Intelligent offline advertisement putting system based on big data and AI

Abstract

The invention discloses an offline advertisement intelligent putting system based on big data and AI, and relates to the technical field of directional advertisement optimization, and the system comprises a causal decision module which constructs a dynamic causal map based on a panoramic situation data flow, carries out the intervention effect analysis, and generates a quantum decision vector; the quantum coding module is used for executing quantum state probability amplitude mapping and entanglement particle binding on the quantum decision vector to generate a quantum decision proposal packet; the collaborative optimization module is used for constructing a global quantum entangled state model, carrying out conflict resolution optimization on the quantum decision proposal package and generating a quantum collaborative execution instruction; and the report generation module is used for executing quantum state collapse operation on the quantum cooperative execution instruction and generating an intelligent putting report. According to the invention, by constructing the dynamic causal atlas and quantum collaborative architecture, normal form upgrading of advertisement decision from correlation analysis to causal intervention is realized, so that the system has the capability of accurately quantifying the real effect of advertisement action and millisecond-level dynamic response.

Description

Technical Field

[0001] This invention relates to the field of targeted advertising optimization technology, and in particular to an intelligent offline advertising delivery system based on big data and AI. Background Technology

[0002] In recent years, with the evolution of big data and artificial intelligence technologies, offline advertising delivery systems have gradually shifted from static display to dynamic optimization. Existing technologies primarily achieve ad placement matching through programmatic buying platforms combined with user profiling models. At the data processing layer, mainstream solutions employ distributed stream processing frameworks to integrate environmental sensor data, POS transaction logs, and customer flow trajectories collected by WiFi probes, generating delivery strategies through ensemble learning algorithms such as random forests. These technologies significantly improve the breadth of ad exposure and partially achieve contextualized reach.

[0003] However, existing technologies have inherent flaws in causal decision-making and global resource coordination. Multi-ad terminal collaboration relies on classical optimization algorithms, whose solution space search efficiency decreases exponentially with node size. When simultaneously optimizing energy consumption, conversion rate, and bidding costs, the NP-hard problem causes the strategy to converge to a suboptimal solution. For example, edge nodes reducing ad rotation frequency to lower power consumption fundamentally conflicts with the cloud's goal of improving conversion rates. Existing collaboration mechanisms can only compromise objectives through weighted summation, failing to achieve Pareto optimality, and their response latency exceeds the business decision window. Summary of the Invention

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

[0005] Therefore, this invention provides an intelligent offline advertising delivery system based on big data and AI to solve the problem of insufficient advertising delivery accuracy caused by causal confusion and distributed decision-making conflicts in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides an intelligent offline advertising delivery system based on big data and AI, comprising: The data acquisition module collects environmental data, user behavior data, business activity data, competitive advertising status, equipment operating parameters, and sales data in real time, performs preprocessing, and outputs a panoramic situational data stream. The causal decision-making module, based on panoramic situational data streams, constructs dynamic causal graphs and performs intervention effect analysis to generate quantum decision vectors; The quantum encoding module performs quantum state probability amplitude mapping and binding with entangled particles on the quantum decision vector to generate a quantum decision proposal package; The collaborative optimization module constructs a global quantum entangled state model, optimizes the quantum decision proposal package by resolving conflicts, and generates quantum collaborative execution instructions. The report generation module performs quantum state collapse operations on quantum cooperative execution instructions to generate intelligent delivery reports.

[0007] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the environmental data includes temperature and humidity values, light intensity and direction, weather warning index, air quality index, and noise decibel value. The user behavior data includes gaze focus coordinates, gait speed, dwell time heatmap, device signal strength distribution, and emotion recognition index; The business activity data includes promotional semantic tags, discount intensity values, brand competition index, inventory turnover rate, and member consumption preference tags; The competitive advertising status includes competitor advertising content, display time percentage, screen brightness value, interaction response rate, and bid consumption rate; The device operating parameters include advertising screen power consumption, network latency, quantum channel bit error rate, fault diagnosis code, and heat sink temperature. The sales data includes real-time transaction volume, product conversion correlation, inventory changes, average order value fluctuations, and return rate.

[0008] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the preprocessing includes data cleaning, spatiotemporal feature extraction, and Z-score normalization.

