Market behavior deduction method and device based on digital twinning and medium

Abstract

The application discloses a market behavior deduction method and device based on digital twinning and a medium, relates to the technical field of digital twinning simulation, and comprises the following steps: collecting market behavior data, performing anonymous association and time sequence calibration on the market behavior data, and concatenating continuous behaviors before and after market stimulation to generate a market behavior time table; continuously comparing historical choices and recent responses of the same consumer object in the market behavior time table, determining a stable behavior boundary and a current market state, limiting a receivable information range according to an actual exposure path, and generating a consumer twinning visible state graph; based on the consumer twinning visible state graph, connecting a consumer digital twinning body and a market environment according to the information receiving range, and respectively loading different market intervention conditions with the same initial state to generate a market twinning branch scene. The application improves the local correction accuracy and branch continuation reliability of digital twinning market behavior deduction.

Description

Technical Field

[0001] This invention relates to the field of digital twin simulation technology, and in particular to a method, device and medium for market behavior extrapolation based on digital twins. Background Technology

[0002] With the development of digital consumer touchpoints and market feedback collection capabilities, related methods are gradually adopting digital twins to construct virtual mappings of consumer objects. This transforms historical choices, recent responses, information exposure, demand changes, and communication behaviors into calculable states. Market intervention conditions such as price adjustments, content delivery, and channel changes are then loaded under unified initial conditions. Through multi-branch simulation, the evolution of consumer decisions and the information diffusion process are observed. The simulation parameters are then calibrated based on real market feedback to support demand forecasting, intervention assessment, and market strategy deduction.

[0003] When existing digital twin market projection methods encounter discrepancies between simulated states and real market feedback, they typically readjust consumer behavior parameters based on the overall error. However, in scenarios where intervention conditions change frequently and consumer information propagates continuously along social connections, channel touchpoints, and group influence paths, this approach is prone to the problem of mixing the starting point of the deviation with subsequent affected states. This is because the existing process does not track the direction of deviation propagation along the state progression sequence and propagation connections, making it difficult to determine the most recent credible verification location and the actual affected range. Consequently, consumer states unaffected by the deviation also participate in the rollback correction, reducing the continuity of states in subsequent branch projections and the credibility of the projection. Summary of the Invention

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

[0005] Therefore, this invention provides a market behavior extrapolation method based on digital twins, which solves the problems of inaccurate positioning of deviation range and difficulty in local continuation correction.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a market behavior extrapolation method based on digital twins, comprising: collecting market behavior data; anonymizing and temporally calibrating the market behavior data and connecting continuous behaviors before and after market stimuli to generate a market behavior time table; continuously comparing the historical choices and recent responses of the same consumer in the market behavior time table to determine the stable behavior boundary and the current market state, and limiting the range of receivable information according to the actual exposure path to generate a visible state map of the consumer twin; based on the visible state map of the consumer twin, connecting the consumer digital twin with the market environment according to the information receiving range, and loading different market intervention conditions with the same initial state. The system generates market twin branch scenarios. Based on these scenarios, each intervention branch is arranged into a unified computable simulation structure. The execution state is advanced and propagated according to the same initial state and simulation step size. The state of each step is verified against the real market feedback. The mismatch location and affected range are determined along the deviation propagation direction, generating a deviation causal cone. Based on this cone, the state and propagation connection of the consumer digital twin affected by the deviation are reverted to the most recently verified position. The behavioral boundaries and information reception conditions are corrected based on the real market feedback. Each market intervention branch is then re-enacted. The demand response and propagation acceptance status after the re-enactment are compared to generate a market behavior projection scheme.

[0007] As a preferred embodiment of the market behavior extrapolation method based on digital twins described in this invention, the specific steps of collecting market behavior data, anonymizing and temporally calibrating the market behavior data, and connecting the continuous behaviors before and after market stimuli to generate a market behavior time table are as follows: The consumer object identifiers in the market behavior data are anonymously replaced. Market behaviors belonging to the same consumer object are associated according to the content and source of the behavior. The recording time of each market behavior is calibrated by the time the market stimulus occurs, generating an anonymously calibrated behavior sequence. Based on anonymous calibration behavior sequences, historical behaviors before market stimulus are continuously linked with response behaviors after market stimulus, and duplicate behaviors and time-series conflict behaviors are eliminated to generate a market behavior time table.

[0008] As a preferred embodiment of the market behavior extrapolation method based on digital twins described in this invention, the steps of continuously comparing the historical choices and recent responses of the same consumer in the market behavior time table to determine the stable behavior boundary and the current market state, and limiting the range of receivable information according to the actual exposure path to generate a visible state map of the consumer twin, are as follows: Based on the market behavior time table, according to the anonymous consumer and the type of market stimulus, the historical choices before the stimulus and the recent responses after the stimulus are matched and corresponding according to the same consumption opportunities, and then continuously linked according to the time of the behavior to generate a continuous behavior comparison chain. Based on the continuous behavior comparison chain, the behavior stability verification elements of each pairing position are continuously verified. Adjacent pairing positions that maintain stable acceptance are merged into stable behavior segments. Based on the stable behavior segments, the range of sustainable behavior of the consumer is limited, and a stable behavior boundary table is generated. Based on the stable behavior boundary table, the position of the consumer's most recent selection state and recent response state is checked against the corresponding stable behavior range. The consumer behavior is determined to be either a stable continuation state or a boundary shift state. The corresponding state mark is written into each behavior position to generate the current market state map. Based on the current market state map, the information exposure status process of market information is tracked according to the actual exposure order in the market behavior time table. Market information that has arrived and been effectively presented before the recent response occurs is retained, market information with invalid exposure status is removed, and an actual exposure reachable sequence is generated. Based on the actual exposure reach sequence, the behavior range in the stable behavior boundary table, the consumption status in the current market status graph, and the actual reachable market information are respectively set as behavior boundary nodes, consumption status nodes, and visible information nodes, and a consumption twin visible status graph is established between the nodes according to the actual exposure order.

[0009] As a preferred embodiment of the market behavior extrapolation method based on digital twins described in this invention, the consumer digital twin is connected to the market environment according to the information reception range based on the visible state diagram of the consumer twin, and different market intervention conditions are loaded with the same initial state. The specific steps are as follows: Based on the visible state diagram of the consumer twin, the effective information nodes in the market environment are connected to the corresponding consumer state nodes according to the actual exposure order and information reception range of the visible information nodes, and the information transmission is cut off at the invalid exposure position to generate a twin environment reception connection table. Select the most recent common verification position before the start of different market interventions from the twin environment receiving connection table, and encapsulate the stable behavior boundary, current market state, last effective exposure position, incomplete response state and effective receiving connection in a unified manner to generate a unified initial state packet; Based on the unified initial state package, each market intervention condition is verified accordingly, and the full loading and boundary loading are determined according to the information receiving scope to generate an intervention condition loading table; Based on the intervention condition loading table, a unified initial state package is copied according to different market intervention conditions. The stable behavior boundaries, current market state and original receiving connection of each branch are kept consistent. The corresponding intervention content and loading state are written to generate the market twin branch initial state table.

[0010] As a preferred embodiment of the market behavior extrapolation method based on digital twins described in this invention, the market twin branch scenario is obtained by connecting each branch to the same market environment benchmark and unifying the initial benchmark of the branches according to the initial state table of the market twin branches, and then solidifying different market intervention conditions to the corresponding branches.

