An advertisement delivery control method, device, equipment and medium

Abstract

The present application relates to the technical field of information processing, and particularly relates to an advertisement putting control method, device, equipment and medium, the method comprises the following steps: determining a deviation correction result corresponding to each putting unit according to a trajectory deviation result between each actual putting state trajectory and a corresponding mirror putting state trajectory; generating candidate putting control information corresponding to each putting unit according to the deviation correction result, and determining a state impact result corresponding to each candidate putting control information; performing constraint correction processing on each candidate putting control information based on the state impact result to obtain target putting control information; and performing control on a corresponding advertisement putting process based on the target putting control information. The present application has solved the problem that the state change of the control object in the prior art is difficult to be effectively characterized, which leads to the difficulty in correcting the control deviation and the lack of control stability.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to an advertising placement control method, apparatus, equipment, and medium. Background Technology

[0002] With the continuous development of internet advertising, the control requirements in the advertising process are gradually shifting from static configuration to dynamic adjustment. Existing advertising technologies typically analyze advertising effectiveness by collecting exposure data, click data, conversion data, and budget consumption data during the advertising process, and adjust the advertising strategy based on the analysis results to improve advertising effectiveness and resource utilization.

[0003] However, most existing advertising delivery control methods are based on feedback results at the current moment or within the current period, typically focusing on responding to single feedback results or changes in local indicators. They lack a holistic portrayal of the continuous changes in the delivery status, making it difficult to accurately distinguish between normal state changes and abnormal state deviations when the advertising delivery environment fluctuates. This can easily lead to control lag, large control deviations, or interference between control actions. Furthermore, when adjusting delivery control actions, existing technologies often focus on the control result itself, neglecting the potential state fluctuations and cascading effects that may result from the control actions. This makes the advertising delivery process prone to poor control stability in dynamically changing scenarios. Summary of the Invention

[0004] In order to address the problem that changes in the state of the controlled object are difficult to be effectively represented in the prior art, resulting in difficulty in correcting control deviations and insufficient control stability, this application provides an advertising delivery control method, device, equipment and medium.

[0005] The above-mentioned objective of this application is achieved through the following technical solution:

[0006] An advertising delivery control method, the advertising delivery control method comprising:

[0007] Obtain real-time feedback data for each ad placement unit during the ad delivery process;

[0008] Based on the real-time feedback data, construct the actual delivery status trajectory corresponding to each delivery unit;

[0009] Based on the historical delivery status information and preset control constraint information corresponding to each delivery unit, a mirror delivery status trajectory corresponding to each delivery unit is generated.

[0010] Based on the trajectory deviation results between each actual delivery state trajectory and the corresponding mirrored delivery state trajectory, the deviation correction result corresponding to each delivery unit is determined;

[0011] Based on the deviation correction results, candidate delivery control information corresponding to each delivery unit is generated, and the state impact result corresponding to each candidate delivery control information is determined.

[0012] Based on the state impact results, constraint correction processing is performed on the candidate delivery control information to obtain the target delivery control information;

[0013] Based on the target delivery control information, control is exercised over the corresponding advertising delivery process.

[0014] By adopting the above technical solution, by acquiring real-time feedback data corresponding to each placement unit during the advertising placement process, and constructing the actual placement state trajectory corresponding to each placement unit based on the real-time feedback data, the state changes during the advertising placement process can have a continuous representation basis. By generating mirror placement state trajectories corresponding to each placement unit based on the historical placement state information and preset control constraint information, and determining the deviation correction results corresponding to each placement unit based on the trajectory deviation results between each actual placement state trajectory and the corresponding mirror placement state trajectory, the identification and correction of deviation content can have a corresponding basis. By generating candidate placement control information based on the deviation correction results and combining it with the state impact results to obtain target placement control information, the control content and state changes during the advertising placement process can be kept connected.

[0015] Preferably, the step of generating a mirrored delivery state trajectory for each delivery unit based on the historical delivery state information and preset control constraint information corresponding to each delivery unit includes:

[0016] Based on the historical delivery status information corresponding to each delivery unit, determine the corresponding historical status migration relationship information;

[0017] Based on the preset control constraint information, constraint filtering processing is performed on the historical state transition relationship information to obtain the target state transition relationship information;

[0018] Based on the target state transition relationship information, the mirrored delivery state sequence information corresponding to each delivery unit is deduced and generated.

[0019] Based on the image delivery state sequence information, construct the image delivery state trajectory corresponding to each delivery unit.

[0020] By adopting the above technical solution, and by determining the corresponding historical state transition relationship information based on the historical state information of each delivery unit, the preceding and following changes in the historical state information can be organized. By performing constraint filtering processing on the historical state transition relationship information according to the preset control constraint information, the target state transition relationship information can be obtained, which can limit the subsequent state unfolding process to the range corresponding to the preset control constraint information. By generating the mirrored delivery state sequence information corresponding to each delivery unit based on the target state transition relationship information, the mirrored state content can be gradually unfolded in chronological order. By constructing the mirrored delivery state trajectory corresponding to each delivery unit based on the mirrored delivery state sequence information, the mirrored state change relationship can form continuous corresponding content, thereby facilitating subsequent corresponding analysis around the actual delivery state trajectory and the mirrored delivery state trajectory.

[0021] Preferably, the step of generating mirrored delivery state sequence information corresponding to each delivery unit based on the target state transition relationship information includes:

[0022] Based on the target state migration relationship information, determine the initial mirror state node information corresponding to each of the deployment units;

[0023] Based on the initial mirror state node information and the target state transition relationship information, a forward recursive process is performed to obtain the forward mirror state chain information corresponding to each deployment unit;

[0024] Perform hierarchical partitioning on each state node in the forward mirror state chain information to obtain state hierarchy result information;

[0025] Based on the state hierarchy result information, freeze processing is performed on state nodes that meet the preset stability conditions, and reverse backtracking verification processing is performed on state nodes that do not meet the preset stability conditions to obtain state adjustment result information.

[0026] Based on the state adjustment result information, iterative chain completion processing is performed on the corresponding unfrozen state nodes to obtain the image delivery state sequence information corresponding to each delivery unit.

[0027] By adopting the above technical solution, and by determining the initial mirror state node information corresponding to each deployment unit based on the target state transition relationship information, the mirror state unfolding can have a starting basis. By performing forward recursion processing based on the initial mirror state node information and the target state transition relationship information, the forward mirror state chain information corresponding to each deployment unit can be obtained, enabling each state node to unfold step by step in chronological order. By performing hierarchical division processing on each state node in the forward mirror state chain information, and combining freezing processing, reverse backtracking verification processing, and iterative chain completion processing, the connection relationship between state nodes can be organized and completed, thereby forming a continuous mirror deployment state sequence information.

[0028] Preferably, determining the deviation correction result for each delivery unit based on the trajectory deviation result between each actual delivery state trajectory and the corresponding mirrored delivery state trajectory includes:

[0029] Based on the trajectory deviation results between the actual delivery state trajectory and the corresponding mirrored delivery state trajectory, the deviation state segment information corresponding to each delivery unit is extracted;

[0030] Based on the deviation state segment information, determine the corresponding mirror reference node information and actual drift node information;

[0031] Based on the corresponding deviation relationship between the mirror reference node information and the actual drift node information, the reverse traction path information corresponding to the deviation state segment is generated;

[0032] Based on the reverse traction path information, a step-by-step pullback process is performed on the actual drift node information to obtain the pullback correction state chain information.

[0033] Based on the pullback correction state chain information, the deviation correction result corresponding to each delivery unit is generated.

[0034] By adopting the above technical solution, by extracting deviation state segment information based on the trajectory deviation result between the actual delivery state trajectory and the corresponding mirror delivery state trajectory, and determining the corresponding mirror reference node information and actual drift node information, the positioning range of the deviation content and the corresponding node relationship can be sorted out. By generating reverse traction path information based on the corresponding deviation relationship, and performing step-by-step pullback processing on the actual drift node information based on the reverse traction path information, the state content within the deviation state segment can be gradually adjusted back along a predetermined direction, thereby forming the corresponding deviation correction result.

[0035] Preferably, the step of generating candidate delivery control information corresponding to each delivery unit based on the deviation correction result, and determining the state impact result corresponding to each candidate delivery control information, includes:

[0036] Based on the deviation correction results, determine the corresponding correction traction node information;

[0037] Based on the corrected traction node information, corresponding candidate delivery control path information is generated;

[0038] Based on the candidate delivery control path information, impact layer expansion processing is performed on the corresponding state nodes to obtain the initial impact layer information, the transition impact layer information, and the final impact layer information.

[0039] Based on the initial impact layer information, the transition impact layer information, and the final impact layer information, identify the impact isolation layer information in each of the candidate delivery control path information;

[0040] Based on the impact isolation layer information, inter-layer peak shaving processing is performed on each of the candidate delivery control path information to obtain the state impact result corresponding to each of the candidate delivery control information.

[0041] By adopting the above technical solution, by determining the corresponding correction traction node information based on the deviation correction result, and generating the corresponding candidate delivery control path information around the correction traction node information, the candidate delivery control content can be unfolded sequentially over time. By performing impact layer unfolding processing on the corresponding state nodes based on each candidate delivery control path information and identifying impact isolation layer information, the state fluctuation relationships at different levels can be distinguished. By performing inter-layer peak shaving processing on each candidate delivery control path information based on the impact isolation layer information, the state impact results corresponding to each candidate delivery control information can be organized, thereby facilitating subsequent constraint adjustment around the candidate delivery control information.

[0042] Preferably, the step of performing constraint correction processing on each of the candidate delivery control information based on each of the state impact results to obtain the target delivery control information includes:

[0043] Based on the state impact result corresponding to the candidate delivery control information, the positive impact constraint interval and the reverse pullback constraint interval are divided.