[0009] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the specific steps for constructing a dynamic causal graph based on panoramic situational data flow are as follows: Perform association rule mining and entity classification on the panoramic situational data stream to generate a set of causal nodes and a set of candidate causal pairs. Perform FCI conditional independence test and do operator strength calculation on the causal node set and candidate causal pair set to generate a dynamic causal graph with edge weight values.

[0010] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the specific steps for generating the quantum decision vector are as follows: Perform causal effect quantification on the dynamic causal graph to generate candidate decision triples; Candidate decision triples are structured and quantum-encoded and bound to causal paths to generate quantum decision vectors.

[0011] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the specific steps for performing quantum state probability amplitude mapping on the quantum decision vector are as follows: Perform causal-probability transformation and normalization verification on the quantum decision vector to generate a probability amplitude allocation table; The probability amplitude allocation table is dynamically labeled with phase angle and encapsulated with quantum state parameters to generate a quantum state parameter table.

[0012] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the specific steps for generating the quantum decision proposal package are as follows: Quantum entangled particle pairs are assigned to the quantum state parameter table to generate a quantum entangled registry; Quantum-safe device binding is performed on the quantum entanglement registry to generate a quantum decision proposal package.

[0013] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the specific steps for constructing a global quantum entangled state model, optimizing the quantum decision proposal package through conflict resolution, and generating quantum cooperative execution instructions are as follows. A global quantum entangled state model is constructed based on the edge node layer, quantum channel layer, and cloud collaboration layer. The edge node layer adapts quantum state parameters to the quantum decision proposal package to generate a set of device-executable quantum instructions. Quantum state synchronization transmission and entanglement verification are performed on the quantum instruction set executable by the device to generate quantum cooperative control signals; The cloud-based collaboration layer performs global strategy optimization and quantum signature encapsulation on the quantum collaborative control signal to generate quantum collaborative execution instructions.

[0014] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the specific steps of performing quantum state collapse operation on the quantum collaborative execution instruction are as follows: Quantum state collapse measurement and execution state verification are performed on quantum cooperative execution instructions, and a quantum state execution feedback report is generated. The quantum state execution feedback report is subjected to full data fusion analysis and execution effect quantification to generate an advertising effect tracking report.

[0015] As a preferred embodiment of the offline advertising intelligent delivery system based on big data and AI described in this invention, the generation of quantum collaborative execution instructions refers to performing attribution analysis and visualization encapsulation on the advertising effect tracking report to generate an intelligent delivery report.

[0016] The beneficial effects of this invention are as follows: by constructing a dynamic causal graph and a quantum collaborative architecture, a paradigm upgrade of advertising decision-making from correlation analysis to causal intervention is realized, enabling the system to accurately quantify the real effect of advertising actions and provide millisecond-level dynamic response. Through global quantum entanglement state modeling and conflict resolution optimization, ultra-distance collaboration of distributed nodes and optimal resource allocation are realized, thus constructing a highly robust advertising delivery network. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of an intelligent offline advertising delivery system based on big data and AI.

[0019] Figure 2 This is a flowchart for the causal decision-making module.

[0020] Figure 3 This is a flowchart for quantum encoding and collaborative optimization.

[0021] Figure 4 This is a schematic diagram of a global quantum cooperative architecture. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides an intelligent offline advertising delivery system based on big data and AI, including the following steps: The data acquisition module collects environmental data, user behavior data, business activity data, competitive advertising status, equipment operating parameters, and sales data in real time, performs preprocessing, and outputs a panoramic situational data stream. The first step involves environmental data including temperature and humidity values, light intensity and direction, weather warning index, air quality index, and noise level in decibels. It should be noted that temperature and humidity values ​​are measured in real time using digital temperature and humidity sensors, and the output of Celsius temperature and relative humidity percentage is based on calibration data from the HygroFlex series probes; light intensity and direction are collected using a combination of LUX-meter and compass sensors, with intensity values ​​in lux and direction angles recorded from 0 to 360 degrees; the meteorological warning index is connected to the China Meteorological Administration API interface and mapped to levels 1-5 according to disaster types such as rainstorms and high temperatures; the air quality index is generated by mapping the measured concentrations of pollutants such as PM2.5, PM10, and SO2 to internationally accepted index classification standards using piecewise linear interpolation of each pollutant concentration; and noise decibel values ​​are obtained using an A-weighted sound level meter to acquire the equivalent continuous sound pressure level within a 1-second sampling period, with a data range of 30-120 dB(A).