[0011] As a preferred embodiment of the market behavior extrapolation method based on digital twins described in this invention, the step of arranging each intervention branch into a unified computable simulation structure according to the market twin branch scenario, executing the state progression and propagation calculation according to the same initial state and simulation step size, and verifying the state of each step calculation with the actual market feedback, and determining the mismatch location and affected range along the deviation propagation direction, is as follows: Based on the market twin branch scenario, the simulation orchestration information of each branch is uniformly configured according to the market intervention branch to obtain a unified simulation orchestration table. Based on the unified simulation orchestration table, the initial benchmark of the branches is used as a common starting point, and the market information simulation evolution state is advanced sequentially according to the same simulation step size. The state propagation trajectory of each state is recorded to generate a step-by-step simulation trajectory. Based on the step-by-step simulation trajectory, the consumption status and propagation changes in each simulation step are compared with the real market feedback in the corresponding time period to determine the deviation status of each simulation position and generate a corresponding deviation verification table. Based on the same position deviation verification table, the consumer state position where the first continuous deviation is formed is located according to the simulation step sequence. The visible information position, propagation connection, market intervention branch and simulation step position corresponding to the first continuous deviation position are checked back to generate the first mismatch positioning table. Based on the first mismatch location table, the deviation extension state is tracked step by step along the state transmission direction in the step-by-step simulation trajectory from the first mismatch position. The deviation state in adjacent simulation steps is continuously transmitted along the same propagation connection direction, and the tracking is terminated at the point where the propagation direction is reversed, thus generating a deviation propagation chain.

[0012] As a preferred embodiment of the market behavior extrapolation method based on digital twins described in this invention, the deviation causal cone is obtained by determining the first mismatch position as the deviation starting point based on the deviation propagation chain, and collecting the affected consumption state positions, simulation step ranges, visible information positions, and market intervention branches according to the propagation sequence.

[0013] As a preferred embodiment of the market behavior extrapolation method based on digital twins described in this invention, the step of using the deviation causal cone to revert the state and propagation connection of the consumer digital twin affected by the deviation to the most recently verified position, correcting the behavior boundaries and information reception conditions based on real market feedback, and continuing to extrapolate each market intervention branch, comparing the demand response and propagation acceptance status after the extrapolation, and generating a market behavior extrapolation scheme, the specific steps are as follows: Based on the deviation causal cone, the positions of each affected consumption state are checked back along the deviation propagation direction. The simulation position that is most recently consistent with the real market feedback before the first mismatch position is determined as the back-off position. The consumption state positions and propagation connections that need to be backed up are delineated, and a local back-off boundary table is generated. Based on the local rollback boundary table, the affected consumption state and propagation connection are restored to the corresponding rollback position, the state changes formed after the rollback position are undone, and the state content that has been verified outside the deviation causal cone is retained to generate a local rollback state package. Based on the local rollback state package, the stable behavior boundary corresponding to the actual selection state and demand response state in the real market feedback and the rollback position is verified in tandem, and the stable behavior range is corrected based on the continuous real response to generate a behavior boundary correction table. Based on the behavior boundary correction table, check the information exposure status of each market information, restore the market information that has completed effective exposure to the receptive connection, and set the propagation connection corresponding to the invalid exposure status to the blocked status, and generate a feedback correction status table. Based on the feedback correction state table, the corrected behavior boundary and information receiving connection are written into the local backoff state packet and merged with the consumption state and propagation connection retained outside the deviation causal cone to generate the branch continuation start table. Based on the branch continuation start table, each market intervention branch is continued from the common retreat position, and the demand response status, propagation acceptance status and remaining deviation range of each branch after continuation are recorded to generate the branch continuation trajectory. Based on the branch evolution trajectory, the demand response status and dissemination acceptance status of each market intervention branch are compared with the corresponding real market feedback to determine the applicable order of each market intervention branch and generate a market behavior projection plan.

[0014] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the market behavior extrapolation method based on digital twins as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the market behavior extrapolation method based on digital twins as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: by using a deviation causal cone to revert the state and propagation connection of the consumer digital twin affected by the deviation to the most recently verified position, and correcting the behavioral boundaries and information reception conditions based on real market feedback, the directional recovery of the state within the deviation range and the continuous evolution of each market intervention branch are completed, so that the unaffected state maintains its original evolution trajectory. At the same time, the demand response and propagation acceptance state after the continuation are consistently verified, providing a reliable basis for the evaluation of the market intervention effect, and improving the accuracy of local correction and the reliability of branch continuation in the digital twin market behavior projection. 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 flowchart of a market behavior extrapolation method based on digital twins.

[0019] Figure 2 A flowchart for constructing the visible state diagram of the consumer twin and the branch scenarios of the market twin.

[0020] Figure 3 This is a flowchart for identifying biased causal cones.

[0021] Figure 4 This is a flowchart for local rollback and branch continuation.

[0022] Figure 5 This is a comparison chart of local correction accuracy under different deviation propagation depths.

[0023] Figure 6 A comparison chart of the relative error of demand response under different simulation steps. Detailed Implementation

[0024] 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.

[0025] 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.

[0026] 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.

[0027] Reference Figures 1-6 As one embodiment of the present invention, this embodiment provides a market behavior extrapolation method based on digital twins, comprising the following steps: S1. Collect market behavior data, perform anonymous correlation and time-series calibration on the market behavior data, and connect the continuous behaviors before and after market stimulus to generate a market behavior time table.

[0028] It should be noted that market behavior data includes consumer object identification, behavior content, source of occurrence, time of behavior recording, type of market stimulus, time of occurrence of market stimulus, market information exposure status, consumption choice status, demand response status, and behavioral interaction feedback, which are used to reconstruct the continuous behavior process of consumers before and after market stimulus.

[0029] S1.1. Anonymize and replace the consumer object identifiers in the market behavior data, associate market behaviors belonging to the same consumer object according to the content and source of the behavior, and calibrate the recording time of each market behavior with the time when the market stimulus occurs to generate an anonymous calibration behavior sequence.

[0030] It should be noted that the consistency of each consumer object identifier in the market behavior data is checked, a unique anonymous identifier is assigned to the same consumer object identifier, and an anonymous replacement correspondence table is established. Based on the anonymous replacement correspondence table, the consumer object identifier in the market behavior data is replaced with the anonymous identifier to form an anonymous market behavior set.

[0031] Anonymous market behavior sets are aggregated according to anonymity identifiers. Market behaviors whose content is interconnected and whose sources can be continuously matched are linked to the same behavior sequence to form a consumer behavior association sequence. Using the market stimulus occurrence time corresponding to each consumer behavior association sequence as a time benchmark, the chronological position of the recording time of each market behavior relative to the market stimulus occurrence time is calculated. The recording time and arrangement order are then corrected according to the calculated chronological position. Finally, the consumer behavior association sequences that have completed time calibration are merged sequentially according to anonymity identifiers to generate anonymized calibrated behavior sequences.

[0032] The expression for calculating the chronological order of each market action's recording time relative to the time of the market stimulus is as follows: ; in, For the first The calibrated position of a market behavior relative to the moment when a market stimulus occurs; For the first The order of market actions after calibration is arranged in ascending order; For the first A record of a market action; For the moment when market stimulus occurs; For median operations; For the first Each market action corresponds to a set of corresponding market actions from its source. For the corresponding source record A moment of identical market behavior; For the first The benchmark moment for market behavior at the same level; To determine the sorting order according to the numerical values ​​from smallest to largest; The first in the consumer object behavior association sequence The serial number of each market action; For the corresponding source used to calculate the clock offset The serial number of each corresponding market behavior; To calibrate the order of market actions Arranged from smallest to largest, the first The sequential number corresponding to each market behavior.

[0033] First, using market behaviors from the same source that correspond to the time benchmark, calculate the time difference between the recording time and the benchmark time for each behavior, and take the median of all time differences as the uniform time correction value for the corresponding source. Then, subtract the market stimulus occurrence time and the uniform time correction value from the recording time of the target market behavior to obtain the calibration order of the target market behavior relative to the market stimulus occurrence time. Finally, arrange the target market behaviors in ascending order of calibration order to determine their ranking in the consumer behavior association sequence.

[0034] S1.2 Based on the anonymous calibration behavior sequence, the historical behavior before the market stimulus and the response behavior after the market stimulus are continuously connected, and repeated behaviors and time-series conflict behaviors are eliminated to generate a market behavior time table.

[0035] It should be noted that, based on the anonymous calibration behavior sequence, market behaviors before and after the occurrence of the corresponding market stimulus are collected according to the anonymity identifier. The market behaviors before the occurrence of the market stimulus are arranged into a historical behavior sequence according to the calibrated record time, and the market behaviors after the occurrence of the market stimulus are arranged into a response behavior sequence according to the calibrated record time.