[0044] Based on the positive impact constraint interval, the candidate delivery control information is forward-constrained and folded to form positive constraint control chain information;

[0045] Based on the reverse pullback constraint interval, the positive constraint control chain information is reverse compensated and folded to form bidirectional folded control chain information;

[0046] Based on the bidirectional folding control chain information, determine the stable output range information;

[0047] Based on the stable output range information, the corresponding bidirectional folding control chain information is locked to obtain the target delivery control information corresponding to each delivery unit.

[0048] By adopting the above technical solution, and by dividing the forward impact constraint interval and the reverse pullback constraint interval according to the state impact result corresponding to the candidate delivery control information, the constraint range of the state impact content in the time sequence can be distinguished. By performing forward constraint folding on the candidate delivery control information based on the forward impact constraint interval, and performing reverse compensation folding on the forward constraint control chain information based on the reverse pullback constraint interval, the control content of the candidate delivery control information in both the forward extension and reverse pullback directions can be organized. By determining the stable output interval information according to the bidirectional folded control chain information and locking the corresponding bidirectional folded control chain information, the target delivery control information corresponding to each delivery unit can form a clear output range.

[0049] Preferably, the step of controlling the corresponding advertising delivery process based on the target delivery control information includes:

[0050] Based on the target delivery control information corresponding to the delivery unit, generate the corresponding master control execution chain information and shadow control takeover chain information;

[0051] Based on the master control execution chain information, control is executed on the corresponding advertising delivery process, and based on the shadow control takeover chain information, reverse approximation is performed on the corresponding mirror delivery state trajectory to obtain control takeover approximation result information;

[0052] Based on the control and approximation result information, determine the corresponding trajectory matching and overtaking result information;

[0053] If the trajectory alignment and overtaking result information meets the preset control seizure conditions, the shadow control seizure chain information replaces the main control execution chain information; if the trajectory alignment and overtaking result information does not meet the preset control seizure conditions, the main control execution chain information is maintained.

[0054] Based on the replaced or retained execution chain information, control over the corresponding advertising delivery process is achieved.

[0055] By adopting the above technical solution, and by generating corresponding main control execution chain information and shadow control takeover chain information based on the target delivery control information corresponding to the delivery unit, the execution control content and the corresponding tracking content can form a chain correspondence. By executing control on the corresponding advertising delivery process based on the main control execution chain information, and performing reverse approximation on the corresponding mirror delivery state trajectory based on the shadow control takeover chain information, the execution content in the advertising delivery process and the approximation content in the mirror delivery state trajectory can form a corresponding comparison basis. By determining the corresponding trajectory fitting and overtaking result information based on the control takeover approximation result information, and replacing it when the preset control takeover conditions are met, and maintaining the main control execution chain information when the preset control takeover conditions are not met, the subsequent control content can be connected according to the predetermined judgment relationship.

[0056] The second objective of this invention is achieved through the following technical solution:

[0057] An advertising delivery control device, the advertising delivery control device comprising:

[0058] The feedback acquisition module is used to acquire real-time feedback data corresponding to each ad delivery unit during the ad delivery process;

[0059] The status construction module is used to construct the actual delivery status trajectory corresponding to each delivery unit based on the real-time feedback data.

[0060] The mirror trajectory module is used to generate mirror delivery status trajectories for each delivery unit based on the historical delivery status information and preset control constraint information corresponding to each delivery unit.

[0061] The deviation correction module is used to determine the deviation correction result for each delivery unit based on the trajectory deviation result between each actual delivery state trajectory and the corresponding mirrored delivery state trajectory.

[0062] The candidate control module is used to generate candidate delivery control information corresponding to each delivery unit based on the deviation correction result, and to determine the state impact result corresponding to each candidate delivery control information.

[0063] The target control module is used to perform constraint correction processing on each of the candidate delivery control information based on the state impact results, so as to obtain the target delivery control information;

[0064] The execution control module is used to control the corresponding advertising delivery process based on the target delivery control information.

[0065] By adopting the above technical solution, by acquiring real-time feedback data corresponding to each placement unit during the advertising placement process, and constructing the actual placement state trajectory corresponding to each placement unit based on the real-time feedback data, the state changes during the advertising placement process can have a continuous representation basis. By generating mirror placement state trajectories corresponding to each placement unit based on the historical placement state information and preset control constraint information, and determining the deviation correction results corresponding to each placement unit based on the trajectory deviation results between each actual placement state trajectory and the corresponding mirror placement state trajectory, the identification and correction of deviation content can have a corresponding basis. By generating candidate placement control information based on the deviation correction results and combining it with the state impact results to obtain target placement control information, the control content and state changes during the advertising placement process can be kept connected.

[0066] The above-mentioned objective three of this application is achieved through the following technical solution:

[0067] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the aforementioned advertising delivery control method.

[0068] The fourth objective of this application is achieved through the following technical solution:

[0069] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned advertising delivery control method.

[0070] In summary, this application includes at least one of the following beneficial technical effects:

[0071] By acquiring real-time feedback data corresponding to each advertising unit during the advertising campaign, and constructing the actual advertising state trajectory corresponding to each advertising unit based on the real-time feedback data, the state changes during the advertising campaign can have a continuous representation basis. By generating mirrored advertising state trajectories corresponding to each advertising unit based on the historical advertising state information and preset control constraint information, and determining the deviation correction results corresponding to each advertising unit based on the trajectory deviation results between the actual advertising state trajectory and the corresponding mirrored advertising state trajectory, the identification and correction of deviation content can have a corresponding basis. By generating candidate advertising control information based on the deviation correction results and combining it with the state impact results to obtain target advertising control information, the control content and state changes during the advertising campaign can be kept connected. Attached Figure Description

[0072] Figure 1This is a flowchart of an advertising delivery control method according to an embodiment of this application.

[0073] Figure 2 This is a principle block diagram of an advertising delivery control device in one embodiment of this application. Detailed Implementation

[0074] The present application will be further described in detail below with reference to the accompanying drawings.

[0075] In one embodiment, such as Figure 1 As shown, this application discloses an advertising placement control method, which specifically includes the following steps:

[0076] S10: Obtain real-time feedback data for each ad delivery unit during the ad delivery process.

[0077] In this embodiment, the advertising delivery process refers to the process of displaying advertising content in multiple delivery locations, delivery time periods, or delivery channels according to preset delivery configurations and receiving user feedback. The delivery unit refers to the basic delivery object that is distinguished according to preset division rules during the advertising delivery process. The delivery unit can be divided according to any dimension of advertising plan, advertising group, advertising material, delivery location, delivery channel, or delivery time period. Real-time feedback data refers to the feedback information that is continuously generated and synchronously recorded during the advertising delivery process as the advertisement is displayed and the user interacts. Real-time feedback data includes one or more of the following: display records, click records, conversion records, consumption records, or interaction records corresponding to the delivery unit.

[0078] Specifically, for each ad placement unit already defined during the advertising campaign, feedback records generated by each unit within the current campaign period are continuously read in chronological order. For each feedback record, the corresponding ad placement unit identifier, feedback type identifier, feedback occurrence time, and feedback value are extracted. Feedback records with the same ad placement unit identifier are grouped into the corresponding ad placement unit. Records with the feedback type identifier "display" are designated as display records, those with the feedback type identifier "click" are designated as click records, those with the feedback type identifier "conversion" are designated as conversion records, and those with the feedback type identifier "spend" are designated as spend records. The interaction records are used as interaction records, and the feedback records in the same campaign unit are sorted according to the time of the feedback occurrence. When there are duplicate uploaded feedback records, the duplicate records are identified by the combination of campaign unit identifier, feedback type identifier, feedback occurrence time and feedback value, and the first record is retained. When there are feedback records with missing fields, the feedback records that are missing campaign unit identifier or feedback occurrence time are removed. After completing the inclusion, sorting, deduplication and removal, the display records, click records, conversion records, consumption records or interaction records corresponding to each campaign unit are collected into the real-time feedback data corresponding to each campaign unit.

[0079] S20: Based on real-time feedback data, construct the actual delivery status trajectory corresponding to each delivery unit.

[0080] In this embodiment, the actual delivery status trajectory refers to the sequence of status records formed by continuously representing the status changes of the delivery unit during the advertising delivery process in chronological order. The status changes in the actual delivery status trajectory are used to reflect the feedback change process of the delivery unit at different time positions. Constructing the actual delivery status trajectory means organizing, arranging and connecting the status content of the delivery unit within a continuous time range based on the various feedback records and feedback occurrence times contained in the real-time feedback data, so as to form a status change path corresponding to the time sequence.

[0081] Specifically, for the real-time feedback data corresponding to each ad placement unit, the display records, click records, conversion records, consumption records, and interaction records are first uniformly sorted according to the feedback occurrence time. Then, the sorted real-time feedback data is sequentially divided into multiple consecutive time segments by a preset time length. Within each time segment, the number of displays corresponding to the display records, the number of clicks corresponding to the click records, the number of conversions corresponding to the conversion records, the consumption value corresponding to the consumption records, and the number of interactions corresponding to the interaction records are counted. The number of displays, clicks, conversions, consumption values, and interactions are collectively used as the ad placement status content under the corresponding time segment. Between adjacent time segments, the ad placement status content corresponding to each time segment is sequentially concatenated according to the order of the feedback occurrence time, so that the ad placement status content of the previous time segment is continuously connected with the ad placement status content of the next time segment. After completing the concatenation of the ad placement status content of all time segments, the ad placement status content arranged in chronological order is aggregated into the actual ad placement status trajectory of the corresponding ad placement unit.

[0082] S30: Based on the historical delivery status information and preset control constraint information of each delivery unit, generate the mirror delivery status trajectory of each delivery unit.

[0083] In this embodiment, historical delivery status information refers to the status changes of the delivery unit recorded in chronological order during the historical delivery process. Historical delivery status information is used to characterize the changes in the number of impressions, clicks, conversions, consumption, and interactions of the delivery unit within a historical time range. Preset control constraint information refers to the status restrictions set in advance during the ad delivery control process. Preset control constraint information is used to limit the range, order, or conditions of status changes of the delivery unit. Mirrored delivery status trajectory refers to the reference status trajectory corresponding to the actual delivery status trajectory, which is formed by combining historical delivery status information with preset control constraint information.

