2. The standardized data set is parsed to extract key features reflecting user behavior patterns, device interaction status and environmental context, forming a multi-dimensional feature vector. Based on the multi-dimensional feature vector, combined with the user's past behavior habits and preference records retrieved from the preset behavior database, a primary interest profile representing the user's current interest tendency is calculated and generated. The primary interest profile is compared and completed with the preset basic user profile template to output a structured interest profile. S3. Receive the structured interest profile and simultaneously acquire the latest real-time contextual data. Calculate the correlation strength between the real-time contextual data and each interest dimension in the structured interest profile. Assign a preset confidence coefficient to each interest dimension based on the correlation strength. Input the structured interest profile with the preset confidence coefficient into a preset intent inference model and output a short-term user intent profile.

4. The personalized and precise marketing push method based on the Internet of Things according to claim 3, characterized in that: The method for generating the continuous business journey framework in step two is as follows: Q1. Based on the preset business goal, create an initial journey framework that includes the starting state and the target state but with no interaction nodes. From the user short-term intent profile, select the core user intents that belong to the same category or have a strong correlation with the business goal under the preset business classification system, and use them as key anchor intents. Inject the key anchor intents into the initial journey framework. Q2. Based on the preset conversion strategy, convert each of the selected core user intentions into sub-business objectives; Q3. Based on the target type, required resources, and estimated time of each sub-business objective, plan expected interaction nodes within a preset time period in the future. Each expected interaction node defines the specific context, content format, and triggering conditions for interaction with the user. According to the natural evolution of user behavior over time and space, arrange all the generated expected interaction nodes into a node sequence according to their contextual logic and triggering conditions. Place the node sequence into and fill the blank interaction node part in the initial journey framework to form a complete path that starts from the current context, runs through multiple sub-business objectives, and ultimately leads to the preset business objective. The complete path is the continuous business journey framework.

5. The personalized and precise marketing push method based on the Internet of Things according to claim 4, characterized in that: The method for generating the currently active business journey instance in step two is as follows: Q4. Acquire the latest contextual data in real time, calculate the matching degree between it and the specific context defined by all the expected interaction nodes in the continuous business journey framework, retain only the candidate expected interaction nodes whose matching degree exceeds the preset activation threshold, and form a set of candidate expected interaction nodes to be activated. For each candidate expected interaction node in the set of candidate expected interaction nodes to be activated, simultaneously extract its predefined initial business value weight and urgency indicator in the continuous business journey framework, combine it with the intensity value of the corresponding intent dimension in the user short-term intent profile acquired in real time, correct the initial business value weight, generate a business value weight, and calculate the real-time urgency based on the urgency indicator and the time constraint strength of the current context; multiply the business value weight and the real-time urgency to obtain the comprehensive priority coefficient of each candidate expected interaction node, and sort all the candidate expected interaction nodes in the set of candidate expected interaction nodes to be activated in descending order according to the comprehensive priority coefficient to generate a dynamic priority sequence. Q5. Extract the top-ranked, preset number of candidate expected interactive nodes from the dynamic priority sequence as core activation nodes. Based on the preset node connection rule base in the continuous business journey framework, find and bind feasible predecessor and successor nodes for each core activation node to form a local behavior path segment with the core activation node as the hub. Integrate all the local behavior path segments, check and repair the scene coherence at the junction of the local behavior path segments, assemble them into an interactive path, and instantiate and encapsulate the definition context, content form, triggering conditions and execution order of the interactive path and all the core activation nodes to generate a currently active business journey instance.

6. The personalized and precise marketing push method based on the Internet of Things according to claim 5, characterized in that: The steps for generating the local behavior path fragment in step Q5 are as follows: Q501. Based on the predefined interaction scenario requirements of the core activation node in the continuous business journey framework and the real-time scenario data of the user, calculate the feasible interaction time window and content presentation mode boundary of the core activation node to form a basic interaction boundary. At the same time, combined with the user's current device status, network environment and attention level data, construct an interaction safety buffer outside the basic interaction boundary. The basic interaction boundary and the interaction safety buffer together constitute the complete interaction safety area of ​​the core activation node. Q502. Within the complete interactive safety area, taking the core activation node as a benchmark, explore forward a set of feasible preceding nodes that are logically consistent with its triggering conditions and contextually compatible, and explore backward a set of feasible succeeding nodes that are naturally connected with its execution results; based on the path dependency relationship and scenario coherence constraints defined in the preset node connection rule base, calculate the behavioral logic coherence score between each feasible preceding node and the core activation node, and between each feasible succeeding node and the core activation node. Node pairs whose behavioral logic coherence scores are higher than a preset connectivity threshold are selected, and the feasible preceding nodes and feasible succeeding nodes in the node pairs are connected with the core activated node as the center to form multiple candidate local behavioral paths. Q503. Calculate the comprehensive efficiency index of each of the local behavioral paths, arrange all the local behavioral paths in descending order based on the comprehensive efficiency index, select the local behavioral path with the highest comprehensive efficiency index as the optimal local behavioral path, optimize the topology of the optimal local behavioral path, and generate local behavioral path segments.

7. The personalized and precise marketing push method based on the Internet of Things according to claim 6, characterized in that: The personalized marketing content in step three includes: based on a preset dynamic content template library, and according to the defined context of the core activation node and the user's real-time context state, dynamically generating multimodal marketing materials that include personalized product recommendations, contextualized promotional information, and interactive guidance content.

8. The personalized and precise marketing push method based on the Internet of Things according to claim 7, characterized in that: The calculation formulas for the push priority and the expected utility value in step three are as follows: Based on the commercial value weight and the real-time urgency of the core activation node, the two are multiplied to obtain the basic priority score; at the same time, a preset adjustment factor is introduced, and the basic priority score is multiplied by the adjustment factor to obtain the comprehensive priority score. The comprehensive priority scores of all content to be pushed are normalized to obtain the push priority. The matching degree between the personalized marketing content and the user's real-time intent is calculated. The matching degree is obtained by analyzing the overlap between the content keywords and the intent dimensions in the user's short-term intent profile where the confidence coefficient exceeds a preset activation threshold. The real-time inventory satisfaction rate and expected conversion rate of the products or services recommended by the personalized marketing content are evaluated. The matching degree, the real-time inventory satisfaction rate, and the expected conversion rate are multiplied to obtain the original utility estimate. A channel efficiency coefficient based on the historical interaction success rate of the push channel is introduced. The original utility estimate is multiplied by the channel efficiency coefficient, and the result is standardized to generate the expected utility value.

9. The personalized and precise marketing push method based on the Internet of Things according to claim 8, characterized in that: The preset push decision engine in step three is specifically a hybrid decision system based on a multi-objective optimization algorithm and a preset real-time rule evaluation module.

10. A personalized and precise marketing push method based on the Internet of Things according to claim 9, characterized in that: In step four, after the push instruction sequence is executed, the user's feedback behavior to the push content is monitored, and the feedback behavior is sent back to step one as new contextual data.

Images