[0036] Based on the content and source of the behavior, the connection between historical behavior sequences and response behavior sequences is verified. End-stage historical behaviors reflecting the same consumption process are continuously connected with the initial response behaviors to form a sequence of behaviors preceding and following the stimulus. Market behaviors in the sequence of behaviors preceding and following the stimulus that share the same anonymous identifier, behavior content, source of occurrence, and recording time are compared item by item. The first occurrence of a market behavior is retained, while duplicate market behaviors are deleted, forming a deduplicated behavior sequence. The order of behaviors in the deduplicated behavior sequence is checked according to the calibrated recording time. Market behaviors with reversed recording times, mutually exclusive behaviors at the same time, or inconsistent behavior sequence are identified as time-conflicting behaviors. Based on the market stimulus occurrence time and the order of preceding and following behaviors, continuously corresponding market behaviors are retained, while time-conflicting behaviors that cannot be continuously connected are eliminated. Finally, the behaviors are merged and arranged according to the anonymous identifier and the calibrated recording time to generate a market behavior time table.

[0037] S2. Continuously compare the historical choices and recent responses of the same consumer in the market behavior time table to determine the stable behavior boundary and the current market state, and limit the range of receivable information according to the actual exposure path to generate a consumer twin visible state map.

[0038] S2.1. Based on the market behavior time table, according to the anonymous consumer and the type of market stimulus, the historical choices before the stimulus and the recent responses after the stimulus are matched and corresponding according to the same consumption opportunities, and then continuously linked according to the time of the behavior to generate a continuous behavior comparison chain.

[0039] It should be noted that, based on the market behavior timetable, the corresponding market behaviors are collected according to the anonymous consumer objects, and then the historical choices before the stimulus and the recent responses after the stimulus are divided according to the market stimulus type, forming a sequence of behaviors to be corresponded.

[0040] The location of the consumption object, consumption content, source of occurrence, type of market stimulus, and time period of the behavior sequence are checked item by item. Historical choices before the stimulus and recent responses after the stimulus that are consistent in consumption content, have a continuous source of occurrence, and are consecutive before and after the behavior occurrence time are identified as the same consumption opportunity. Corresponding positions are established according to the same consumption opportunity to form a consumption opportunity correspondence sequence. The historical choices before the stimulus and recent responses after the stimulus in the consumption opportunity correspondence sequence are sorted according to the time of occurrence of the behavior. Each historical choice before the stimulus is connected to the corresponding recent response after the stimulus. Then, adjacent consumption opportunity correspondence sequences are connected in sequence according to the anonymous consumption object and the type of market stimulus to generate a continuous behavior comparison chain.

[0041] S2.2 Based on the continuous behavior comparison chain, the behavior stability verification elements of each pairing position are continuously verified. Adjacent pairing positions that maintain stable acceptance are merged into stable behavior segments. Based on the stable behavior segments, the range of sustainable behavior of the consumer object is limited, and a stable behavior boundary table is generated.

[0042] It should be noted that, based on the continuous behavior comparison chain, the historical selection status, recent response direction, selection duration, response change magnitude, and adjacent behavior interval of each pairing position are read sequentially according to the anonymous consumer object and market stimulus type, and the read content is collected into behavior stability verification elements; the behavior stability verification elements of adjacent pairing positions are compared item by item, and the pairing positions that have consistent historical selection, continuous recent response direction, response change have not exceeded the previous pairing position, and adjacent behavior intervals remain continuous are marked as stable acceptance positions, forming a stable acceptance judgment sequence.

[0043] Based on the stable acceptance determination sequence, consecutive adjacent stable acceptance positions are merged sequentially from the first stable position to the last stable position to determine the starting position, ending position, and acceptance direction of each stable behavior segment. Then, based on the maximum continuous change range of historical selection status and recent response status within the stable behavior segment, the behavior range that the consumer can sustain without interruption is determined. Stable behavior segments and corresponding behavior ranges are then aggregated according to anonymous consumer and market stimulus type to generate a stable behavior boundary table.

[0044] S2.3. Based on the stable behavior boundary table, the position of the consumer's most recent selection state and recent response state is checked against the corresponding stable behavior range. The consumer behavior is determined to be either a stable continuation state or a boundary wandering state. The corresponding state mark is written into each behavior position to generate the current market state diagram.

[0045] It should be noted that, based on the stable behavior boundary table, the corresponding stable behavior range is located according to the anonymous consumer and the market stimulus type, and the consumer's most recent selection status and recent response status are read from the market behavior time table to form the current status check sequence.

[0046] The most recent selection state and the recent response state in the current state check sequence are respectively placed into the corresponding behavior positions within the stable behavior range. The continuity direction of the most recent selection state and the stable behavior segment is checked to see if they are consistent. The movement position of the recent response state relative to the edge of the stable behavior range is also checked to form a behavior position check sequence. Based on the behavior position check sequence, behavior positions where both the most recent selection state and the recent response state are within the stable behavior range and the continuity direction is consistent are determined to be stable continuation. Behavior positions where the recent response state is close to the edge of the stable behavior range or changes back and forth at a position adjacent to the edge of the stable behavior range are determined to be boundary wandering. Then, according to the time of the behavior, stable continuation and boundary wandering are written into the corresponding behavior positions, and the state continuity order of adjacent behavior positions is connected to generate the current market state diagram.

[0047] It should also be noted that the current market state diagram includes anonymous consumer objects, market stimulus types, most recent choice state, recent response state, corresponding stable behavior range, behavior location, direction of transition, and markers for stable continuation state or boundary shifting state. It is used to map the current behavior of consumer objects to the corresponding position in the stable behavior boundary and connect the state transition order of adjacent behavior positions, thereby identifying whether consumer behavior is continuous and stable, whether it is close to the boundary, and whether it has been shifting back and forth. This provides a basis for subsequently limiting the scope of market information that can be received and constructing a visible consumer twin state diagram, improving the continuity and accuracy of current market state determination.

[0048] S2.4 Based on the current market state map, track the information exposure status process of market information according to the actual exposure order in the market behavior time table, retain market information that has arrived and been effectively presented before the recent response occurs, remove market information with invalid exposure status, and generate the actual exposure reach sequence.

[0049] It should be noted that, based on the current market state map, the recent response time is located according to the anonymous consumer object and behavior location, and the arrival time, presentation start time, presentation end time and information exposure status of each market information before the recent response time are read sequentially from the market behavior time table, and arranged according to the actual exposure order to form an exposure status tracking sequence.

[0050] Based on the exposure status tracking sequence, each market information item is checked to see if it arrived before the recent response occurred, and whether the market information was presented continuously. Market information whose arrival time is earlier than the recent response time and whose information exposure status indicates that it has been effectively presented is marked as reachable market information, forming an exposure reachability verification sequence. Based on the exposure reachability verification sequence, all reachable market information is retained, and market information that arrived after the recent response occurred, whose presentation process was interrupted, whose effective presentation status was not achieved, or whose information exposure status is missing is removed. Then, the retained market information is connected according to the actual exposure order to generate the actual exposure reachability sequence.

[0051] S2.5. Based on the actual exposure reach sequence, set the behavior range in the stable behavior boundary table, the consumption status in the current market status diagram, and the actual reachable market information as behavior boundary nodes, consumption status nodes, and visible information nodes, respectively, and establish a consumption twin visible status diagram between nodes according to the actual exposure order.

[0052] It should be noted that, based on the actual exposure reach sequence, the behavior range in the stable behavior boundary table, the consumption status in the current market status diagram, and the actual reach market information in the actual exposure reach sequence are read according to the anonymous consumer object and the market stimulus type. The corresponding positions of the three types of content are established according to the location of the behavior and the time of the recent response, forming a state content correspondence sequence.

[0053] Based on the state content correspondence sequence, each behavior range is set as a behavior boundary node, the consumption state falling into the corresponding behavior range is set as a consumption state node, and the actual reachable market information that was effectively presented before the recent response occurred is set as a visible information node. The corresponding anonymous consumer object, market stimulus type, and behavior location are written into the behavior boundary node, consumption state node, and visible information node to form a node correspondence sequence. According to the node correspondence sequence, the consumption state node is connected to the corresponding behavior boundary node, and then each visible information node is sequentially connected to the corresponding consumption state node according to the actual exposure order. Connections are disconnected between different anonymous consumer objects and different market stimulus types to form a node connection sequence. The node connection sequence is arranged according to the order of behavior occurrence. After verifying that each consumption state node has a corresponding behavior boundary node and a visible information node, a consumer twin visible state graph is generated.