[0026] The second step involves user behavior data, including gaze focus coordinates, gait speed, dwell time heatmap, device signal strength distribution, and emotion recognition index. It should be noted that the gaze focus coordinates are obtained by capturing the user's iris position using a binocular camera and outputting pixel coordinates in the screen coordinate system; gait speed is calculated using LiDAR point cloud sequence analysis, estimating the movement speed value through the displacement of joint points in the human skeleton across consecutive frames; the dwell time heatmap integrates WiFi probe signal strength and infrared sensor data to generate a rasterized heat distribution matrix; the device signal strength distribution records the strength values ​​of surrounding Bluetooth and WiFi signals, categorizing and statistically analyzing feature parameters according to device identification; the emotion recognition index analyzes the activation intensity and skin conductance waveform of 52 facial muscle action units using a pre-trained convolutional neural network, extracting multi-scale spatiotemporal features and then regressing them through a fully connected layer to output a standardized emotion score in the 0-1 range, where 0 represents completely negative emotion and 1 represents completely positive emotion. All data is time-synchronized and encapsulated in a structured format, including the collection time, spatial location, and data source identifier.

[0027] The third step involves business activity data, including promotional semantic tags, discount levels, brand competition index, inventory turnover rate, and member consumption preference tags. It should be noted that promotional semantic tags extract key promotional information such as "buy one get one free" and "limited-time discount" by parsing mall broadcasts and poster text, and store them in categories; the discount strength value records the percentage difference between the original price and the promotional price of the product, retaining two decimal places of precision; the brand competition index counts the display frequency and time period proportion of competitor advertisements in the same category, calculated comprehensively by exposure duration and position weight; the inventory turnover rate is updated regularly according to the formula "sales quantity / average inventory" based on the time-series data of product warehousing and sales records; the member consumption preference tag integrates historical purchase records and browsing behavior to generate structured classification labels such as "preference for maternal and infant products" and "preference for high-end cosmetics". All data is linked to product IDs through timestamps to form a complete record of commercial activities with spatiotemporal attributes.

[0028] The fourth step is to analyze the competitive advertising status, including competitor ad content, display time percentage, screen brightness, interaction response rate, and bid burn rate. It should be noted that competitor advertising content is analyzed by extracting visual elements from adjacent ad screens, including brand logos, product images, and promotional copy; the display time ratio is recorded as the proportion of the cumulative display time of competitor ads within a unit of time to the total time period; screen brightness is measured using a photometer to measure the nit value of competitor ad screens, distinguishing the brightness characteristics of static images and dynamic videos; interaction response rate is calculated as the ratio of the number of touchscreen operations to the number of impressions of competitor ads; and bid burn rate is monitored to track the speed at which competitors consume their budget in real-time bidding ad slots, recording changes in budget at the minute level. All data is indexed using ad slot geocoding and timestamps, forming a structured log with competitive landscape markers.

[0029] Step 5: Equipment operating parameters include advertising screen power consumption, network latency, quantum channel bit error rate, fault diagnosis codes, and heat sink temperature; It should be noted that the power consumption of the advertising screen is monitored in real time by a smart meter, which measures the wattage of the display unit to distinguish the energy consumption differences between static images and dynamic video playback. Network latency is recorded as the millisecond-level time difference between the instruction issuance and the advertising screen's response, including both transmission and processing delay components. The quantum channel bit error rate is statistically analyzed to show the ratio of erroneous bits to the total number of transmitted bits during quantum key distribution, conforming to communication protocol standards. Fault diagnosis codes are based on a fault coding manual, identifying fault types such as power abnormalities and signal loss. Heat sink temperature is measured in Celsius using surface-mount thermocouples. All parameters are organized by device serial number and timestamp, forming a time-series data stream with status markers.

