An advertisement intelligent decision method and system based on feature matching

Abstract

The application relates to the technical field of advertisement decision-making, and discloses an advertisement intelligent decision-making method and system based on feature matching, which comprises the following steps: obtaining a feature association sequence by associating a user, a scene and advertisement features; comparing adjacent records from the end of the sequence to the front; merging records with the same scene and advertisement features to form a stable association section, and setting a first advertisement feature difference as a conflict boundary; reserving a section before the boundary and marking the section as a reserved state; disconnecting a rear-end user and advertisement association and marking the association as an excluded state; taking a user feature at the end of the section as a starting point to generate a to-be-reconnected association chain; screening out illegal advertisements, reserving the starting point and the scene association, connecting a compliant advertisement to obtain a candidate decision chain; connecting a stable section of the candidate chain in the reserved state, disconnecting and replacing the candidate chain in the excluded state, and finally outputting an advertisement; and the application can effectively improve advertisement decision-making efficiency.

Description

Technical Field

[0001] This invention relates to the field of advertising decision-making technology, and in particular to an intelligent advertising decision-making method and system based on feature matching. Background Technology

[0002] Current intelligent advertising decision-making technologies primarily rely on the feature relationships between users, scenarios, and advertisements to complete ad matching decisions. In scenarios where ad feature conflicts or disordered relationships occur in the feature association sequence, a global reset and iterative approach is commonly used. This involves re-matching, verifying, and re-determining the entire ad feature association process. This approach repeatedly calculates and processes verified and stable ad association records in the sequence, significantly increasing the data processing volume and computational resource consumption for ad decision-making, resulting in low overall processing efficiency.

[0003] Meanwhile, the existing global iterative decision-making process, when correcting conflict associations, simultaneously breaks the inherent effective association between users and scenarios, completely resetting the matching state of the overall association link. This not only further amplifies the redundant workload of decision adjustment, but also disrupts the stable binding relationship between user features and scenario features in the same scenario. This can easily lead to problems such as disordered advertising decision matching logic and fluctuations in decision results, making it difficult to balance the computational efficiency and stability of advertising decisions. Therefore, how to effectively improve the efficiency of advertising decisions has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides an intelligent advertising decision-making method and system based on feature matching to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an intelligent advertising decision-making method based on feature matching, comprising: S1: Associate user features, scene features, and advertising features that have the same scene identifier to obtain a feature association sequence; S2: Compare the associated records with the adjacent associated records in the feature association sequence forward from the end of the feature association sequence, integrate the associated records with the same scene features and advertising features into a stable association segment, and take the position of the first different advertising feature as the conflict boundary; S3: Retain the stable association segment before the conflict boundary, and the decision association state is the retained state. Remove the association between user features and advertising features after the conflict boundary, and the decision association state is the excluded state. Use the user feature at the end of the stable association segment as the reconnection starting point to form a reconnection chain to be reconnected. S4: Filter out advertisements that do not conform to the preset advertisement constraints, maintain the association between the reconnection starting point and the scene features, and connect the advertisements that have not been filtered out under the same scene identifier to the reconnection starting point of the association chain to be reconnected, forming a candidate decision chain; S5: When the decision association state of the candidate decision chain is in the reserved state, the candidate decision chain is continued to the stable association segment; S6: When the decision association state of the candidate decision chain is excluded, the association between the candidate decision chain and the reconnection starting point is removed, the next candidate decision chain is connected, the candidate decision chain that has been connected is taken as the target decision chain, and the corresponding advertisement is output.

[0006] To address the above problems, the present invention also provides an intelligent advertising decision-making system based on feature matching, the system comprising: The feature association module is used to associate user features, scene features, and advertising features that have the same scene identifier to obtain a feature association sequence; The stable boundary module is used to compare the associated records with adjacent records in the feature association sequence along the end of the feature association sequence, integrate the associated records with the same scene features and advertising features into a stable association segment, and take the position of the first different advertising feature as the conflict boundary. The association reconnection module is used to retain the stable association segment before the conflict boundary, and the decision association state is to retain it. After the conflict boundary, the association between user features and advertising features is to be removed, and the decision association state is to exclude it. The user feature at the end of the stable association segment is used as the reconnection starting point to form the association chain to be reconnected. The candidate construction module is used to filter out advertisements that do not meet the preset advertising constraints, maintain the association between the reconnection starting point and the scene features, and connect the advertisements that have not been filtered out under the same scene identifier to the reconnection starting point of the association chain to be reconnected, forming a candidate decision chain. The retention continuation module is used to continue the candidate decision chain to the stable association segment when the decision association status of the candidate decision chain is in the retention state; The replacement decision module is used to remove the association between the candidate decision chain and the reconnection starting point when the decision association status of the candidate decision chain is in the exclusion state, connect to the next candidate decision chain, take the completed candidate decision chain as the target decision chain, and output the corresponding advertisement.