[0054] It should also be noted that the visible state diagram of consumer twins is used to establish a correspondence between stable behavior boundaries, current consumption status, and actual available market information according to behavioral location and exposure order. This clarifies the range of information that different consumer objects can receive under specific market stimuli and the current position of their behavior. By excluding information that has not been effectively exposed, it can prevent invisible market information from participating in subsequent inferences, ensuring that market intervention branches are based on real available information. This improves the authenticity of the consumption status description, the accuracy of market response inferences, and the credibility of comparisons between different intervention conditions.

[0055] S3. Based on the visible state diagram of the consumer twin, connect the consumer digital twin with the market environment according to the information receiving range, and load different market intervention conditions with the same initial state to generate market twin branch scenarios.

[0056] S3.1 Based on the consumer twin visible state diagram, according to the actual exposure order and information reception range of the visible information nodes, connect the effective information nodes in the market environment to the corresponding consumer state nodes, and cut off the information transmission at the invalid exposure position to generate a twin environment reception connection table.

[0057] It should be noted that, based on the visible state diagram of the consumer twin, visible information nodes and corresponding consumer state nodes are located according to anonymous consumer objects and market stimulus types. The information reception range determined and written by the market behavior time table (origin, exposure location, presentation interval, and effective exposure status) of the visible information nodes is read, and the actual exposure order is read to form an information reception verification sequence. According to the information reception verification sequence, each information node in the market environment is checked against the visible information nodes. Information nodes with consistent information content, exposure locations falling within the information reception range, and effective exposure are identified as effective information nodes. Effective information nodes are then connected sequentially to their corresponding consumer state nodes according to the actual exposure order, and the starting position, arrival position, and receiving order of information transmission are recorded to form an information transmission path. The information transmission path is verified level by level. Locations where effective exposure is not completed, exposure locations exceed the information reception range, or the presentation process is interrupted are identified as invalid exposure locations. Information transmission is terminated at invalid exposure locations, and subsequent receiving connections are canceled. Finally, the effective information nodes, corresponding consumer state nodes, information transmission paths, and invalid exposure locations are collected to generate a twin environment reception connection table.

[0058] S3.2 Select the most recent common verification position before the start of different market interventions from the twin environment receiving connection table, and encapsulate the stable behavior boundary, current market state, last effective exposure position, incomplete response state and effective receiving connection in a unified manner to generate a unified initial state packet.

[0059] It should be noted that, from the twin environment receiving connection table, valid receiving connections are checked back according to the start time of different market interventions. The consistency of the state and the integrity of the connection of each consumer state node before the start of the market intervention are checked in turn. The position closest to the start time of each market intervention and which has been verified is determined as the closest common verification position before the start of different market interventions.

[0060] Based on the most recent common verification location before the start of different market interventions, the corresponding stable behavior boundary and current market state are located. The market information transmission sequence is traced back along the effective receiving connection, and the exposure location that most recently completed effective reception is determined as the last effective exposure location. The response process that has not yet formed a complete consumer feedback after the last effective exposure location is then checked to determine the incomplete response state. According to the same consumer state node and the same verification time, the stable behavior boundary, current market state, last effective exposure location, incomplete response state, and effective receiving connection are written into the same encapsulated content item by item, and the temporal and positional continuity between the encapsulated content is verified to be consistent, generating a unified initial state package.

[0061] S3.3. Based on the unified initial state package, verify each market intervention condition accordingly, determine the complete loading and boundary loading according to the information receiving scope, and generate the intervention condition loading table.

[0062] It should be noted that, based on the unified initial state package, the corresponding intervention content, intervention start position, intervention duration range, and information delivery position are read according to different market intervention conditions. Each market intervention condition is then matched with the stable behavior boundary, current market state, last effective exposure position, incomplete response status, and effective receiving connection in the unified initial state package to form an intervention condition verification sequence.

[0063] Based on the intervention condition verification sequence, the information delivery location of each market intervention condition is verified along the effective receiving connection to ensure it is within the information receiving range of the corresponding consumer status node. The intervention start location is also verified to be later than the last effective exposure location and not to overlap with the incomplete response status in time, thus forming an intervention receiving determination sequence. According to this sequence, market intervention conditions where all intervention content is within the information receiving range and can continuously reach the corresponding consumer status node along the effective receiving connection are defined as complete loading. Market intervention conditions where only part of the intervention content is within the information receiving range or where the reception is interrupted at the edge of the information receiving range are defined as boundary loading. The receptive portion and truncation position corresponding to the boundary loading are marked. Then, complete loading, boundary loading, receptive portion, truncation position, and corresponding consumer status node are collected item by item according to the market intervention conditions to generate an intervention condition loading table.

[0064] It should also be noted that different market intervention conditions are used to define the intervention content, starting position, duration, information delivery position, and loading status of each market intervention branch, so that each branch retains only one differentiated intervention factor under the same initial state and the same market environment benchmark. By independently loading, advancing, and verifying the feedback of different intervention conditions, the influence of differences in the initial state on the extrapolation process can be eliminated, the impact of various market interventions on consumer choices, demand response, and communication reception status can be accurately compared, and the comparability and credibility of market intervention effect evaluation can be improved.

[0065] S3.4 Based on the intervention condition loading table, copy the unified initial state package according to different market intervention conditions, keep the stable behavior boundary, current market state and original receiving connection of each branch consistent, write the corresponding intervention content and loading state, and generate the market twin branch initial state table.

[0066] It should be noted that, based on the intervention condition loading table, branch identifiers are established according to different market intervention conditions, and the unified initial state package is copied one by one using the branch identifiers as indexes to form the initial state copy corresponding to each market intervention condition.

[0067] Each initial state copy is checked against its stable behavior boundaries, current market state, and existing receiving connections to ensure consistency in the range of stable behavior boundaries, the position of the current market state, and the order of the existing receiving connections, thus forming a branch-consistent initial state. The intervention content and loading status corresponding to each market intervention condition are read from the intervention condition loading table. All intervention content corresponding to a complete loading is written into the branch-consistent initial state, and the receivable intervention content corresponding to boundary loading is written into the branch-consistent initial state, with unloaded content marked at truncation positions, forming an intervention writing initial state. Then, the stable behavior boundaries, current market state, existing receiving connections, intervention content, and loading status in the intervention writing initial state are aggregated according to the branch identifier, and it is verified that each branch contains only one market intervention condition and that the initial content of each branch remains consistent, generating a market twin branch initial state table.

[0068] S3.5. Based on the initial state table of market twin branches, connect each branch to the same market environment benchmark and unify the initial benchmark of the branches. Then, solidify different market intervention conditions to the corresponding branches to generate market twin branch scenarios.

[0069] It should be noted that, based on the initial state table of the market twin branches, the initial state content of each branch is connected to the same market environment benchmark according to the branch identifier, forming the branch environment connection content.

[0070] Based on the branch environment connection content, the branch start positions of each branch are checked for alignment, and the consistent branch start positions are determined as the branch initial benchmark, forming branch benchmark alignment content. According to the branch benchmark alignment content, all market intervention conditions corresponding to the complete loading are solidified to the corresponding branch, and the market intervention conditions corresponding to the boundary loading are solidified to the corresponding branch according to the acceptable content and truncation position, and the cross-transmission between different market intervention conditions is blocked, forming branch intervention solidified content. According to the branch identifier, the branch intervention solidified content is merged with the same market environment benchmark and the branch initial benchmark to generate market twin branch scenarios.

[0071] S4. Based on the market twin branch scenario, arrange each intervention branch into a unified computable simulation structure, execute the state advancement and propagation calculation according to the same initial state and simulation step size, and verify the state of each step calculation with the real market feedback, determine the mismatch position and affected range along the deviation propagation direction, and generate the deviation causal cone.