[0030] Step 6: Sales data includes real-time transaction volume, product conversion correlation, inventory changes, average order value fluctuations, and return rate; It should be noted that the real-time transaction log records every transaction generated by the POS terminal, including the product code, transaction time, and amount; the product conversion correlation statistics show the ratio of the number of purchases of a specific product after advertising exposure to the number of exposures; inventory change monitoring tracks the increase or decrease in inventory quantity due to product warehousing and sales, updating the current inventory value on an hourly basis; average order value fluctuation analysis shows the standard deviation and mean change of the customer's single transaction amount within a unit of time; and the return rate statistics show the proportion of orders with returned goods, recorded by product category. All data is linked through transaction IDs and timestamps to form a complete sales behavior tracking record.

[0031] Step 7: Preprocessing includes data cleaning, spatiotemporal feature extraction, and Z-score normalization.

[0032] It should be noted that after environmental data, user behavior data, business activity data, competitive advertising status, equipment operating parameters, and sales data are acquired through distributed acquisition nodes, data cleaning is first performed: sensor drift data in temperature and humidity values ​​are eliminated using sliding window filtering; invalid coordinate points in line-of-sight focus coordinates are corrected using Kalman filtering; and ambiguous text in promotional semantic tags is standardized using natural language processing techniques. Subsequently, spatiotemporal feature extraction is performed: data from different acquisition frequencies (such as line-of-sight focus coordinates at 200Hz and temperature and humidity values ​​at 1Hz) are unified to a 10Hz time base using cubic spline interpolation; and geographic coordinates are converted to a Cartesian coordinate system using UTM projection. Z-score normalization is implemented separately for each data type: environmental data is based on the historical average of each sensor; user behavior data is standardized by user ID grouping; and business activity data is normalized by product category.

[0033] The causal decision-making module, based on panoramic situational data streams, constructs dynamic causal graphs and performs intervention effect analysis to generate quantum decision vectors; The first step is to perform association rule mining and entity classification on the panoramic situational data stream to generate a set of causal nodes and a set of candidate causal pairs. It should be noted that the panoramic situational data stream identifies stable association patterns between environmental data fields such as temperature and humidity values ​​and user behavior data fields such as residence heatmaps by analyzing the co-occurrence frequency and time-series correlation between data fields (a stable association pattern refers to a co-occurrence frequency or time-series correlation value between data fields that is consistently higher than the 95th percentile of the historical data statistical distribution within a continuous observation period). It then filters association rules that meet the minimum support and confidence conditions (minimum support is determined by statistically analyzing the distribution of field co-occurrence frequencies in historical data, using the 95th percentile as a threshold, representing the minimum frequency requirement for a rule to appear in the dataset; minimum confidence is determined by analyzing the distribution of conditional probabilities in historical data, using the 90th percentile as a threshold). Percentiles serve as the threshold, representing the minimum standard for the reliability of association rule inferences. Entity classification follows a predefined classification system, categorizing meteorological warning indices in environmental data as weather entities, gaze focus coordinates in user behavior data as user attention entities, and promotional semantic tags in commercial activity data as marketing entities. The causal node set consists of the classified entities, with each node clearly labeled with its entity type, data source, and valid time range. The candidate causal pair set is determined by examining the lead-lag relationship of time series, retaining entity combinations with temporal correlation, such as entity pairs where changes in meteorological warning indices lead changes in dwell heatmap density. Finally, the causal node set and the candidate causal pair set are generated.

[0034] It should be noted that the specific process of predefining entity classification is as follows: by analyzing the naming characteristics, numerical distribution range and co-occurrence patterns of historical data fields, a clustering algorithm is used to automatically summarize three major entity categories: environment, behavior and business. The classification criteria for each category are determined by the similarity measure of data features.

[0035] The second step involves performing the FCI conditional independence test and calculating the strength of the do operator on the causal node set and the candidate causal pair set to generate a dynamic causal graph with edge weights. The expression is as follows: ; in, For the first in the dynamic causal graph One cause node, For the first in the dynamic causal graph Result node, for Control of latent variables at any given time for Time, Reason, Node To the result node edge weight values, To set cause nodes for proactive intervention And control the hidden variables Under the conditions, The probability of occurrence To control hidden variables Under the conditions, The probability of occurrence; It should be noted that, firstly, for each pair of candidate causal nodes ( , In a given set of latent variables Conditional independence tests are performed under the given conditions, and statistical methods are used to assess the degree of deviation between the observed data and the independence hypothesis. When the test results meet the predetermined confidence level, the hypothesis is considered valid. and A conditional dependency exists. Causal pairs that pass the test proceed to the do operator strength analysis stage, where the probability distribution under the intervention state is determined using a backdoor adjustment method. Probability distribution under natural conditions The arithmetic difference between the two is used as the weight value. And record it. The generated dynamic causal graph is stored in a directed acyclic graph structure, with each edge labeled with its source node. Target node and weight values The node attributes include the entity type and data source identifier.