[0084] Specifically, for the historical delivery status information corresponding to each delivery unit, the status content corresponding to each time position is first arranged in historical time sequence. Then, the corresponding status restriction content in the preset control constraint information is read, and the status content corresponding to each time position is matched with the status restriction content one by one. If the status content corresponding to a certain time position meets the status restriction content, the status content corresponding to the current time position is retained. If the status content corresponding to a certain time position does not meet the status restriction content, the number of impressions, clicks, conversions, consumption, and interactions corresponding to the current time position are adjusted according to the status restriction content. Among them, values ​​exceeding the upper limit of the status restriction content are compressed, values ​​below the lower limit of the status restriction content are padded, and status content that does not conform to the status change sequence is corrected. After completing the filtering, adjustment, and correction of the status content corresponding to each time position, the status content corresponding to each time position is connected sequentially in historical time sequence to form a mirrored delivery status trajectory that matches the corresponding delivery unit.

[0085] S40: Based on the trajectory deviation results between each actual delivery status trajectory and the corresponding mirrored delivery status trajectory, determine the deviation correction result for each delivery unit.

[0086] In this embodiment, the trajectory deviation result refers to the difference formed after comparing the actual delivery state trajectory and the mirrored delivery state trajectory in the same time sequence. The trajectory deviation result is used to characterize the deviation of the delivery unit in terms of the position of state change, the magnitude of state change, and the continuity of state change. The deviation correction result refers to the correction content formed after sorting out the deviation content in the actual delivery state trajectory based on the trajectory deviation result. The deviation correction result is used to characterize the state part of each delivery unit that needs to be corrected before subsequent control.

[0087] Specifically, for the actual and mirrored delivery status trajectories corresponding to each delivery unit, the content of each status in the two trajectories is first matched one-to-one according to the same time sequence. The number of impressions, clicks, conversions, consumption, and interactions corresponding to each time position in the actual delivery status trajectory are compared item by item with the number of impressions, clicks, conversions, consumption, and interactions corresponding to the same time position in the mirrored delivery status trajectory. During the item-by-item comparison, the differences in the number of impressions, clicks, conversions, consumption, and interactions are recorded. Deviation segments are identified based on whether each difference appears continuously in consecutive time positions. If the same direction difference appears in multiple consecutive time positions, the multiple consecutive time positions are merged into the same deviation segment. Within the deviation segment, the order of the difference change corresponding to each time position is read to determine whether the difference in the deviation segment gradually increases, gradually decreases, or fluctuates alternately. Then, based on the position range, difference size, and difference change order of the deviation segment, deviation content matching the corresponding delivery unit is compiled, and all deviation content is aggregated into the deviation correction result for the corresponding delivery unit.

[0088] S50: Based on the deviation correction results, generate candidate delivery control information for each delivery unit and determine the state impact result corresponding to each candidate delivery control information.

[0089] In this embodiment, candidate delivery control information refers to the candidate control content that is organized based on the deviation correction results and used to change the state change content of the delivery unit. The candidate delivery control information is used to characterize the adjustment position, adjustment direction and adjustment magnitude corresponding to different deviation content. The state impact result refers to the impact of the candidate delivery control information on the continuous relationship of state change after it acts on the corresponding deviation content. The state impact result is used to characterize the degree of fluctuation caused by the candidate delivery control information on the state change of adjacent time positions when changing the deviation content.

[0090] Specifically, for the deviation correction results corresponding to each delivery unit, the time location range, difference direction, and difference magnitude corresponding to each deviation content are first read. The time location range is used as the adjustment position corresponding to the candidate delivery control information, the difference direction is used as the adjustment direction corresponding to the candidate delivery control information, and the difference magnitude is used as the adjustment magnitude corresponding to the candidate delivery control information. Multiple candidate delivery control information are formed around the same deviation content. After forming each candidate delivery control information, the adjustment position, adjustment direction, and adjustment magnitude in the candidate delivery control information are sequentially mapped to the deviation content corresponding to the deviation correction result to obtain the state after the candidate effect. The system first analyzes the changes in state content, then reads the differences between the state changes after the candidate's action and the deviation before the candidate's action in chronological order at the previous, current, and next time positions. It records the changes in display count, click count, conversion count, consumption value, and interaction count. Based on whether each change value shows a sudden increase, sudden decrease, or continuous jump between adjacent time positions, it determines the state fluctuation content corresponding to the candidate's delivery control information. After completing the organization of the state fluctuation content corresponding to each candidate's delivery control information, the state fluctuation content is collected into the state impact results corresponding to each candidate's delivery control information.

[0091] S60: Based on the impact results of each state, perform constraint correction processing on each candidate delivery control information to obtain the target delivery control information.

[0092] In this embodiment, the target delivery control information refers to the control content formed after screening and adjustment based on the candidate delivery control information and the state impact results. The target delivery control information is used to characterize the adjustment position, adjustment direction and adjustment range that actually enter the execution stage in the corresponding delivery unit. The constraint correction process refers to the process of restricting, compressing or correcting the adjustment content in the candidate delivery control information based on the state fluctuations reflected in the state impact results.

[0093] Specifically, for each candidate delivery control information corresponding to each delivery unit and the state impact result corresponding to each candidate delivery control information, the state fluctuation content corresponding to each state impact result is first read. The constraint position corresponding to the candidate delivery control information is determined according to the time position range of the state fluctuation content. The constraint strength corresponding to the candidate delivery control information is determined according to the magnitude of the change value in the state fluctuation content. Within the constraint position, the adjustment range in the candidate delivery control information is reduced. When there is continuous state fluctuation content between the constraint position and the adjacent time position, the adjustment position in the candidate delivery control information is shrunk so that the adjustment position avoids the continuous state fluctuation content. When the state fluctuation content covers multiple adjacent time positions, the adjustment direction in the candidate delivery control information is corrected to ensure that the state change content at adjacent time positions maintains a consistent change order. After completing the adjustment amplitude reduction, adjustment position contraction, and adjustment direction correction, the adjustment position, adjustment direction, and adjustment amplitude corresponding to each candidate delivery control information are reread. Each candidate delivery control information is then used as the candidate control content under the corresponding delivery unit. Control content with fewer constraint positions and continuous adjustment content is retained from each candidate control content, and the retained control content is aggregated into the target delivery control information of the corresponding delivery unit.

[0094] S70: Based on the target delivery control information, control is exercised over the corresponding advertising delivery process.

[0095] In this embodiment, execution control refers to the process of actually adjusting the state content corresponding to the advertising unit during the advertising delivery process according to the adjustment position, adjustment direction and adjustment range determined in the target delivery control information.

[0096] Specifically, for each ad placement unit, the target ad placement control information is first read, including the adjustment position, adjustment direction, and adjustment range. The control order for each ad placement unit is determined according to the time sequence of the adjustment position during the ad placement process. When the time position corresponding to the adjustment position is reached, the adjustment direction is mapped to the status content at the current time position. If the number of impressions changes according to the adjustment direction, the number of impressions is increased or decreased according to the adjustment range. If the number of clicks changes according to the adjustment direction, the number of clicks is increased or decreased according to the adjustment range. If the number of conversions changes according to the adjustment direction, the number of conversions is increased or decreased according to the adjustment range. If the cost changes according to the adjustment direction, the cost is increased or decreased according to the adjustment range. If the number of interactions changes according to the adjustment direction, the number of interactions is increased or decreased according to the adjustment range. If multiple adjustment positions correspond to the same ad placement unit, the adjustment content corresponding to each adjustment position is executed sequentially according to the control order. After completing the adjustment content corresponding to each adjustment position, the executed status content is written into the current status record of the corresponding ad placement unit during the ad placement process, thus completing the control of the corresponding ad placement process based on the target ad placement control information.

[0097] In one embodiment, step S30, namely generating a mirrored delivery state trajectory for each delivery unit based on the historical delivery state information and preset control constraint information corresponding to each delivery unit, includes:

[0098] S301: Determine the corresponding historical status migration relationship information based on the historical delivery status information of each delivery unit.

[0099] In this embodiment, historical state migration relationship information refers to the relationship between the state content changes between adjacent time positions in the historical delivery state information corresponding to the same delivery unit. Historical state migration relationship information is used to characterize the direction and magnitude of change of the number of impressions, clicks, conversions, consumption values ​​and interaction counts in time sequence.

[0100] Specifically, for the historical delivery status information corresponding to each delivery unit, the status content corresponding to each time position is first read in historical time sequence, and the status content corresponding to the previous time position is recorded as the first status content, and the status content corresponding to the next time position is recorded as the second status content. Then, the number of impressions, clicks, conversions, consumption, and interactions in the first status content, and the number of impressions, clicks, conversions, consumption, and interactions in the second status content are extracted respectively. Before and after comparison is performed for the same type of status content. Among them, the number of impressions in the second status content is subtracted from the number of impressions in the first status content to obtain the change in the number of impressions. The calculation formula is as follows: Subtract the number of clicks in the first state from the number of clicks in the second state content to obtain the change in click count. The calculation formula is as follows: Subtract the number of transformations in the first state from the number of transformations in the second state to obtain the change in the number of transformations. The calculation formula is as follows: Subtract the consumption value from the consumption value in the first state from the consumption value in the second state to obtain the change in consumption value. The calculation formula is as follows: The change in interaction count is obtained by subtracting the number of interactions in the first state from the number of interactions in the second state content. The calculation formula is as follows: After obtaining each change value, the corresponding change direction is determined according to the sign of each change value. If the change value is greater than zero, it is determined to be an upward direction; if the change value is less than zero, it is determined to be a downward direction; and if the change value is equal to zero, it is determined to be a stable direction. Then, the change values ​​of the number of impressions, clicks, conversions, consumption, and interactions are associated with the corresponding change directions to form the state transition content between adjacent time positions before and after the current time. The state transition content of each group is then collected along the historical time sequence to obtain the historical state transition relationship information of the corresponding delivery unit.