[0072] S4.1 Based on the market twin branch scenario, the simulation arrangement information of each branch is uniformly configured according to the market intervention branch, and a unified simulation arrangement table is obtained. Based on the unified simulation arrangement table, the initial benchmark of the branch is used as a common starting point, and the market information simulation evolution state is advanced sequentially according to the same simulation step size. The state propagation trajectory of each state is recorded, and a step-by-step simulation trajectory is generated.

[0073] It should be noted that, based on the market twin branch scenario, the simulation orchestration information is mapped to each branch according to the market intervention branch, and the execution order, simulation step position, market information delivery order, consumption status advancement position, and dissemination acceptance position of each branch are aligned and calibrated. The calibrated simulation orchestration information is merged according to the branch identifier to obtain a unified simulation orchestration table.

[0074] Based on a unified simulation orchestration table, the initial benchmark of each branch is set as the common starting point of all branches. The consumption state position, response direction, state duration, incomplete response state, and stable behavior boundary formed in the previous simulation step are used as the basis for the state advancement of the next simulation step. The market information that can be effectively received and connected to the corresponding consumption state position in this simulation step is determined according to the same simulation step size.

[0075] When there is no valid market information to be reached in this simulation step, the consumption state position and response direction of the previous simulation step are maintained, and the unfinished response state that has not yet ended continues; when the market information to be reached is consistent with the response direction of the previous simulation step, the consumption state is advanced to the adjacent behavior position along the original response direction. If it is still within the stable behavior boundary after advancement, it is marked as a stable continuation state. If it reaches the edge of the stable behavior boundary after advancement, it is marked as a boundary wandering state.

[0076] When the arriving market information is opposite to the response direction of the previous simulation step, first check whether there is an incomplete response state in the previous simulation step. If there is an incomplete response state, maintain the original response direction and extend the reverse market information to the simulation step after the incomplete response ends. If there is no incomplete response state, adjust the response direction to the influence direction corresponding to the market information and advance the consumption state along the adjusted response direction. When the consumption state exceeds the stable behavior boundary after advancement, keep the consumption state at the corresponding boundary position and write the excess part into the incomplete response state.

[0077] When multiple market information messages arrive at the same consumption state position within the same simulation step, they are processed sequentially according to their actual exposure time. For messages with the same actual exposure time, they are processed according to the order of valid receiving connections. Market information messages with the same direction of influence are merged consecutively and then a state progression is performed. Market information messages with opposite directions of influence are progressed first, and subsequent reverse market information is carried over to the next simulation step. After each simulation step's consumption state update, the pre-update consumption state position, market information arrival position, valid receiving connection, response direction, post-update consumption state position, stable continuation state, boundary wandering state, and incomplete response state are recorded according to the simulation step order, forming a continuous simulation state sequence. Based on this continuous simulation state sequence, the market information arrival position is used as the starting position of the state change, the updated consumption state position is used as the receiving position, and the propagation direction is determined according to the valid receiving connection. State changes that can be continuously received between adjacent simulation steps are connected to generate a step-by-step simulation trajectory.

[0078] S4.2 Based on the step-by-step simulation trajectory, the consumption status and propagation changes in each simulation step are compared with the real market feedback in the corresponding time period to determine the deviation status of each simulation position and generate a deviation verification table.

[0079] It should be noted that, based on the step-by-step simulation trajectory, real market feedback is selected from the market behavior time table according to the market intervention conditions. Consumers whose stable behavior boundary, current market state, last effective exposure position, and incomplete response state are consistent with the initial state of the corresponding branch, and whose actual market intervention conditions are consistent with the content loaded in the corresponding branch, are classified into the same real feedback group. The real feedback group identifier is then associated with the market intervention branch identifier to form a branch feedback correspondence table.

[0080] Based on the branch feedback correspondence table, according to the market intervention branch, simulation step sequence, and simulation position, the consumption status and dissemination changes in each simulation step are aligned with the actual choice status, demand response status, and information dissemination status formed by the corresponding real feedback group in the same period. When the same branch corresponds to multiple real feedback objects, the feedback content of each real feedback object is collected according to the behavior position, change direction, and continuous status. The feedback status that appears most frequently and is continuously received is taken as the real market feedback of the corresponding simulation position, forming a peer verification sequence.

[0081] Based on the same-position verification sequence, the behavioral position, direction of change, and duration of consumption status are first verified. Then, the receiving position, direction of propagation, and arrival order of propagation changes are verified. Simulation positions where all verified content is consistent are marked as unbiased positions, and simulation positions where any verified content is inconsistent are marked as deviation positions. Based on the difference type of consumption status and propagation changes in the deviation positions, the deviation direction, deviation start time, and continuous deviation status of each simulation position are determined. The deviation status of each simulation position is collected according to the market intervention branch and simulation step sequence to generate the same-position deviation verification table.

[0082] S4.3 Based on the same position deviation verification table, locate the consumption state position where the first continuous deviation is formed according to the simulation step sequence, and check the visible information position, propagation connection, market intervention branch and simulation step position corresponding to the first continuous deviation position to generate the first mismatch positioning table.

[0083] It should be noted that, based on the same-position deviation verification table, the deviation states of each simulation position are arranged sequentially according to the market intervention branch and simulation step order. The consumer state positions with consistent deviation directions and continuous deviation states in adjacent simulation steps are linked together to form a continuous deviation sequence. The unbiased position of the previous simulation step is checked in reverse along the continuous deviation sequence, and the first consumer state position that changes from unbiased to continuous deviation is determined as the first mismatch candidate position, forming the first deviation positioning sequence. According to the first deviation positioning sequence, the state source, propagation receiving position, and market information arrival position corresponding to the first mismatch candidate position are checked back along the step-by-step simulation trajectory, and the checked content that can explain the cause of the continuous deviation is collected into mismatch source tracing information. Then, the consumer state position that first forms continuous deviation and the mismatch source tracing information are merged according to the market intervention branch and simulation step order to generate the first mismatch positioning table.

[0084] S4.4 According to the first mismatch location table, track the deviation extension state step by step from the first mismatch position along the state transmission direction in the step-by-step simulation trajectory, continuously transmit the deviation state in adjacent simulation steps along the same propagation connection direction, and terminate the tracking at the point where the propagation direction reverses, thus generating a deviation propagation chain.

[0085] It should be noted that, based on the first mismatch location table, the first mismatch position is used as the starting point for deviation tracking. The consumer state positions in the step-by-step simulation trajectory are verified for propagation and acceptance according to the behavioral interaction feedback, market information forwarding records, and common channel touchpoints corresponding to anonymous consumer objects. A directed propagation connection is established between two positions where the previous consumer state position generates information forwarding or interaction influence, and the subsequent consumer state position responds after the corresponding information arrives. The directed propagation connection is then programmed with the propagation start position, propagation arrival position, effective simulation step range, propagation sequence, and acceptance status. Following the direction of the directed propagation connection, downstream consumer state positions in the next simulation step, located at the propagation arrival position and with an effective acceptance status, are located level by level from the first mismatch position. When the same consumer state position connects to multiple downstream consumer state positions, they are tracked sequentially according to the actual arrival time of the market information. For positions with the same actual arrival time, they are tracked sequentially according to the propagation sequence in the step-by-step simulation trajectory. The downstream consumer state positions are then connected in series according to the simulation step sequence to form a deviation tracking sequence.

[0086] Based on the deviation tracking sequence, the deviation direction of each downstream consumption state position is continuously checked against the deviation direction of the previous consumption state position. Downstream consumption state positions with consistent deviation directions, valid directed propagation connections, and no propagation interruption between adjacent simulation steps are included in the direction acceptance sequence. The deviation direction of subsequent consumption state positions that is opposite to the original propagation direction of the corresponding directed propagation connection is determined as a propagation direction reversal. Tracking stops at the position where the propagation direction reversal occurs for the first time, the propagation connection fails, or the deviation state is not accepted by the next simulation step. Then, the continuously accepted consumption state positions before the tracking stops are connected according to the propagation sequence to generate a deviation propagation chain.