[0036] It should be noted that the predetermined confidence level is set based on error rate analysis of historical experimental data, with an example value range of (0.05, 0.01). The backdoor adjustment method uses a stratified statistical set of latent variables. A standardized causal inference method that uses different combinations of values ​​to eliminate the influence of confounding factors on causal relationship estimation.

[0037] The third step is to perform causal effect quantification on the dynamic causal graph and generate candidate decision triples. It should be noted that the edge weight values ​​in the dynamic causal graph... Causal paths exceeding a preset causal threshold are extracted as valid causal relationships and quantified into triples through causal effects: advertising actions (such as "advertising coffee") originate from cause nodes. The entity type, the target location (e.g., "LED screen A3"), and the corresponding result node. The spatial attributes and time window (e.g., "09:00-11:00") are calculated based on the duration of the causal effect. Each candidate decision triple carries the weight values ​​of the original causal path. The system calculates timestamps to ensure that decisions can be traced back to specific edge relationships in the dynamic causal graph. Triples are arranged in descending order of effect size to form a priority decision queue.

[0038] It should be noted that the causal threshold is set based on the normal distribution characteristics of historical causal effect values, with an exemplary value range of (0.55, 0.65).

[0039] The fourth step is to perform structured quantum encoding and causal path binding on the candidate decision triples to generate quantum decision vectors.

[0040] It should be noted that the advertising action, target location, and time window fields in the candidate decision triple are transformed through qubit encoding: the advertising action is mapped to a superposition of the ground state |0> and the excited state |1>, the target location is converted into the quantum phase angle of three-dimensional spatial coordinates, and the time window is encoded as the parameters of the time evolution operator; the causal path binding assigns the weight values ​​of the corresponding edges in the original dynamic causal graph. As the probability amplitude coefficient, it is encapsulated synchronously with the quantum state, and the generated quantum decision vector contains three parts: quantum state parameters (probability amplitude and phase), causal path identifier (source node), and so on. With the target node The hash value) and spatiotemporal constraints (geographic coordinate range and time validity stamp).

[0041] The quantum encoding module performs quantum state probability amplitude mapping and binding with entangled particles on the quantum decision vector to generate a quantum decision proposal package; The first step is to perform causal-probability transformation and normalization verification on the quantum decision vector to generate a probability amplitude allocation table; It should be noted that the quantum state parameters in the quantum decision vector are converted into classical probability values ​​through quantum measurement operations, and the selection probability of the advertising action is obtained by projection measurement of the quantum superposition state; the causal path identifier is retained as metadata to ensure that the probability value remains associated with the edge weights of the dynamic causal graph; normalization verification confirms that the sum of the probability values ​​equals 1. The generated probability amplitude allocation table contains the advertising action identifier, probability value, and causal path hash value, with all values ​​retained to four decimal places.

[0042] The second step is to dynamically label the probability amplitude allocation table with phase angles and encapsulate quantum state parameters to generate a quantum state parameter table. It should be noted that the probability values ​​in the probability amplitude allocation table are converted into phase angle parameters, and the selection probability of an advertising action corresponds to a specific phase angle. The quantum state parameter encapsulation combines the probability values ​​and phase angles into a complete quantum state description. For example, the quantum state descriptions of coffee ads and milk tea ads include their respective probability values ​​and phase angle parameters. The generated quantum state parameter table contains three fields: advertising action identifier, probability value, and phase angle. All values ​​retain four decimal places of precision.

[0043] The third step is to allocate quantum entangled particle pairs to the quantum state parameter table and generate a quantum entangled registry. It should be noted that the advertising action identifier in the quantum state parameter table is assigned to entangled particle pairs through a quantum entanglement generator, and each advertising action is bound to a unique entangled particle identifier (such as QID_88923). The probability amplitude and phase angle parameters in the quantum state description are synchronously associated with the corresponding entangled particle pairs, forming a quantum entanglement registry containing the advertising action identifier, entangled particle identifier, probability amplitude, and phase angle. All fields conform to the quantum key distribution protocol specification, the entangled particle identifier is generated using SHA-256 hashing to ensure uniqueness, and the probability amplitude and phase angle values ​​retain four decimal places of precision.