[0101] S302: Based on the preset control constraint information, perform constraint filtering processing on the historical state transition relationship information to obtain the target state transition relationship information.

[0102] In this embodiment, the target state transition relationship information refers to the state transition content retained in the historical state transition relationship information after filtering out state transition content that does not meet the constraints according to the preset control constraint information. The target state transition relationship information is used to characterize the state change relationship that meets the conditions limited by the preset control constraint information.

[0103] Specifically, regarding historical state transition information, firstly, the changes in display counts, click counts, conversion counts, consumption values, and interaction counts, along with the direction of change for each value, are read sequentially from the historical timeline of each state transition content. Then, the corresponding state restriction content is read from the preset control constraint information, and each state transition content is mapped one-to-one with the state restriction content at the same time point. During this mapping process, it is determined whether the changes in display counts, click counts, conversion counts, consumption values, and interaction counts fall within the allowed range of change corresponding to the state restriction content. Simultaneously, it is determined whether the direction of change for each value aligns with the allowed range of change corresponding to the state restriction content. If the changes in display count, click count, conversion count, consumption value, and interaction count all fall within the allowed range, and the direction of change for each value is consistent with the allowed direction, the current state transition content is retained. If any value exceeds the allowed range or any direction of change is inconsistent with the allowed direction, the current state transition content is adjusted. Specifically, values ​​exceeding the upper limit of the allowed range are compressed to that limit, values ​​below the lower limit are padded to that limit, and values ​​with inconsistent directions are adjusted according to the allowed direction. The adjusted formula for calculating the change in display count is as follows: The formula for calculating the change in click count after adjustment is as follows: The formula for calculating the change in the number of conversions after adjustment is Δ. The formula for calculating the change in the adjusted consumption value is as follows: The formula for calculating the change in the number of interactions after adjustment is as follows: , where L Zi L Di L Ti L Xi and L Ji These represent the lower bounds of the allowed range of changes for the following values: display count, click count, conversion count, cost, and interaction count. Zi U Di U Ti U Xi and U Ji These represent the upper bounds of the allowed ranges for changes in display counts, click counts, conversion counts, consumption values, and interaction counts, respectively. After preserving or modifying the content of each state transition, the preserved and modified state transition content are aggregated in chronological order to obtain the final result.

[0104] S303: Based on the target state transition relationship information, deduce and generate the mirrored delivery state sequence information corresponding to each delivery unit.

[0105] In this embodiment, the mirrored projection state sequence information refers to the continuous state content sequence formed by gradually unfolding the target state transition relationship information in chronological order. The mirrored projection state sequence information is used to characterize the state change content unfolded by the projection unit along the time direction under the limitation of preset control constraint information. The deduction generation refers to the process of taking the state content corresponding to the previous time position as the starting point of the unfolding and combining it with the state transition content corresponding to the next time position to gradually obtain the state content corresponding to each subsequent time position.

[0106] Specifically, for the target state transition relationship information corresponding to each delivery unit, firstly, read the change values ​​of display count, click count, conversion count, consumption value, and interaction count in each state transition content according to historical time sequence. Then, read the starting state content under the corresponding time sequence in the historical delivery state information, and record the display count as Z1, click count as D1, conversion count as T1, consumption value as X1, and interaction count as J1 in the starting state content. Then, use the state content corresponding to the previous time position as the current expansion basis, and add the display count change value in the current expansion basis to the display count change value in the current state transition content to obtain the display count corresponding to the next time position. The calculation formula is as follows: The number of clicks at the current expanded base is added to the change in the number of clicks at the current state transition content to obtain the number of clicks at the next time position. The calculation formula is as follows: The number of transformations in the current expanded base is added to the change in the number of transformations in the current state transition content to obtain the number of transformations corresponding to the next time position. The calculation formula is as follows: The consumption value in the current expanded base is added to the change in consumption value in the current state transition content to obtain the consumption value corresponding to the next time position. The calculation formula is as follows: The number of interactions in the current expanded base is added to the change in the number of interactions in the current state transition content to obtain the number of interactions corresponding to the next time position. The calculation formula is as follows: ,in, , , , and These represent the number of impressions, clicks, conversions, cost, and interactions corresponding to the current time position in the mirror delivery status sequence information. , , , and These represent the number of impressions, clicks, conversions, consumption, and interactions corresponding to the next time position in the mirror delivery state sequence information. After obtaining the state content corresponding to the next time position, the state content corresponding to the next time position is used as the current expansion basis for the subsequent state transition content, and the same expansion is continued until the state expansion corresponding to all state transition content is completed. Then, the state content corresponding to each time position obtained by expanding along the time order is arranged in sequence to form the mirror delivery state sequence information corresponding to each delivery unit.

[0107] S304: Construct the image delivery status trajectory corresponding to each delivery unit based on the image delivery status sequence information.

[0108] In this embodiment, constructing the mirror projection state trajectory refers to continuously associating the state contents arranged in chronological order in the mirror projection state sequence information, so that the state contents corresponding to adjacent time positions form a sequential connection, and thereby forming trajectory content that can characterize the path of state change.

[0109] Specifically, for the mirrored delivery state sequence information corresponding to each delivery unit, the state content corresponding to each time position is first read in chronological order, and the state content corresponding to the previous time position is sequentially connected with the state content corresponding to the next time position. During the connection process, the number of impressions, clicks, conversions, consumption, and interactions corresponding to the previous time position, as well as the number of impressions, clicks, conversions, consumption, and interactions corresponding to the next time position, are extracted respectively. The state content corresponding to the previous time position is taken as the preceding state content, and the state content corresponding to the next time position is taken as the following state content. Then, the corresponding trajectory connection content is recorded around the sequential relationship between the preceding and following state content, so that a corresponding connection relationship is formed between each group of adjacent time positions. After completing the connection between all adjacent time positions, the connection relationships are sequentially connected in chronological order, so that the first state content, each intermediate state content, and the last state content in the mirrored delivery state sequence information form a continuous state change path, and the state change path is organized into the mirrored delivery state trajectory corresponding to each delivery unit.

[0110] In one embodiment, step S303, namely, based on the target state transition relationship information, deducing and generating the mirrored delivery state sequence information corresponding to each delivery unit, includes:

[0111] S3031: Determine the initial mirror state node information corresponding to each deployment unit based on the target state migration relationship information.

[0112] In this embodiment, the initial mirror state node information refers to the state content used as the starting point for unfolding the mirror delivery state sequence information at the starting position of the time sequence corresponding to the target state migration relationship information. The state node information refers to the node content formed by combining the number of displays, clicks, conversions, consumption values ​​and interaction counts at a single time position.

[0113] Specifically, for the target state transition relationship information corresponding to each delivery unit, the state transition content is first read in chronological order, and the first state transition content at the starting position is located. Then, the preceding state content corresponding to the starting position is extracted from the first state transition content. The number of impressions, clicks, conversions, consumption, and interactions in the preceding state content are read out respectively, and it is checked whether the number of impressions, clicks, conversions, consumption, and interactions completely correspond at the same time position. If the number of impressions, clicks, conversions, consumption, and interactions all exist, the number of impressions, clicks, conversions, consumption, and interactions are combined according to the same time position to form the state node content corresponding to the starting position. When there are multiple delivery units, the same reading, extraction, checking, and combination actions are performed on the first state transition content corresponding to each delivery unit to obtain the initial mirror state node information corresponding to each delivery unit.

[0114] S3032: Based on the initial mirror state node information and the target state transition relationship information, perform forward recursion processing to obtain the forward mirror state chain information corresponding to each deployment unit.

[0115] In this embodiment, forward recursion processing refers to the process of taking the initial mirror state node information as the starting node content and gradually unfolding the content of each node along the time sequence corresponding to the target state migration relationship information. Forward mirror state chain information refers to the chain-like state content composed of multiple mirror state node information connected in chronological order during the forward recursion processing.

[0116] Specifically, for the initial mirror state node information and target state transition relationship information corresponding to each deployment unit, the state node content corresponding to the initial mirror state node information is first determined as the forward recursion starting point. Then, each state transition content in the target state transition relationship information is read sequentially according to time. After reading the current state transition content, the number of impressions at the current forward recursion starting point is added to the change in the number of impressions in the current state transition content to obtain the number of impressions corresponding to the next node. The calculation formula is as follows: The click count for the next node is obtained by adding the click count change in the current forward recursion starting point to the click count change in the current state transition content. The calculation formula is as follows: The number of transformations at the current forward recursion starting point is added to the change in the number of transformations in the current state transition content to obtain the number of transformations for the next node. The calculation formula is as follows: The consumption value at the current forward recursion starting point is added to the change in consumption value in the current state transition content to obtain the consumption value for the next node. The calculation formula is as follows: The interaction count for the next node is obtained by adding the interaction count change in the current state transition content to the number of interactions at the current forward recursion starting point. The calculation formula is as follows: ,in, , , , and These represent the number of impressions, clicks, conversions, costs, and interactions corresponding to the current forward recursion starting point. , , , and These represent the number of impressions, clicks, conversions, costs, and interactions corresponding to the next node. After obtaining these data, the impressions, clicks, conversions, costs, and interactions are combined to form the mirror state node information for the next node. The current forward recursion starting point is then connected to the next node in chronological order to form the current chain segment content. The next node is then updated to the forward recursion starting point corresponding to the subsequent state transition content, and the same recursion operation continues until all state transition content has completed the corresponding node expansion and sequential connection. Finally, all node content and all chain segment content obtained by connecting in chronological order are aggregated into the forward mirror state chain information corresponding to each delivery unit.

[0117] S3033: Perform hierarchical partitioning on each state node in the forward mirror state chain information to obtain state hierarchy result information.