[0087] S4.5 Based on the deviation propagation chain, the first mismatch position is determined as the deviation starting point, and the affected consumption state positions, simulation step ranges, visible information positions and market intervention branches are collected in the order of propagation to generate the deviation causal cone.

[0088] It should be noted that, based on the deviation propagation chain, the initial mismatch position is determined as the deviation starting point, and the state positions, simulation steps, visible information positions, and market intervention branches covered by the deviation propagation chain are uniformly recorded as deviation impact content according to the propagation sequence. The deviation impact content is then matched level by level along the deviation propagation chain, marking the entry position, acceptance position, and termination position of each level of deviation to form a deviation impact trajectory. Based on the deviation impact trajectory, the simulation steps within the continuous propagation range of the deviation are determined as the simulation step range, and the positions where deviations are no longer accepted are determined as deviation boundaries. Then, the deviation impact content at each level is sequentially connected from the deviation starting point to the deviation boundary, and the deviation propagation direction and affected range are merged according to the market intervention branches to generate a deviation causal cone.

[0089] It should also be noted that the deviation causal cone is used to centrally identify the affected consumption state location, simulation step range, visible information location, and market intervention branch along the deviation propagation direction, starting from the initial mismatch position. This clarifies the deviation starting point, propagation path, and termination boundary, narrows the scope of backtracking and correction, avoids repeated calculations on unaffected content, and improves the accuracy of deviation positioning and the efficiency of local continuation.

[0090] S5. Based on the deviation causal cone, the state and propagation connection of the consumer digital twin affected by the deviation are reverted to the most recently verified position. The behavior boundary and information reception conditions are corrected according to the real market feedback, and each market intervention branch is continued. The demand response and propagation acceptance status after the continued performance are compared to generate a market behavior projection plan.

[0091] S5.1 Based on the deviation causal cone, backtrack along the deviation propagation direction to check the positions of each affected consumption state. Determine the simulation position that is most recently consistent with the real market feedback before the first mismatch position as the backtrack position, and delineate the consumption state positions and propagation connections that need to be backtracked, generating a local backtrack boundary table.

[0092] It should be noted that, based on the deviation causal cone, the deviation impact content is checked step by step from the end to the first mismatch position according to the deviation propagation direction, and the content covered by the deviation causal cone is determined as the rollback object; the simulation positions before the first mismatch position are verified with the corresponding real market feedback along the step-by-step simulation trajectory, and the simulation positions where the consumption state and propagation state are consistent are marked as valid verification positions, forming a rollback candidate sequence.

[0093] According to the simulation step sequence, select the effective verification position closest to the first mismatch position from the backoff candidate sequence, and determine the selected effective verification position as the backoff position; take the backoff position as the backoff start point and the deviation boundary of the deviation causal cone as the backoff end point, classify the backoff objects between the backoff start point and the backoff end point into the local backoff range, and retain the content that has been verified outside the local backoff range, and generate a local backoff boundary table.

[0094] S5.2. Based on the local rollback boundary table, restore the affected consumption state and propagation connection to the corresponding rollback position, cancel the state changes formed after the rollback position, and retain the state content that has been verified outside the deviation causal cone, and generate a local rollback state package.

[0095] It should be noted that, based on the local rollback boundary table, the rollback objects are determined according to the rollback position and the local rollback range, and each rollback object is restored to the consumption state and propagation acceptance state corresponding to the rollback position, thus forming the rollback restoration content.

[0096] Based on the rollback recovery content, the state changes formed after the rollback position is canceled along the simulation step sequence, the corresponding propagation inheritance is released and the deviation extension content is cleared to form rollback cleanup content; according to the rollback cleanup content, the state content that has been verified outside the deviation causal cone is retained, and the retained content is connected with the recovered rollback object according to the simulation position to form local rollback state content; then the state continuity and propagation integrity of the local rollback state content at the rollback position are checked, and the verified local rollback state content is uniformly encapsulated to generate a local rollback state package.

[0097] S5.3 Based on the local rollback state package, the stable behavior boundary corresponding to the actual selection state and demand response state in the real market feedback and the rollback position is verified in tandem, and the stable behavior range is corrected based on the continuous real response to generate a behavior boundary correction table.

[0098] It should be noted that, based on the local rollback state packet, the continuous real responses in the real market feedback are arranged in the same position with the corresponding stable behavior boundaries according to the rollback position, and the behavior position, change direction and duration of the continuous real responses are verified item by item to form a response boundary verification sequence.

[0099] Based on the response boundary verification sequence, positions that continuously fall within the stable behavior boundary and whose direction of change is continuous are identified as boundary retention positions; positions that continuously exceed the stable behavior boundary but maintain the same direction of continuity are identified as boundary expansion positions; and positions that cannot be continuously connected after deviation are identified as boundary contraction positions, forming boundary offset segments. According to the boundary offset segments, the stable behavior range corresponding to the boundary retention position is retained, the stable behavior range is extended along the boundary expansion position, and the stable behavior range is withdrawn inward from the boundary contraction position to form the corrected behavior range. Then, the corrected behavior range and corresponding correction direction are collected according to the retreat position and the order of behavior occurrence to generate a behavior boundary correction table.

[0100] S5.4. Based on the behavior boundary correction table, check the information exposure status of each market information, restore the market information that has completed effective exposure to the receptive connection, and set the propagation connection corresponding to the invalid exposure status to the blocked status, and generate a feedback correction status table.

[0101] It should be noted that, according to the behavior boundary correction table, and based on the corrected stable behavior range and the order of behavior occurrence, the information exposure status of each market information is mapped to the consumption status position. The arrival position, presentation process, exposure completion status, source of occurrence, and exposure position of each market information are verified item by item to form an exposure status verification sequence. Based on this sequence, market information that has arrived before the corresponding consumption status is formed, has been continuously and effectively presented, and whose exposure position falls within the corrected stable behavior range is determined as valid exposure information. Market information that has not arrived, has interrupted presentation, has not been effectively presented, or whose exposure position exceeds the corrected stable behavior range is determined as invalid exposure information. Based on the source of occurrence, exposure position, presentation interval, and propagation path corresponding to valid exposure information, the range of market information that the consumption status position can receive is redefined. The redefined information reception range, valid exposure requirements, receivable propagation connections, and invalid exposure blocking positions are uniformly defined as the corrected information reception conditions. Based on the corrected information reception conditions, the propagation connection corresponding to the valid exposure information is restored to the receptive state, the propagation connection corresponding to the invalid exposure information is set to the blocked state, the information transmission after the invalid exposure position is terminated, and the corrected information reception conditions are then aggregated with the behavior boundary correction table to generate a feedback correction status table.

[0102] It should also be noted that information reception conditions include information reception range, effective exposure requirements, receivable transmission links, and invalid exposure blocking locations. These conditions are used to limit the sources, locations, presentation status, and transmission paths of market information that can enter the corresponding consumption status. By retaining only market information that meets the effective exposure requirements and is within the information reception range, and cutting off transmission at invalid exposure locations, it is possible to prevent information that has not actually arrived or has not been fully presented from participating in the subsequent development of the branch, thereby improving the accuracy of consumption status advancement, demand response prediction, and market intervention effect evaluation.

[0103] S5.5 Based on the feedback correction state table, the corrected behavior boundary and information receiving connection are written into the local rollback state packet and merged with the consumption state and propagation connection retained outside the deviation causal cone to generate the branch continuation start table.

[0104] It should be noted that, based on the feedback correction status table, the feedback correction content is written into the local rollback status package according to the rollback position and the order of behavior occurrence, replacing the content to be corrected at the corresponding position in the local rollback status package, thus forming the continuation correction status.

[0105] Based on the continuation correction status, the content retained outside the cone is matched item by item according to the simulation position and the order of propagation. The continuity of consumption status and the consistency of propagation direction on both sides of the retreat boundary are checked to form the corresponding connection content. Based on the corresponding connection content, the continuation correction status and the content retained outside the cone are merged at the corresponding simulation position, and a unified continuation starting point is retained at the retreat position. Continuous propagation is restored at the deviation boundary to form the initial content of the branch continuation. Then, the initial content of the branch continuation is collected according to the market intervention branch and the consistency of the continuation starting point of each branch is verified to generate the branch continuation starting table.