[0044] The fourth step is to bind the quantum entanglement registry to quantum-safe devices and generate a quantum decision proposal package.

[0045] It should be noted that the entangled particle identifier in the quantum entanglement registry is bound to the physical address of the target advertising screen through a quantum key distribution protocol. The advertising action identifier, probability amplitude, and phase angle parameters are transmitted after being encrypted by a quantum channel. The quantum security device binding process generates a quantum decision proposal package containing the entangled particle identifier, the device physical address, encrypted quantum state parameters, and a timestamp.

[0046] The collaborative optimization module constructs a global quantum entangled state model, optimizes the quantum decision proposal package by resolving conflicts, and generates quantum collaborative execution instructions. The first step is to construct a global quantum entangled state model based on the edge node layer, quantum channel layer, and cloud collaboration layer; It should be noted that the quantum decision proposal package deployed at the edge node layer is transmitted through the fiber optic network of the quantum channel layer and aggregated into quantum states at the cloud collaboration layer. The quantum channel layer uses an entangled particle distribution protocol to establish quantum correlations between nodes, forming a cross-device Bell state entanglement network. The cloud collaboration layer uses quantum gate operations to fuse the quantum state parameters of each node into a global superposition state, constructing a global quantum entangled state model that includes all advertising action options and their probability amplitudes and phase angles.

[0047] The second step involves the edge node layer adapting the quantum decision proposal package to quantum state parameters to generate a set of quantum instructions that can be executed by the device. It should be noted that the quantum state parameters in the quantum decision proposal package are converted into an executable instruction format for the target advertising screen through a device characteristic matching process: the probability amplitude is adapted to the screen brightness level (e.g., a probability amplitude of 0.7 corresponds to a brightness of 300 nits), the phase angle is mapped to the content switching interval (e.g., a phase of 0.1π corresponds to a 2-second carousel cycle), and the entangled particle identifier is bound to the screen controller MAC address. The generated device-executable quantum instruction set includes advertising content identifiers, brightness parameters, timing parameters, and device binding identifiers.

[0048] The third step is to perform quantum state synchronization transmission and entanglement verification on the quantum instruction set executable by the device to generate quantum cooperative control signals. It should be noted that the device can execute a set of quantum instructions, which are transmitted to the target device through a quantum communication channel. The quantum state parameters (probability amplitude and phase angle) are synchronized across devices through entangled photon pairs to ensure that the instruction states received by each node are consistent. The entanglement verification process monitors the bit error rate and entanglement fidelity of the qubits to verify that the quantum state has not decohered or been tampered with during the transmission process. The final generated quantum cooperative control signal (synchronization confirmation flag, channel quality index, and device execution parameters) is used to verify that the quantum state has not been decohered or tampered with.

[0049] The fourth step involves the cloud-based collaborative layer performing global strategy optimization and quantum signature encapsulation on the quantum collaborative control signal to generate quantum collaborative execution instructions.

[0050] It should be noted that the cloud-based collaboration layer performs global strategy optimization on the quantum collaboration control signal. The advertising revenue objective determines priorities based on historical click-through rate and conversion rate data for each ad slot. The resource competition objective assesses conflict intensity by analyzing the overlap of quantum states of adjacent ad slots; that is, it obtains the descriptive parameters of the current quantum states of each ad slot and then compares the matching degree of these quantum state feature vectors. When the similarity exceeds a preset conflict judgment threshold, a resource competition conflict is confirmed. The user experience objective sets constraints based on user fatigue thresholds and visual comfort standards. The quantum signature encapsulation process uses a quantum hash function to process the optimization strategy parameters, generating an unforgeable instruction identifier, and finally generating a quantum collaborative execution instruction (containing three parts: optimization strategy parameters, quantum state synchronization requirements, and quantum signature).

[0051] It should be noted that the conflict determination threshold is set based on the correlation analysis of the quantum state similarity of ad slots and the decrease in conversion rate in historical data, with an example value range of (0.65, 0.75).