[0118] In this embodiment, the state hierarchy result information refers to the result content formed after the state nodes in the forward mirror state chain information are sorted and hierarchically according to the state change relationship. The hierarchical division process refers to the action of grouping and classifying each state node according to the connection position, change range and continuity relationship of each state node in the forward mirror state chain information.

[0119] Specifically, for each state node in the forward mirror state chain information, firstly, the number of impressions, clicks, conversions, consumption, and interactions corresponding to each state node are read in chronological order. Then, the current state node is compared with the previous state node, and the differences in impressions, clicks, conversions, consumption, and interactions between the current and previous state nodes are calculated. Finally, the absolute values ​​of these differences are summarized to obtain the state change amount corresponding to the current state node. The calculation formula is as follows: Where Bi represents the state change amount corresponding to the current state node. , , , and These represent the differences in display counts, click counts, conversion counts, consumption values, and interaction counts between the current state node and the previous state node, respectively. After obtaining the state change amount corresponding to each state node, the connection relationship between adjacent state nodes is read in chronological order. If the state change amounts corresponding to adjacent state nodes are small and the connection order remains continuous, the corresponding state nodes are classified into the same level. If the state change amounts corresponding to adjacent state nodes increase significantly or the connection order is broken, the next state node is classified into the next level. After completing the comparison and classification of all state nodes one by one, the level position corresponding to each state node and the connection relationship between each state node are associated and organized to obtain the state level result information.

[0120] S3034: Based on the state hierarchy result information, freeze the state nodes that meet the preset stability conditions and perform reverse backtracking verification on the state nodes that do not meet the preset stability conditions to obtain the state adjustment result information.

[0121] In this embodiment, the preset stability condition refers to the pre-set condition used to determine whether the state node maintains a stable change relationship. The freeze process refers to the action of keeping the current state content and current connection relationship of the state node that meets the preset stability condition unchanged. The reverse backtracking verification process refers to the action of reading the contents of the preceding state node in reverse time for the state node that does not meet the preset stability condition and performing corresponding comparisons. The state adjustment result information refers to the state node adjustment result content formed after the freeze process and the reverse backtracking verification process are completed.

[0122] Specifically, regarding the state hierarchy result information, the hierarchical position and connection relationships of each state node are read in hierarchical order. Each state node is then checked to ensure continuous connection within the current level, consistent change direction with the previous and next state nodes, and absence of interruptions between adjacent levels. If continuous connection remains unchanged, change direction is consistent, and no interruptions occur, the current state node is deemed to meet the preset stability conditions. The number of impressions, clicks, conversions, consumption, interactions, and current connection relationships corresponding to the current state node are directly retained, and the current state node is frozen. If continuous connection is interrupted, change direction is inconsistent, or interruptions occur between adjacent levels, the current state node is deemed not to meet the preset stability conditions. Then, the preceding states are read sequentially backwards from the current state node. The system compares the number of impressions, clicks, conversions, consumption, and interactions of a node with its adjacent preceding state nodes, identifying the differences between them. It then continues reading preceding state nodes from even earlier times until a backtracking position is found that maintains a continuous connection with the current state node. Based on the state content corresponding to the backtracking position, the number of impressions, clicks, conversions, consumption, and interactions in the current state node are adjusted. The connection relationships between the current state node and its preceding and following state nodes are also reorganized. After completing the freezing action for each state node that meets the preset stability conditions, and the reverse reading, item-by-item comparison, backtracking location, state adjustment, and connection reorganization for each state node that does not meet the preset stability conditions, the freezing results and verification adjustment results for each state node are compiled into state adjustment result information.

[0123] S3035: Based on the state adjustment result information, perform iterative chain completion processing on the corresponding unfrozen state nodes to obtain the mirrored delivery state sequence information corresponding to each delivery unit.

[0124] In this embodiment, iterative chain completion processing refers to repeatedly performing node insertion and connection completion in chronological order around the state nodes that are not yet in a frozen state in the state trimming result information, so as to form a continuous connection relationship between the unfrozen state nodes and the preceding and following state nodes.

[0125] Specifically, for each unfrozen state node in the state adjustment result information, the preceding and succeeding state nodes corresponding to each unfrozen state node are first read in chronological order. Then, it is checked whether there are connection gaps between each unfrozen state node and its preceding and succeeding state nodes. If a connection gap exists, the number of impressions, clicks, conversions, consumption, and interactions corresponding to both ends of the gap are read. The intermediate state content to be added is determined based on the state content at both ends of the gap. Finally, the added intermediate state content is inserted into the corresponding gap position, ensuring that the preceding state node, intermediate state content, unfrozen state node, and succeeding state node move along the timeline. The sequence is re-established to form continuous connections. After completing one insertion, it continues to check whether there are still new connection gaps between the inserted intermediate state content and adjacent state nodes. If there are still new connection gaps, the same reading, confirming, inserting, and connecting actions are performed around the new connection gaps. If there are no new connection gaps, the connection results corresponding to the current unfrozen state node are retained, and the same chain-filling action is performed on the next unfrozen state node. After all unfrozen state nodes have completed the corresponding connection gap checks and insertion connections, the frozen state nodes, the unfrozen state nodes after chain filling, and the connection relationships between each node arranged in chronological order are collected to obtain the mirrored delivery state sequence information corresponding to each delivery unit.

[0126] In one embodiment, step S40, namely, determining the deviation correction result for each delivery unit based on the trajectory deviation result between each actual delivery state trajectory and the corresponding mirrored delivery state trajectory, includes:

[0127] S401: Based on the trajectory deviation results between the actual delivery status trajectory and the corresponding mirrored delivery status trajectory, extract the deviation status segment information corresponding to each delivery unit.

[0128] In this embodiment, the deviation state segment information refers to the information formed around the state segments where there is deviation content at continuous time positions after the actual delivery state trajectory and the mirror delivery state trajectory are compared. The deviation state segment information is used to characterize the location range and continuous distribution of the deviation content.

[0129] Specifically, regarding the trajectory deviation between the actual delivery status trajectory and the corresponding mirrored delivery status trajectory, the deviation content corresponding to each time position is first read in chronological order, and it is checked whether the deviation content corresponding to each time position appears continuously in adjacent time positions. If the deviation content corresponding to the current time position is continuous with the deviation content corresponding to the previous time position, the current time position is merged into the deviation segment of the previous time position. If the deviation content corresponding to the current time position is not continuous with the deviation content corresponding to the previous time position, the current time position is taken as the starting position of the new deviation segment, and the deviation content corresponding to the subsequent time positions is read one by one in chronological order. The starting time position and ending time position are recorded around the continuous time positions in each deviation segment. Then, the deviation content of the number of impressions, clicks, conversions, consumption, and interactions in each deviation segment are collected to maintain the correspondence of each deviation content in the same continuous time segment. After completing the division and collection of all continuous time segments, the position range corresponding to each continuous time segment and the collected deviation content are organized into the deviation status segment information corresponding to each delivery unit.

[0130] S402: Based on the deviation state segment information, determine the corresponding mirror reference node information and actual drift node information.

[0131] In this embodiment, the mirror reference node information refers to the state node content extracted from the mirror delivery state trajectory within the time position range corresponding to the deviation state segment information, which is used as the benchmark for comparison of deviation content. The actual drift node information refers to the state node content extracted from the actual delivery state trajectory within the time position range corresponding to the deviation state segment information, which deviates from the mirror reference node information.

[0132] Specifically, for the deviation state segment information, firstly, the start time position and end time position corresponding to each deviation state segment are read. Then, the corresponding state node content is located one by one in the mirrored delivery state trajectory around the start time position to the end time position. The located state node content is organized in chronological order and used as the mirror reference node information for the corresponding deviation state segment. At the same time, the corresponding state node content is located one by one in the actual delivery state trajectory around the start time position to the end time position. The located state node content is organized in chronological order and used as the actual drift node information for the corresponding deviation state segment. During the organization process, the one-to-one correspondence between the mirror reference node information and the actual drift node information at the same time position is maintained, so that each mirror reference node information corresponds to the actual drift node information at the same time position. After completing the location, extraction and corresponding organization of the node content corresponding to each deviation state segment, the corresponding mirror reference node information and actual drift node information are obtained.

[0133] S403: Based on the corresponding deviation relationship between the mirror reference node information and the actual drift node information, generate the reverse traction path information corresponding to the deviation state segment.

[0134] In this embodiment, the corresponding deviation relationship refers to the difference between the mirror reference node information and the actual drift node information at the same time position, and the reverse traction path information refers to the node backtracking path content formed when the actual drift node information gradually approaches the mirror reference node information based on the corresponding deviation relationship.

[0135] Specifically, regarding the mirror reference node information and the actual drift node information, the content of the mirror reference node and the actual drift node at the same time position are read sequentially according to time. The number of impressions, clicks, conversions, consumption values, and interactions in the mirror reference node content are compared with the number of impressions, clicks, conversions, consumption values, and interactions in the actual drift node content to determine the deviation direction and deviation position at each time position. Then, starting from the end time position of the deviation state segment, the content of the actual drift node and the mirror reference node corresponding to the previous time position are read sequentially in reverse time. During each reverse reading, the deviation direction corresponding to the next time position is extended to the previous time position. It is determined whether the previous time position continues to maintain the same deviation direction. If it continues to maintain the same deviation direction, the previous time position is included in the same reverse traction segment. If it does not continue to maintain the same deviation direction, the previous time position is used as the starting point of a new reverse traction segment. Then, the node positions and deviation directions arranged in reverse time within each reverse traction segment are sequentially connected to form the reverse traction path information of the corresponding deviation state segment.

[0136] S404: Based on the reverse traction path information, perform step-by-step pullback processing on the actual drift node information to obtain the pullback correction state chain information.

[0137] In this embodiment, the step-by-step pullback process refers to the action of gradually moving the actual drift node information from the deviation position to the mirror reference node information according to the node order determined in the reverse traction path information. The pullback correction state chain information refers to the corrected state chain content that is continuously connected in time sequence after the step-by-step pullback process is completed.