[0106] like Figure 5 The figure shows a comparison of local correction accuracy at different deviation propagation depths. The figure uses deviation propagation depth as the horizontal axis and local correction accuracy as the vertical axis to compare three methods of the present invention: local rollback of the deviation causal cone, rollback within a fixed simulation step range, and rollback of all consumption states. As the deviation propagation depth increases, the present invention can still accurately define the location and propagation connection of the affected consumption states along the deviation propagation chain, only rolling back the states within the deviation causal cone while retaining the verified state content outside the deviation causal cone. Therefore, the local correction accuracy is consistently higher than the comparison methods, demonstrating the innovation of the present invention in transforming "overall error correction" into "directional recovery along the deviation propagation direction."

[0107] The present invention provides a local rollback method for the deviation causal cone: the method first delineates the deviation causal cone based on the first mismatch position and the deviation propagation chain, rolls back only the affected consumption state and propagation connection within the cone to the most recently verified position, and retains the state content that has been verified outside the cone. Therefore, it can maintain a high local correction accuracy even when the deviation propagation depth increases.

[0108] Fixed simulation step range backoff: This method determines the backoff range according to a preset fixed number of simulation steps, and no longer adjusts the boundary based on the actual propagation direction and termination position of the deviation. When the deviation propagation depth is less than or close to the fixed range, partial correction can be completed, but when the deviation exceeds the fixed range, it is easy to miss the downstream affected state, resulting in a gradual decrease in the accuracy of local correction.

[0109] All consumption states rollback: After detecting simulation deviations, this method restores all consumption states and propagation connections to their previous positions. Although this can cover the actual deviation range, it also cancels a large number of state changes that were not affected by the deviations, destroying the original evolution trajectory. Therefore, the overall local correction accuracy is significantly lower than that of this invention.

[0110] S5.6. Based on the branch continuation start table, continue each market intervention branch from the common retreat position, and record the demand response status, propagation acceptance status and remaining deviation range after each branch continuation, generating the branch continuation trajectory.

[0111] It should be noted that, according to the branch continuation start table, the common retreat position is used as the starting point for the continuation of each market intervention branch. The corresponding market intervention conditions are delivered step by step according to the unified simulation step size, so that market information enters the consumption state position along the corrected information receiving connection, and the next simulation step is advanced according to the state formed by the previous simulation step, forming a branch continuation state sequence. Based on the branch continuation state sequence, the response acceptance and propagation direction of adjacent consumption state positions are continuously checked according to the market intervention branch and simulation step order, and the state positions that can be continuously accepted are connected in sequence to form a continuation state trajectory. According to the continuation state trajectory, the continuous simulation positions that have not yet recovered consistency are included in the remaining deviation range, and the continuation state content corresponding to each market intervention branch is collected according to the simulation step order to generate the branch continuation trajectory.

[0112] It should also be noted that the branch continuation trajectory is used to record the demand response state, propagation acceptance state, state change sequence, and remaining deviation range formed after each market intervention branch advances again from the common retreat position. It also reflects the actual role of the corrected behavioral boundaries and information reception conditions in subsequent simulations. By continuously connecting the consumption state positions that can be stably accepted and locating the simulation positions that have not yet returned to consistency, it is possible to determine whether the deviations of each market intervention branch have converged and whether the propagation process has returned to stability. This provides a basis for determining the applicable order of different market intervention branches, improving the accuracy of market behavior projection and the credibility of intervention effect evaluation.

[0113] like Figure 6 As shown in the figure, this is a comparison of the relative error of demand response under different numbers of simulation steps. The figure uses the number of simulation steps as the horizontal axis and the relative error of demand response as the vertical axis to compare three methods of deviation causal cone local regression, fixed simulation step range regression, and full consumption state regression of the present invention. As the number of simulation steps increases, the relative error of demand response formed by the present invention remains at a low level. This indicates that after the present invention corrects the behavioral boundaries and information reception conditions based on real market feedback, it enables each market intervention branch to continuously inherit the original state from the common regression position, reducing the accumulation and diffusion of deviations in subsequent simulations. This reflects the innovation of connecting local correction with branch regression and improves the credibility of market intervention effect evaluation.

[0114] The present invention features a partial rollback of the deviation causal cone: This method delineates the deviation causal cone based on the initial mismatch position and the deviation propagation chain, rolls back only the consumption state and propagation connection affected by the deviation, and combines real market feedback to correct the behavioral boundary and information reception conditions, so that each market intervention branch continues to evolve from the common rollback position. Therefore, as the number of simulation steps increases, the relative error of demand response remains at a low level.

[0115] Fixed simulation step range backoff: This method determines the backoff interval according to the preset number of simulation steps, and no longer dynamically adjusts the backoff range based on the actual propagation depth of the deviation and the termination boundary. When the deviation exceeds the fixed range, some affected states cannot be completely corrected, and the remaining deviation will continue to accumulate during the simulation process. Therefore, the relative error of the demand response continues to rise as the number of simulation steps increases.

[0116] All consumption states rollback: This method restores all consumption states and propagation connections to their previous positions. Although it can cover known deviations, it also cancels a large amount of the original evolution content of unaffected states, causing a new difference between the branch continuation starting point and the real market state. Therefore, its demand response relative error is higher than that of this invention and gradually increases during long-term continuation.

[0117] S5.7 Based on the branch evolution trajectory, the demand response status and dissemination acceptance status of each market intervention branch are compared with the corresponding real market feedback to determine the applicable order of each market intervention branch and generate a market behavior projection plan.

[0118] It should be noted that, based on the branch continuation trajectory, the demand response status and propagation acceptance status of each market intervention branch are unified as the branch continuation status. These are then arranged in tandem with the corresponding real market feedback according to the market intervention branch, simulation step order, and consumption status position, forming a continuation tandem verification sequence. Based on this sequence, the direction of change, acceptance position, and continuity of the branch continuation status are checked level by level against the corresponding real market feedback. Positions that match are marked as recovery positions, and positions that still differ are marked as remaining deviation positions, forming the branch applicability verification content. Based on this verification content, each market intervention branch is ranked according to the continuity of the recovery position, the degree of contraction of the remaining deviation range, and the stability of propagation acceptance. Branches with a high degree of consistency with the corresponding real market feedback and a small remaining deviation range are ranked first, determining the applicability order of each market intervention branch. Finally, the applicable basis and continuation status of each market intervention branch are collected according to the applicability order to generate a market behavior projection scheme.

[0119] This embodiment also provides a computer device applicable to the market behavior extrapolation method based on digital twins, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the market behavior extrapolation method based on digital twins as proposed in the above embodiment.

[0120] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0121] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the market behavior extrapolation method based on digital twins as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0122] In summary, this invention achieves directional recovery of states within the deviation range and continuous evolution of each market intervention branch by: reverting the state and propagation connection of the consumer digital twin affected by the deviation to the most recently verified position based on the deviation causal cone; and correcting the behavioral boundaries and information reception conditions according to real market feedback. This ensures that the unaffected state maintains its original evolutionary trajectory, while consistently verifying the demand response and propagation acceptance state after the evolution. This provides a reliable basis for evaluating the effectiveness of market intervention and improves the accuracy of local correction and the reliability of branch evolution in the digital twin market behavior projection.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for extrapolating market behavior based on digital twins, characterized in that, include: Collect market behavior data, perform anonymous correlation and time-series calibration on the market behavior data, and connect the continuous behaviors before and after market stimulus to generate a market behavior time table; By continuously comparing the historical choices and recent responses of the same consumer in the market behavior time table, the boundaries of stable behavior and the current market status are determined, and the scope of receivable information is limited according to the actual exposure path to generate a visible state map of consumer twins. Based on the visible state diagram of the consumer twin, the consumer digital twin and the market environment are connected according to the information receiving range, and different market intervention conditions are loaded with the same initial state to generate market twin branch scenarios; Based on the market twin branch scenario, each intervention branch is arranged into a unified computable simulation structure. The state is advanced and propagated according to the same initial state and simulation step size. The state of each step calculation is verified against the real market feedback. The mismatch location and affected range are determined along the deviation propagation direction, and the deviation causal cone is generated. Based on the bias causal cone, the state and communication connection of the consumer digital twin affected by the bias are reverted to the most recently verified position. The behavioral boundaries and information reception conditions are corrected according to the real market feedback, and each market intervention branch is continued. The demand response and communication acceptance status after the continued simulation are compared to generate a market behavior projection plan.