[0052] The user fatigue threshold is determined by analyzing the user attention decay curve in historical behavior data, with an example value range of (30, 45).

[0053] Visual comfort standards are determined by combining objective physiological data such as pupil diameter changes and blink frequency collected from subjects in a laboratory environment with correlation analysis of physical parameters such as brightness and color temperature of advertising screens and user behavior feedback.

[0054] The constraints refer to the user experience parameter boundaries that must be followed during the advertising process, including the user fatigue threshold (maximum exposure time of a single ad) and the visual comfort standard (brightness / color temperature range).

[0055] The report generation module performs quantum state collapse operations on quantum cooperative execution instructions to generate intelligent delivery reports.

[0056] The first step is to perform quantum state collapse measurement and execution state verification on the quantum cooperative execution instructions, and generate a quantum state execution feedback report; It should be noted that the quantum cooperative execution instruction achieves state collapse through quantum projection measurement operations, converting the quantum superposition state of the advertising strategy into a deterministic execution action (such as "coffee advertisement @ screen A3"). The execution state verification process monitors the consistency between the actual advertising screen display content and the quantum instruction, while recording the quantum channel bit error rate and device response latency, and generating a quantum state execution feedback report (collapse result (measured data of advertising content and location), quantum fidelity (matching degree between instruction state and execution state), device status log (power consumption, network latency, and other parameters)).

[0057] The second step is to perform full data fusion analysis and execution effect quantification on the quantum state execution feedback report to generate an advertising effect tracking report. It should be noted that the collapse results, quantum fidelity, and device status log data in the quantum state execution feedback report are fused and analyzed through spatiotemporal alignment and feature correlation: the advertising content identifier in the collapse results is matched with POS transaction records to form a conversion rate indicator; quantum fidelity data is used to correct the attribution weight of advertising effectiveness; and the network latency parameter in the device status log is correlated with the user interaction response rate. The execution effect quantification process forms key indicators: advertising exposure is verified through camera traffic statistics; conversion rate is based on the correlation between scanning behavior and transaction data; user engagement is evaluated based on dwell time and interaction frequency, ultimately generating an advertising effect tracking report.

[0058] The third step is to perform attribution analysis and visualization of the advertising performance tracking report to generate an intelligent delivery report.

[0059] It should be noted that, firstly, ad impressions, conversion rates, and user engagement metrics are aligned along a spatiotemporal dimension to establish a mapping relationship between ad performance and placement strategies. Secondly, through comparative experiments, under controlled conditions, the difference in performance between a single ad placement when it is closed and open is measured to obtain the net effect value of the ad placement when it is closed or open. Simultaneously, quantum fidelity data is used to calibrate the measurement results to eliminate noise interference during quantum state transmission. Finally, the contribution weight of each ad placement is determined, with weight values ​​ranging from 0 to 1, representing the relative influence of the ad placement on the overall performance.

[0060] It should be noted that other variables refer to influencing factors that need to be kept constant during the comparative experiment, including but not limited to: environmental data (temperature and humidity, light intensity), user behavior data (staying heatmap, device signal strength), commercial activity data (promotional labels, discount level), competitive advertising status (competitor display frequency, brightness value), and device operating parameters (network latency, quantum error rate).

[0061] In summary, this invention achieves a paradigm shift in advertising decision-making from correlation analysis to causal intervention by constructing a dynamic causal graph and a quantum collaborative architecture. This enables the system to accurately quantify the true effects of advertising actions and provide millisecond-level dynamic responses. Furthermore, through global quantum entanglement state modeling and conflict resolution optimization, it realizes long-distance collaboration of distributed nodes and optimal resource allocation, thus constructing a highly robust advertising delivery network.

[0062] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An intelligent offline advertising delivery system based on big data and AI, characterized in that: include, The data acquisition module collects environmental data, user behavior data, business activity data, competitive advertising status, equipment operating parameters, and sales data in real time, performs preprocessing, and outputs a panoramic situational data stream. The causal decision-making module, based on panoramic situational data streams, constructs dynamic causal graphs and performs intervention effect analysis to generate quantum decision vectors; The quantum encoding module performs quantum state probability amplitude mapping and binding with entangled particles on the quantum decision vector to generate a quantum decision proposal package; The collaborative optimization module constructs a global quantum entangled state model, optimizes the quantum decision proposal package by resolving conflicts, and generates quantum collaborative execution instructions. The report generation module performs quantum state collapse operations on quantum cooperative execution instructions to generate intelligent delivery reports.