[0138] Specifically, for each reverse traction path information and each actual drift node information, the positions of each node arranged in reverse order in the reverse traction path information are first read. Starting from the actual drift node content corresponding to the end of the deviation state segment, the number of impressions, clicks, conversions, consumption values, and interactions in the actual drift node content at the end are aligned with the corresponding mirror reference node content item by item. After the alignment of the end node is completed, the position of the previous node in the reverse traction path information is read, and the actual drift node content corresponding to the previous node position is adjusted in order according to the correction results already completed by the next node, so that the state content corresponding to the previous node position and the corrected state content corresponding to the next node position are connected. Then, the same node reading, content alignment, and connection arrangement actions are performed one by one along the reverse traction path information to the previous position, so that the actual drift node content is gradually pulled back in the reverse order in the reverse traction path information. After all node positions have been pulled back, the corrected node content is reconnected in the forward time direction to obtain the pullback correction state chain information.

[0139] S405: Generate deviation correction results for each delivery unit based on the pullback correction state chain information.

[0140] In this embodiment, the pullback correction state chain information refers to the corrected state chain content formed after step-by-step pullback processing, and the deviation correction result refers to the result content formed based on the pullback correction state chain information, which is used to characterize the correction of deviation content.

[0141] Specifically, for the pullback correction state chain information, firstly, the content of each corrected node and the connection relationships between the corrected node contents are read in chronological order. Then, around the start and end time positions corresponding to the deviation state segments, the content of each corrected node is matched one by one with the original deviation content to determine the number of impressions, clicks, conversions, consumption values, and interactions that have been corrected at each time position. The connection relationships that have been corrected between adjacent time positions are also included in the corresponding correction segment. After completing the collection of content for a single correction segment, the same reading, matching, and collection actions are performed on the pullback correction state chain information corresponding to other deviation state segments. Finally, the corrected node content and connection relationships collected in each correction segment are gathered according to the delivery unit to form the deviation correction result corresponding to each delivery unit.

[0142] In one embodiment, step S50, namely, generating candidate delivery control information for each delivery unit based on the deviation correction result, and determining the state impact result corresponding to each candidate delivery control information, includes:

[0143] S501: Based on the deviation correction results, determine the corresponding correction traction node information.

[0144] In this embodiment, the corrected traction node information refers to the state node content that plays a guiding role in the generation of subsequent candidate deployment control path information in the deviation correction result. The corrected traction node information is used to characterize the node position and node content that need to be used as the control starting point or control connection base point in the deviation correction result.

[0145] Specifically, regarding the deviation correction results, the corrected node content and connection relationships of each delivery unit are first read in chronological order. Then, the changes in node content are checked one by one around the start time, end time, and intermediate time positions corresponding to each deviation state segment. If there is a significant adjustment between the content of a node and the content of the previous node, and the content of the next node is continuously connected, the current node is determined as a candidate traction node. If the content of a node is at the boundary of the correction segment and is directly adjacent to the uncorrected segment, the current node is determined as a boundary traction node. Then, the candidate traction nodes and boundary traction nodes are organized in chronological order, and the display count, click count, conversion count, consumption value, interaction count, and node position of each node are collected to form the corresponding correction traction node information.

[0146] S502: Generate corresponding candidate delivery control path information based on the corrected traction node information.

[0147] In this embodiment, the candidate delivery control path information refers to the control extension path content formed by organizing the corrected traction node information in chronological order. The candidate delivery control path information is used to characterize the node connection range and node arrangement order that the control content passes through when it expands forward or backward from the corrected traction node information.

[0148] Specifically, for the information of the correction traction nodes, the node position and node content corresponding to each correction traction node are first read. Taking each correction traction node as the starting position of the path, the content of the preceding node directly adjacent to the current correction traction node is read forward in time sequence, and the content of the following node directly adjacent to the current correction traction node is read backward in time sequence. If the content of the preceding node is continuously connected to the current correction traction node, the content of the preceding node is merged into the same path range. If the content of the following node is continuously connected to the current correction traction node, the content of the following node is merged into the same path range. Then, a candidate path is formed around the preceding node content, the current node content, and the following node content corresponding to the current correction traction node. The candidate paths corresponding to different correction traction nodes are sorted separately so that each candidate path maintains an independent time sequence and node connection relationship, thereby generating the corresponding candidate delivery control path information.

[0149] S503: Based on the control path information of each candidate delivery, perform impact layer expansion processing on the corresponding state nodes to obtain the initial impact layer information, the transition impact layer information, and the final impact layer information.

[0150] In this embodiment, the impact layering and unfolding process refers to the action of dividing the state nodes in the candidate delivery control path information into layers according to the order and diffusion position of state changes in the path. The initial impact layer information refers to the content of the first layer of state nodes directly corresponding to the correction traction node. The transition impact layer information refers to the content of the intermediate state nodes located between the initial impact layer information and the final impact layer information. The final impact layer information refers to the content of the state nodes located at the boundary position of the candidate delivery control path information.

[0151] Specifically, for each candidate delivery control path information, all status nodes in the path are first read, and the status node at the same position as the correction traction node is determined as the starting position node. Then, the status node directly adjacent to the starting position node and located in the middle of the path is determined as the intermediate position node. The status nodes located at both ends of the candidate delivery control path information are determined as the boundary position nodes. After completing the node position division, the node content corresponding to the starting position node is organized into the starting impact layer information, the node content corresponding to the intermediate position node is organized into the transition impact layer information in chronological order, and the node content corresponding to the boundary position node is organized into the end impact layer information according to the positions at both ends of the path. This allows each status node in the same path to fall into the corresponding level, thereby obtaining the starting impact layer information, the transition impact layer information, and the end impact layer information.

[0152] S504: Identify the impact isolation layer information in each candidate delivery control path information based on the initial impact layer information, the transition impact layer information, and the terminal impact layer information.

[0153] In this embodiment, the impact isolation layer information refers to the interlayer node content that serves to separate state changes between the initial impact layer information, the transitional impact layer information, and the final impact layer information. The impact isolation layer information is used to characterize the location content in the path that can block or slow down the continuous diffusion of state changes.

[0154] Specifically, for the initial impact layer information, transitional impact layer information, and final impact layer information, the state node content in each layer is first read in the order of the path, and the node changes between adjacent layers are compared layer by layer. When the node content in the initial impact layer information extends to the node content in the transitional impact layer information, it is checked whether the changes between adjacent nodes slow down. When the node content in the transitional impact layer information extends to the node content in the final impact layer information, it is checked whether the changes between adjacent nodes continue to slow down. If the node content corresponding to a certain layer changes and shrinks relative to the node content of the previous layer and continues to shrink relative to the node content of the next layer, the current layer is determined as the impact isolation position. Then, the node content in the impact isolation position and the corresponding layer position are collected to identify the impact isolation layer information in each candidate delivery control path information.

[0155] S505: Based on the impact isolation layer information, perform inter-layer peak shaving processing on the information of each candidate delivery control path to obtain the state impact result corresponding to each candidate delivery control information.

[0156] In this embodiment, interlayer peak shaving refers to the action of compressing and consolidating the state change content between adjacent layers in the candidate delivery control path information around the impact isolation layer information. The state impact result refers to the result content formed after the state fluctuation content caused by the candidate delivery control information is sorted out after the interlayer peak shaving is completed.

[0157] Specifically, regarding the impact isolation layer information and the control path information for each candidate delivery, firstly, the hierarchical position and node content corresponding to the impact isolation layer information are read. Then, the change range of the node content in the adjacent layers on both sides of the impact isolation layer information is checked one by one. If there is a significant change between the node content in the layer preceding the impact isolation layer information and the node content corresponding to the impact isolation layer information, the node content in the layer preceding the impact isolation layer is shrunk and reorganized, so that the display count, click count, conversion count, consumption value, and interaction count content in the layer preceding the impact isolation layer are closer to the node content corresponding to the impact isolation layer information. If there is a significant change between the node content in the layer following the impact isolation layer information and the node content corresponding to the impact isolation layer information, then... The node content in the next layer is shrunk and reorganized to bring the display count, click count, conversion count, consumption value, and interaction count content in the next layer closer to the node content corresponding to the impact isolation layer information. After completing the shrunk and reorganization of the adjacent layers on both sides of the impact isolation layer information, the other layers are checked sequentially along the path to see if there are still significant changes. If there are still significant changes, the same shrunk and reorganization action is performed around the corresponding impact isolation layer information. After completing the inter-layer peak shaving for all candidate delivery control path information, the node changes, inter-layer connection changes, and path boundary changes before and after peak shaving for each path are sorted out to form the state impact result corresponding to each candidate delivery control information.

[0158] In one embodiment, in step S60, based on the impact results of each state, constraint correction processing is performed on each candidate delivery control information to obtain target delivery control information, including:

[0159] S601: Based on the state impact results corresponding to the candidate delivery control information, divide the positive impact constraint interval and the reverse pullback constraint interval.

[0160] In this embodiment, the forward impact constraint interval refers to the continuous time position range that needs to be restricted when expanding backward along the time sequence corresponding to the candidate delivery control information in the state impact result, and the reverse pullback constraint interval refers to the continuous time position range that needs to be pulled back when tracing backward along the time sequence corresponding to the candidate delivery control information in the state impact result.

[0161] Specifically, regarding the state impact results corresponding to the candidate delivery control information, the state fluctuation content corresponding to each time position is first read in chronological order. Then, the state fluctuation content at each subsequent time position is checked one by one around the adjustment position in the candidate delivery control information to see if it continues to extend. If the state fluctuation content at the subsequent time position is continuous with the state fluctuation content at the current time position, the current time position and the subsequent time position are included in the same positive extension range. If the state fluctuation content at the subsequent time position is not continuous with the state fluctuation content at the current time position, the current time position is taken as the termination position of the positive extension range. Then, the state fluctuation content at each subsequent time position is checked one by one around the adjustment position in the candidate delivery control information. Check whether the state fluctuation content at the previous time position is continuously traced back. If the state fluctuation content at the previous time position is continuous with the state fluctuation content at the current time position, the current time position and the previous time position are included in the same reverse traceback range. If the state fluctuation content at the previous time position is not continuous with the state fluctuation content at the current time position, the current time position is used as the starting position of the reverse traceback range. After completing the forward extension check at the next time position and the reverse traceback check at the previous time position, the range of time positions that are continuously extended backward is organized into the forward impact constraint range, and the range of time positions that are continuously traced forward is organized into the reverse pullback constraint range.