[0007] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves precise local correction of ad association conflicts by accurately dividing stable association segments and conflict boundaries of feature association sequences, effectively avoiding the problem of repetitive processing of all association data. By retaining effective and stable associations, it selectively resolves user and ad associations only at conflicting locations, significantly reducing the data computation and processing workload of intelligent ad decision-making, effectively improving the overall computational efficiency of ad decision-making, ensuring efficient output of ad placement decisions, and simultaneously maintaining the inherent association between users and scenarios, thus preserving the integrity and continuity of ad matching scenarios.

[0008] 2. This invention achieves accurate access and dynamic adaptation of compliant advertisements by constructing reconnectable association links and a candidate decision link continuation and replacement mechanism. Relying on constraint screening and state verification, it completes the optimal continuation of candidate decision chains, forming a stable and compliant target decision chain. This effectively ensures the stability and accuracy of advertising decision results, avoids decision fluctuations caused by association resets, and can dynamically adapt to the iterative update needs of advertisements in the same scenario, significantly improving the scenario adaptability and decision accuracy of the intelligent advertising decision-making system. Attached Figure Description

[0009] Figure 1 A flowchart illustrating an intelligent advertising decision-making method based on feature matching, provided in an embodiment of the present invention; Figure 2 This is a functional block diagram of an advertising intelligent decision-making system based on feature matching, provided in an embodiment of the present invention.

[0010] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0011] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0012] This application provides a feature-matching-based intelligent advertising decision-making method. The executing entity of this feature-matching-based intelligent advertising decision-making method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the feature-matching-based intelligent advertising decision-making method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0013] Reference Figure 1 The diagram shown is a flowchart illustrating an intelligent advertising decision-making method based on feature matching, according to an embodiment of the present invention. In this embodiment, the intelligent advertising decision-making method based on feature matching includes: S1: Associate user features, scene features, and advertising features that have the same scene identifier to obtain a feature association sequence; Based on the source identifiers carried by user characteristics, scenario characteristics, and advertising characteristics, user characteristics, scenario characteristics, and advertising characteristics with the same source identifier are grouped into the same ad placement association group. Each ad placement record in the ad placement association group contains user characteristics, scenario characteristics, advertising characteristics, and the source identifier of the record itself. The source identifier is used to identify the source system or source data table of the ad placement record. Each ad placement record in the ad placement association group is extracted one by one and its source identifier is verified. When the source identifiers of all ad placement records in the ad placement association group are completely consistent, the ad placement association group is used as the source verification group.

[0014] In one feasible implementation, the source verification group can also be constructed by using source identifier aggregation and comparison.

[0015] Specifically, one ad placement group is selected from all ad placement groups as a pending group. The source identifier of the first ad placement record in the pending group is read as the baseline source identifier. Starting from the second ad placement record in the pending group, the source identifier of each ad placement record is read and compared with the baseline source identifier. When the source identifiers of all ad placement records in the pending group are the same as the baseline source identifier, the source identifiers in the pending group are determined to be consistent, and the pending group is marked as a source verification group. When the source identifier of any ad placement record in the pending group is different from the baseline source identifier, the source identifiers of the pending group are determined to be inconsistent, and the pending group is removed from the subsequent processing flow.

[0016] For each source verification group, each ad delivery record within the group is retrieved sequentially. The user characteristic field, scene characteristic field, and ad characteristic field of the record are checked to see if they all contain non-empty data. If any field is empty, the record is deemed to have an incomplete correspondence. The source verification group is then marked as failing the integrity verification and removed from subsequent processing. When all ad delivery records within the source verification group contain non-empty user characteristics, scene characteristics, and ad characteristics, the source verification group is deemed to have passed the integrity verification. Subsequently, adjacent characteristic combinations within the source verification group are continuously verified according to the ad delivery order.

[0017] In one feasible implementation, continuity verification is accomplished by comparing the timestamp differences between adjacent ad placement records.