2. The offline advertising intelligent delivery system based on big data and AI as described in claim 1, characterized in that: The environmental data includes temperature and humidity values, light intensity and direction, weather warning index, air quality index, and noise decibel value; The user behavior data includes gaze focus coordinates, gait speed, dwell time heatmap, device signal strength distribution, and emotion recognition index; The business activity data includes promotional semantic tags, discount intensity values, brand competition index, inventory turnover rate, and member consumption preference tags; The competitive advertising status includes competitor advertising content, display time percentage, screen brightness value, interaction response rate, and bid consumption rate; The device operating parameters include advertising screen power consumption, network latency, quantum channel bit error rate, fault diagnosis code, and heat sink temperature. The sales data includes real-time transaction volume, product conversion correlation, inventory changes, average order value fluctuations, and return rate.

3. The offline advertising intelligent delivery system based on big data and AI as described in claim 2, characterized in that: The preprocessing includes data cleaning, spatiotemporal feature extraction, and Z-score normalization.

4. The offline advertising intelligent delivery system based on big data and AI as described in claim 3, characterized in that: The specific steps for constructing a dynamic causal graph based on panoramic situational data stream are as follows. Perform association rule mining and entity classification on the panoramic situational data stream to generate a set of causal nodes and a set of candidate causal pairs. Perform FCI conditional independence test and do operator strength calculation on the causal node set and candidate causal pair set to generate a dynamic causal graph with edge weight values.

5. The offline advertising intelligent delivery system based on big data and AI as described in claim 4, characterized in that: The specific steps for generating the quantum decision vector are as follows: Perform causal effect quantification on the dynamic causal graph to generate candidate decision triples; Candidate decision triples are structured and quantum-encoded and bound to causal paths to generate quantum decision vectors.

6. The offline advertising intelligent delivery system based on big data and AI as described in claim 5, characterized in that: The specific steps for performing quantum state probability amplitude mapping on the quantum decision vector are as follows. Perform causal-probability transformation and normalization verification on the quantum decision vector to generate a probability amplitude allocation table; The probability amplitude allocation table is dynamically labeled with phase angle and encapsulated with quantum state parameters to generate a quantum state parameter table.

7. The offline advertising intelligent delivery system based on big data and AI as described in claim 6, characterized in that: The specific steps for generating the quantum decision proposal package are as follows: Quantum entangled particle pairs are assigned to the quantum state parameter table to generate a quantum entangled registry; Quantum-safe device binding is performed on the quantum entanglement registry to generate a quantum decision proposal package.

8. The offline advertising intelligent delivery system based on big data and AI as described in claim 7, characterized in that: The specific steps for constructing a global quantum entangled state model, optimizing the quantum decision proposal package through conflict resolution, and generating quantum cooperative execution instructions are as follows. A global quantum entangled state model is constructed based on the edge node layer, quantum channel layer, and cloud collaboration layer. The edge node layer adapts quantum state parameters to the quantum decision proposal package to generate a set of device-executable quantum instructions. Quantum state synchronization transmission and entanglement verification are performed on the quantum instruction set executable by the device to generate quantum cooperative control signals; The cloud-based collaboration layer performs global strategy optimization and quantum signature encapsulation on the quantum collaborative control signal to generate quantum collaborative execution instructions.

9. The offline advertising intelligent delivery system based on big data and AI as described in claim 8, characterized in that: The specific steps for performing quantum state collapse operations on quantum cooperative execution instructions are as follows: Quantum state collapse measurement and execution state verification are performed on quantum cooperative execution instructions, and a quantum state execution feedback report is generated. The quantum state execution feedback report is subjected to full data fusion analysis and execution effect quantification to generate an advertising effect tracking report.

10. The offline advertising intelligent delivery system based on big data and AI as described in claim 9, characterized in that: The aforementioned generation of quantum collaborative execution instructions refers to attribution analysis and visualization encapsulation of advertising performance tracking reports to generate intelligent delivery reports.

Images