[0162] S602: Based on the positive impact constraint interval, the candidate delivery control information is forward-constrained and folded to form positive constraint control chain information.

[0163] In this embodiment, forward constraint folding refers to the action of consolidating and merging the control content of candidate delivery control information in the backward direction of time sequence around the positive impact constraint interval. The positive constraint control chain information refers to the control chain content formed by connecting the forward constraint information in the backward direction of time sequence after the forward constraint folding is completed.

[0164] Specifically, for the positive impact constraint interval and candidate delivery control information, the start and end time positions corresponding to the positive impact constraint interval are first read, and the adjustment positions, adjustment directions, and adjustment magnitudes of each candidate delivery control information falling within the positive impact constraint interval are located. When there are multiple adjustment positions within the same positive impact constraint interval, the control contents corresponding to multiple adjustment positions are sequentially arranged in chronological order. Then, it is checked whether the adjustment directions corresponding to adjacent adjustment positions are consistent. If the adjustment directions are consistent, the control contents corresponding to adjacent adjustment positions are merged and organized so that multiple adjacent adjustment positions together form a continuous control content. If the adjustment directions are inconsistent, the control content corresponding to the previous adjustment position is retained as the previous control content, and the control content corresponding to the next adjustment position is retained as the next control content. The connection range between the previous control content and the next control content is shortened so that the previous control content and the next control content are arranged sequentially within the positive impact constraint interval. After completing the merging, sequential arrangement, and connection range reduction of each control content within the positive impact constraint interval, the continuous control content formed sequentially along the time line is organized into positive constraint control chain information.

[0165] S603: Based on the reverse pullback constraint interval, the positive constraint control chain information is reverse compensated and folded to form bidirectional folded control chain information.

[0166] In this embodiment, reverse compensation folding refers to the action of compensating and consolidating the control content of the forward constraint control chain information in the forward direction of time sequence around the reverse pull-back constraint interval. Bidirectional folded control chain information refers to the bidirectional consolidated control chain content formed after simultaneously undergoing forward constraint folding and reverse compensation folding.

[0167] Specifically, for the reverse pullback constraint interval and the forward constraint control chain information, first read the start and end time positions corresponding to the reverse pullback constraint interval, and locate each control content in the forward constraint control chain information that intersects with the reverse pullback constraint interval. Then, read the adjustment position, adjustment direction, and adjustment magnitude corresponding to each control content in reverse time order. After the control content corresponding to the later time position has been settled, check whether the control content corresponding to the earlier time position maintains continuity with the control content corresponding to the later time position. If continuity is maintained, adjust the control content corresponding to the earlier time position... The control content corresponding to a later time position is moved closer and organized, so that the control content corresponding to an earlier time position and the control content corresponding to a later time position form a pullback connection in reverse time. If the continuity cannot be maintained, the control content corresponding to an earlier time position is contracted and compressed in amplitude, so that the control content corresponding to an earlier time position avoids the interruption position and moves back to the control content that has been folded. After completing the reverse reading, convergence, position contraction and amplitude compression of each control content within the reverse pullback constraint interval, the control content that simultaneously satisfies the positive folding and reverse back-movement relationship is organized into bidirectional folded control chain information.

[0168] S604: Determine the stable output range information based on the bidirectional folding control chain information.

[0169] In this embodiment, stable output interval information refers to the time position range information in the bidirectional folding control chain information where the control content remains continuously connected and the order of changes is stable.

[0170] Specifically, for the bidirectional folding control chain information, the adjustment position, adjustment direction, and adjustment magnitude corresponding to each control content are first read in chronological order. Then, it is checked whether adjacent control contents are continuously connected. If there is no disconnection between adjacent control contents and the adjustment direction maintains the same change order, the corresponding time position range is assigned to the same continuous interval. Then, it is checked whether the adjustment magnitudes corresponding to each control content in the same continuous interval are connected. If there is no sudden interruption in the adjustment magnitude along the time sequence, the current time position is kept in the same continuous interval. If there is a significant jump in the adjustment magnitude along the time sequence, the current time position is taken as the starting position of the new continuous interval. After completing the sequential check of all control contents, the continuous time position range that simultaneously satisfies continuous connection and stable change order is organized into stable output interval information.

[0171] S605: Based on the stable output range information, lock the corresponding bidirectional folding control chain information to obtain the target delivery control information for each delivery unit.

[0172] In this embodiment, locking refers to the action of fixing and retaining the control content in the bidirectional folded control chain information that falls within the stable output interval information, around the stable output interval information.

[0173] Specifically, for stable output interval information and bidirectional folded control chain information, the start and end time positions corresponding to each stable output interval information are first read. Then, the corresponding control content is located segment by segment in the bidirectional folded control chain information around each start and end time position. The adjustment position, adjustment direction, and adjustment range within the stable output interval information are completely preserved, while the control content outside the stable output interval information is stripped from the current bidirectional folded control chain information. In the case of multiple stable output interval information, the same positioning and preservation actions are performed on the control content corresponding to each stable output interval information. Then, the preserved control content within each stable output interval information is connected sequentially according to the time order, so that each preserved control content forms the final fixed control chain content. After the preservation and connection of the control content corresponding to all stable output interval information are completed, the final fixed control chain content is organized into the target delivery control information corresponding to each delivery unit.

[0174] In one embodiment, step S70, namely, controlling the corresponding advertising delivery process based on the target delivery control information, includes:

[0175] S701: Generate the corresponding master control execution chain information and shadow control takeover chain information based on the target delivery control information corresponding to the delivery unit.

[0176] In this embodiment, the master control execution chain information refers to the actual execution control chain content organized in chronological order based on the adjustment position, adjustment direction and adjustment range determined in the target deployment control information. The shadow control chain information refers to the tracking control chain content organized in chronological order and node position corresponding to the master control execution chain information, around the same target deployment control information.

[0177] Specifically, for the target delivery control information corresponding to the delivery unit, the adjustment direction and adjustment range corresponding to each adjustment position are first read in chronological order, and the adjustment positions are arranged in chronological order. Then, the arranged adjustment positions, adjustment directions, and adjustment ranges are directly connected to form the first control chain content corresponding to the target delivery control information. The first control chain content is determined as the main control execution chain information. At the same time, the corresponding adjustment direction and adjustment range are copied one by one around each adjustment position in the main control execution chain information, so that the copied adjustment positions, adjustment directions, and adjustment ranges maintain the same chronological order and the same node position as the main control execution chain information. Then, the copied adjustment positions, adjustment directions, and adjustment ranges are separately organized into another control chain content, and the other control chain content is determined as the shadow control chain information. After completing the corresponding organization of the main control execution chain information and the shadow control chain information, the main control execution chain information and the shadow control chain information corresponding to the delivery unit are obtained.

[0178] S702: Based on the master control execution chain information, control is executed on the corresponding advertising delivery process, and based on the shadow control takeover chain information, the reverse approximation is performed on the corresponding mirror delivery state trajectory to obtain the control takeover approximation result information.

[0179] In this embodiment, reverse approximation refers to the action of gradually comparing and narrowing the difference positions by reversing the relationship of each node position in the shadow control chain information along the mirrored delivery state trajectory. Control approximation result information refers to the corresponding approximation result content formed after the main control execution chain information acts on the advertising delivery process and the shadow control chain information acts on the mirrored delivery state trajectory.

[0180] Specifically, for the main control execution chain information, the adjustment positions, adjustment directions, and adjustment ranges are read in chronological order. When the ad delivery process reaches the corresponding adjustment position, the adjustment direction and adjustment range are mapped to the number of impressions, clicks, conversions, consumption, and interactions at the current time position. The control actions corresponding to each adjustment position are completed sequentially. At the same time, for the shadow control seizure chain information, the position, adjustment direction, and adjustment range of each node are read in the same chronological order as the main control execution chain information. Starting from the state content corresponding to each node position in the mirrored delivery state trajectory, the state content corresponding to the previous time position is read one by one in reverse chronological order. During each reverse reading, the adjustment direction and adjustment range corresponding to the current node position are matched with the state content obtained in reverse reading, so that the state content obtained in reverse reading gradually approaches the node content in the shadow control seizure chain information. After completing the reverse reading and gradual approach of all node positions, the execution content of the main control execution chain information during the ad delivery process and the reverse approach content of the shadow control seizure chain information in the mirrored delivery state trajectory are matched and aggregated to form the control seizure approach result information.

[0181] S703: Based on the control and approach result information, determine the corresponding trajectory matching and overtaking result information.

[0182] In this embodiment, the trajectory alignment and overtaking result information refers to the overtaking relationship content formed by comparing the execution result corresponding to the main control execution chain information with the approximation result corresponding to the shadow control seizure chain information, based on the control seizure and approximation result information.

[0183] Specifically, regarding the control and approximation result information, the main control execution content and shadow approximation content corresponding to each node position are first read in chronological order. Then, the correspondence between the main control execution content and the shadow approximation content in terms of display count, click count, conversion count, consumption value, and interaction count is compared one by one around the same node position. If the shadow approximation content is closer to the state content corresponding to the mirrored delivery state trajectory than the main control execution content, the current node position is determined as the overtaking node position. If the shadow approximation content does not exceed the corresponding fit of the main control execution content, the current node position is determined as the non-overtaking node position. Then, the overtaking node positions are checked in chronological order to see if there are consecutive overtaking node positions. If multiple consecutive node positions are overtaking node positions, the multiple consecutive node positions are merged into the same overtaking segment. If no consecutive overtaking node positions appear, a single node position is retained as an independent overtaking position. Finally, the time position ranges corresponding to each overtaking node position, non-overtaking node position, and each overtaking segment are organized to obtain the corresponding trajectory fitting overtaking result information.