2. The market behavior extrapolation method based on digital twins as described in claim 1, characterized in that, The process of collecting market behavior data, anonymizing and calibrating the data over time, and connecting consecutive behaviors before and after market stimuli to generate a market behavior time table is as follows: The consumer object identifiers in the market behavior data are anonymously replaced. Market behaviors belonging to the same consumer object are associated according to the content and source of the behavior. The recording time of each market behavior is calibrated by the time the market stimulus occurs, generating an anonymously calibrated behavior sequence. Based on anonymous calibration behavior sequences, historical behaviors before market stimulus are continuously linked with response behaviors after market stimulus, and duplicate behaviors and time-series conflict behaviors are eliminated to generate a market behavior time table.

3. The market behavior extrapolation method based on digital twins as described in claim 2, characterized in that, The process involves continuously comparing the historical choices and recent responses of the same consumer in the market behavior time table to determine the stable behavior boundary and the current market state. Based on the actual exposure path, the scope of receivable information is limited, and a consumer twin visible state map is generated. The specific steps are as follows: Based on the market behavior time table, according to the anonymous consumer and the type of market stimulus, the historical choices before the stimulus and the recent responses after the stimulus are matched and corresponding according to the same consumption opportunities, and then continuously linked according to the time of the behavior to generate a continuous behavior comparison chain. Based on the continuous behavior comparison chain, the behavior stability verification elements of each pairing position are continuously verified. Adjacent pairing positions that maintain stable acceptance are merged into stable behavior segments. Based on the stable behavior segments, the range of sustainable behavior of the consumer is limited, and a stable behavior boundary table is generated. Based on the stable behavior boundary table, the position of the consumer's most recent selection state and recent response state is checked against the corresponding stable behavior range. The consumer behavior is determined to be either a stable continuation state or a boundary shift state. The corresponding state mark is written into each behavior position to generate the current market state map. Based on the current market state map, the information exposure status process of market information is tracked according to the actual exposure order in the market behavior time table. Market information that has arrived and been effectively presented before the recent response occurs is retained, market information with invalid exposure status is removed, and an actual exposure reachable sequence is generated. Based on the actual exposure reach sequence, the behavior range in the stable behavior boundary table, the consumption status in the current market status graph, and the actual reachable market information are respectively set as behavior boundary nodes, consumption status nodes, and visible information nodes, and a consumption twin visible status graph is established between the nodes according to the actual exposure order.

4. The market behavior extrapolation method based on digital twins as described in claim 3, characterized in that, The process involves connecting the consumer digital twin with the market environment based on the visible state diagram of the consumer twin, according to the information reception range, and loading different market intervention conditions with the same initial state. The specific steps are as follows: Based on the visible state diagram of the consumer twin, the effective information nodes in the market environment are connected to the corresponding consumer state nodes according to the actual exposure order and information reception range of the visible information nodes, and the information transmission is cut off at the invalid exposure position to generate a twin environment reception connection table. Select the most recent common verification position before the start of different market interventions from the twin environment receiving connection table, and encapsulate the stable behavior boundary, current market state, last effective exposure position, incomplete response state and effective receiving connection in a unified way to generate a unified initial state packet; Based on the unified initial state package, each market intervention condition is verified accordingly, and the full loading and boundary loading are determined according to the information receiving scope to generate an intervention condition loading table; Based on the intervention condition loading table, a unified initial state package is copied according to different market intervention conditions. The stable behavior boundaries, current market state and original receiving connection of each branch are kept consistent. The corresponding intervention content and loading state are written to generate the market twin branch initial state table.

5. The market behavior extrapolation method based on digital twins as described in claim 4, characterized in that, The market twin branch scenario is obtained by connecting each branch to the same market environment benchmark and unifying the initial benchmark of the branches according to the initial state table of the market twin branches, and then solidifying different market intervention conditions to the corresponding branches.

6. The market behavior extrapolation method based on digital twins as described in claim 1, characterized in that, Based on the market twin branch scenario, each intervention branch is arranged into a unified computable simulation structure. The state is advanced and propagated according to the same initial state and simulation step size. The state of each step is then verified against the actual market feedback. The mismatch location and affected range are determined along the deviation propagation direction. The specific steps are as follows: Based on the market twin branch scenario, the simulation orchestration information of each branch is uniformly configured according to the market intervention branch to obtain a unified simulation orchestration table. Based on the unified simulation orchestration table, the initial benchmark of the branches is used as a common starting point, and the market information simulation evolution state is advanced sequentially according to the same simulation step size. The state propagation trajectory of each state is recorded to generate a step-by-step simulation trajectory. Based on the step-by-step simulation trajectory, the consumption status and propagation changes in each simulation step are compared with the real market feedback in the corresponding time period to determine the deviation status of each simulation position and generate a corresponding deviation verification table. Based on the same position deviation verification table, the consumer state position where the first continuous deviation is formed is located according to the simulation step sequence. The visible information position, propagation connection, market intervention branch and simulation step position corresponding to the first continuous deviation position are checked back to generate the first mismatch positioning table. Based on the first mismatch location table, the deviation extension state is tracked step by step along the state transmission direction in the step-by-step simulation trajectory from the first mismatch position. The deviation state in adjacent simulation steps is continuously transmitted along the same propagation connection direction, and the tracking is terminated at the point where the propagation direction is reversed, thus generating a deviation propagation chain.

7. The market behavior extrapolation method based on digital twins as described in claim 6, characterized in that, The deviation causal cone is obtained by determining the first mismatch position as the deviation starting point based on the deviation propagation chain, and collecting the affected consumption state positions, simulation step ranges, visible information positions, and market intervention branches according to the propagation sequence.

8. The market behavior extrapolation method based on digital twins as described in claim 1, characterized in that, The deviation-based causal cone method reverts the consumer digital twin state and propagation connections affected by deviations to the most recently verified position, corrects behavioral boundaries and information reception conditions based on real market feedback, and continues to simulate each market intervention branch. By comparing the demand response and propagation acceptance status after the simulation, a market behavior projection plan is generated. The specific steps are as follows: Based on the deviation causal cone, the positions of each affected consumption state are checked back along the deviation propagation direction. The simulation position that is most recently consistent with the real market feedback before the first mismatch position is determined as the back-off position. The consumption state positions and propagation connections that need to be backed up are delineated, and a local back-off boundary table is generated. Based on the local rollback boundary table, the affected consumption state and propagation connection are restored to the corresponding rollback position, the state changes formed after the rollback position are undone, and the state content that has been verified outside the deviation causal cone is retained to generate a local rollback state package. Based on the local rollback state package, the stable behavior boundary corresponding to the actual selection state and demand response state in the real market feedback and the rollback position is verified in tandem, and the stable behavior range is corrected based on the continuous real response to generate a behavior boundary correction table. Based on the behavior boundary correction table, check the information exposure status of each market information, restore the market information that has completed effective exposure to the receptive connection, and set the propagation connection corresponding to the invalid exposure status to the blocked status, and generate a feedback correction status table. Based on the feedback correction state table, the corrected behavior boundary and information receiving connection are written into the local backoff state packet and merged with the consumption state and propagation connection retained outside the deviation causal cone to generate the branch continuation start table. Based on the branch continuation start table, each market intervention branch is continued from the common retreat position, and the demand response status, propagation acceptance status and remaining deviation range of each branch after continuation are recorded to generate the branch continuation trajectory. Based on the branch evolution trajectory, the demand response status and dissemination acceptance status of each market intervention branch are compared with the corresponding real market feedback to determine the applicable order of each market intervention branch and generate a market behavior projection plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the market behavior inference method based on digital twins as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the market behavior inference method based on digital twins as described in any one of claims 1 to 8.

Images