[0184] S704: If the trajectory matching and overtaking result information meets the preset control conditions, replace the main control execution chain information with the shadow control chain information, and if the trajectory matching and overtaking result information does not meet the preset control conditions, maintain the main control execution chain information.

[0185] In this embodiment, the preset control takeover condition refers to the pre-set conditions used to determine whether the shadow control takeover chain information is qualified to replace the master control execution chain information.

[0186] Specifically, for the trajectory-fitting overtaking result information, firstly, the corresponding overtaking node positions, non-overtaking node positions, and overtaking segments are read. Then, the continuous length of each overtaking segment is checked, and the distribution density of each overtaking node position in the time sequence is checked. If the continuous length of the overtaking segment reaches the preset length and the overtaking node positions remain continuously distributed within the preset time range, the trajectory-fitting overtaking result information is determined to meet the preset control conditions. Then, the positions, adjustment directions, and adjustment ranges of each node in the shadow control chain information are read. The node content corresponding to the same time position in the main control execution chain information is replaced one by one with the node content in the shadow control chain information. If the continuous length of the overtaking segment does not reach the preset length or the overtaking node positions do not remain continuously distributed within the preset time range, the trajectory-fitting overtaking result information does not meet the preset control conditions. Then, the positions, adjustment directions, and adjustment ranges of each node in the main control execution chain information remain unchanged, thereby completing the judgment and execution of replacing the main control execution chain information with the shadow control chain information or maintaining the main control execution chain information.

[0187] S705: Based on the replaced execution chain information or the retained execution chain information, complete the control of the corresponding advertising delivery process.

[0188] In this embodiment, the replaced execution chain information refers to the execution chain content formed after replacing the master execution chain information with the shadow takeover chain information, and the retained execution chain information refers to the master execution chain information that is retained even if no replacement occurs.

[0189] Specifically, for the replaced or retained execution chain information, the adjustment direction and adjustment range corresponding to each node position are first read in chronological order, and the node positions are arranged sequentially according to time. When the ad delivery process reaches the time position corresponding to each node position, the adjustment direction and adjustment range corresponding to the current node position are mapped to the number of impressions, clicks, conversions, consumption, and interactions in the current state content. This ensures that the number of impressions, clicks, conversions, consumption, and interactions are adjusted item by item according to the corresponding node order in the execution chain information. In the case of multiple consecutive node positions, the adjustment content corresponding to the previous node position and the adjustment content corresponding to the next node position are executed sequentially, so that the state content after execution is continuously updated along the time order. After the adjustment content corresponding to all node positions is completed, the state content after execution is written into the corresponding ad delivery process record in chronological order, thereby completing the control of the corresponding ad delivery process.

[0190] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0191] In one embodiment, an advertising delivery control device is provided, which corresponds one-to-one with the advertising delivery control method described in the above embodiments. For example... Figure 2 As shown, the advertising delivery control device includes a feedback acquisition module, a state construction module, a mirror trajectory module, a deviation correction module, a candidate control module, a target control module, and an execution control module.

[0192] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0193] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An advertisement distribution control method characterized by comprising: The advertising delivery control method includes: Obtain real-time feedback data for each ad placement unit during the ad delivery process; Based on the real-time feedback data, construct the actual delivery status trajectory corresponding to each delivery unit; Based on the historical delivery status information and preset control constraint information corresponding to each delivery unit, a mirror delivery status trajectory corresponding to each delivery unit is generated. Based on the trajectory deviation results between each actual delivery state trajectory and the corresponding mirrored delivery state trajectory, the deviation correction result corresponding to each delivery unit is determined; Based on the deviation correction results, candidate delivery control information corresponding to each delivery unit is generated, and the state impact result corresponding to each candidate delivery control information is determined. Based on the state impact results, constraint correction processing is performed on the candidate delivery control information to obtain the target delivery control information; Based on the target delivery control information, control is exercised over the corresponding advertising delivery process.

2. The advertisement delivery control method of claim 1, wherein The step of generating a mirrored delivery state trajectory for each delivery unit based on its historical delivery status information and preset control constraint information includes: Based on the historical delivery status information corresponding to each delivery unit, determine the corresponding historical status migration relationship information; Based on the preset control constraint information, constraint filtering processing is performed on the historical state transition relationship information to obtain the target state transition relationship information; Based on the target state transition relationship information, the mirrored delivery state sequence information corresponding to each delivery unit is deduced and generated. Based on the image delivery state sequence information, construct the image delivery state trajectory corresponding to each delivery unit.

3. The advertising placement control method according to claim 2, characterized in that, The step of generating mirrored delivery state sequence information corresponding to each delivery unit based on the target state transition relationship information includes: Based on the target state migration relationship information, determine the initial mirror state node information corresponding to each of the deployment units; Based on the initial mirror state node information and the target state transition relationship information, a forward recursive process is performed to obtain the forward mirror state chain information corresponding to each deployment unit; Perform hierarchical partitioning on each state node in the forward mirror state chain information to obtain state hierarchy result information; Based on the state hierarchy result information, freeze processing is performed on state nodes that meet the preset stability conditions, and reverse backtracking verification processing is performed on state nodes that do not meet the preset stability conditions to obtain state adjustment result information. Based on the state adjustment result information, iterative chain completion processing is performed on the corresponding unfrozen state nodes to obtain the image delivery state sequence information corresponding to each delivery unit.

4. The advertising placement control method according to claim 1, characterized in that, The step of determining the deviation correction result for each delivery unit based on the trajectory deviation result between each actual delivery state trajectory and the corresponding mirrored delivery state trajectory includes: Based on the trajectory deviation results between the actual delivery state trajectory and the corresponding mirrored delivery state trajectory, the deviation state segment information corresponding to each delivery unit is extracted; Based on the deviation state segment information, determine the corresponding mirror reference node information and actual drift node information; Based on the corresponding deviation relationship between the mirror reference node information and the actual drift node information, the reverse traction path information corresponding to the deviation state segment is generated; Based on the reverse traction path information, a step-by-step pullback process is performed on the actual drift node information to obtain the pullback correction state chain information. Based on the pullback correction state chain information, the deviation correction result corresponding to each delivery unit is generated.

5. The advertising placement control method according to claim 1, characterized in that, The step of generating candidate delivery control information for each delivery unit based on the deviation correction result, and determining the state impact result corresponding to each candidate delivery control information, includes: Based on the deviation correction results, determine the corresponding correction traction node information; Based on the corrected traction node information, corresponding candidate delivery control path information is generated; Based on the candidate delivery control path information, impact layer expansion processing is performed on the corresponding state nodes to obtain the initial impact layer information, the transition impact layer information, and the final impact layer information. Based on the initial impact layer information, the transition impact layer information, and the final impact layer information, identify the impact isolation layer information in each of the candidate delivery control path information; Based on the impact isolation layer information, inter-layer peak shaving processing is performed on each of the candidate delivery control path information to obtain the state impact result corresponding to each of the candidate delivery control information.

6. The advertising placement control method according to claim 1, characterized in that, Based on the state impact results, constraint correction processing is performed on the candidate delivery control information to obtain the target delivery control information, including: Based on the state impact result corresponding to the candidate delivery control information, the positive impact constraint interval and the reverse pullback constraint interval are divided. Based on the positive impact constraint interval, the candidate delivery control information is forward-constrained and folded to form positive constraint control chain information; Based on the reverse pullback constraint interval, the positive constraint control chain information is reverse compensated and folded to form bidirectional folded control chain information; Based on the bidirectional folding control chain information, determine the stable output range information; Based on the stable output range information, the corresponding bidirectional folding control chain information is locked to obtain the target delivery control information corresponding to each delivery unit.

7. The advertising placement control method according to claim 1, characterized in that, The step of controlling the corresponding advertising delivery process based on the target delivery control information includes: Based on the target delivery control information corresponding to the delivery unit, generate the corresponding master control execution chain information and shadow control takeover chain information; Based on the master control execution chain information, control is executed on the corresponding advertising delivery process, and based on the shadow control takeover chain information, reverse approximation is performed on the corresponding mirror delivery state trajectory to obtain control takeover approximation result information; Based on the control and approximation result information, determine the corresponding trajectory matching and overtaking result information; If the trajectory alignment and overtaking result information meets the preset control seizure conditions, the shadow control seizure chain information replaces the main control execution chain information; if the trajectory alignment and overtaking result information does not meet the preset control seizure conditions, the main control execution chain information is maintained. Based on the replaced or retained execution chain information, control over the corresponding advertising delivery process is achieved.

8. An advertising delivery control device, characterized in that, The advertising delivery control device includes: The feedback acquisition module is used to acquire real-time feedback data corresponding to each ad delivery unit during the ad delivery process; The status construction module is used to construct the actual delivery status trajectory corresponding to each delivery unit based on the real-time feedback data. The mirror trajectory module is used to generate mirror delivery status trajectories for each delivery unit based on the historical delivery status information and preset control constraint information corresponding to each delivery unit. The deviation correction module is used to determine the deviation correction result for each delivery unit based on the trajectory deviation result between each actual delivery state trajectory and the corresponding mirrored delivery state trajectory. The candidate control module is used to generate candidate delivery control information corresponding to each delivery unit based on the deviation correction result, and to determine the state impact result corresponding to each candidate delivery control information. The target control module is used to perform constraint correction processing on each of the candidate delivery control information based on the state impact results, so as to obtain the target delivery control information; The execution control module is used to control the corresponding advertising delivery process based on the target delivery control information.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the advertising delivery control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the advertising delivery control method as described in any one of claims 1 to 7.

Images