[0018] Specifically, the timestamp of the first ad placement record is read from the source verification group that has passed the integrity verification as the preceding timestamp. Starting from the second ad placement record, the timestamp of the current ad placement record is read one by one as the current timestamp. The difference between the current timestamp and the preceding timestamp is calculated to obtain the time difference value. When the time difference value is greater than the preset time threshold, it is determined that the adjacent feature combination of the group is not continuous, the source verification group is marked as failing the continuity verification and removed from subsequent processing. When the time difference value is less than or equal to the preset time threshold, it is determined that the adjacent feature combination of the group is continuous, the current timestamp is updated to the new preceding timestamp, and the verification of the next group of adjacent records continues. The preset time threshold is set by the technicians according to the actual collection frequency of the ad placement system. For example, when the ad placement system works at a frequency of collecting one placement record every 10 seconds, the preset time threshold is set to 30 seconds. When the time difference value of all adjacent records in the source verification group is less than or equal to the preset time threshold, the source verification group is determined to have passed the continuity verification.

[0019] The source verification group that has passed both integrity and continuity verification is designated as the trusted association group. Each ad delivery record is read from the trusted association group, and the scene identifier carried in the record is used as the association primary key. The user features, scene features, and ad features in the record are established according to the same scene identifier. The user features, scene features, and ad features that have established the correspondence are combined into an association record. Multiple ad delivery records with the same scene identifier in a trusted association group form multiple association records. All association records formed in the trusted association group are arranged according to the ad delivery order to obtain the feature association sequence.

[0020] In one feasible implementation, the construction of feature association sequences can also be achieved by first grouping by scene identifier and then sorting by time.

[0021] Specifically, the scene identifiers in each ad delivery record are read from the trusted association group. Ad delivery records with the same scene identifier are grouped into the same scene group. An ad delivery record is taken from the same scene group, and the delivery timestamp in the record is read. All ad delivery records in the scene group are sorted in order of delivery timestamp from earliest to latest. Each sorted ad delivery record is converted into an association record. That is, each association record contains corresponding user features, scene features and ad features. All the sorted association records in the scene groups are sequentially connected in order of delivery start timestamp from earliest to latest according to the scene group to form a feature association sequence. Each association record carries the source identifier and scene identifier of its respective delivery association group.

[0022] S2: Compare the associated records with the adjacent associated records in the feature association sequence forward from the end of the feature association sequence, integrate the associated records with the same scene features and advertising features into a stable association segment, and take the position of the first different advertising feature as the conflict boundary; In one feasible implementation, the integration of stable related segments is achieved by comparing scene features and advertising features one by one.

[0023] Specifically, the last associated record in the feature association sequence is read as the baseline associated record, and the second-to-last associated record is read as the comparison record. Scene feature values ​​and advertising feature values ​​are extracted from the baseline associated record and from the comparison record. The scene feature values ​​of the comparison record are compared character by character with the scene feature values ​​of the baseline associated record, and the advertising feature values ​​of the comparison record are compared character by character with the advertising feature values ​​of the baseline associated record. When the scene feature values ​​and advertising feature values ​​are completely identical, it is determined that the scene feature and advertising feature of the comparison record are the same as those of the baseline associated record. The comparison record is then merged into a stable association segment and used as the new baseline associated record. The process continues to select the associated record adjacent to the new baseline associated record as the new comparison record and repeats the above comparison process.

[0024] Each associated record in the feature association sequence contains a scene feature field and an advertising feature field. The scene feature field stores data describing the scene type, and the advertising feature field stores data describing the advertising attributes. During character comparison, all character sequences stored in the two fields are compared character by character. When the lengths of the two character sequences are the same and the characters at each position are consistent, they are determined to be the same.

[0025] When the scene feature value of the new record to be compared is the same as the scene feature value of the benchmark associated record, the comparison continues between the advertising feature value of the record to be compared and the advertising feature value of the benchmark associated record.

[0026] When the advertising feature value of a new record to be compared is the same as the advertising feature value of the benchmark associated record, the record to be compared is merged into the stable associated segment and the benchmark associated record is updated.

[0027] When the advertising feature value of a new record to be compared is different from the advertising feature value of the benchmark associated record, the merging operation is stopped, all currently merged associated records are taken as stable associated segments, and the record to be compared with the different advertising feature value is taken as the conflict boundary.

[0028] The feature association sequence contains a total of 97 associated records. Starting from the 97th associated record, comparisons are made backwards. The 97th associated record is used as the baseline associated record, and the 96th associated record is used as the record to be compared. When the scene features and advertising features of the 96th and 97th records are the same, the 96th record is merged into the stable association segment and used as the new baseline associated record. The comparison continues with the 95th and 96th records, and so on, moving forward one by one. When the comparison reaches the 78th record, it is found that the advertising feature value of the 78th record is different from the advertising feature value of the 79th record. Merging stops, and records 79 to 97 are used as the stable association segment, which contains a total of 19 associated records. The 78th record is used as the conflict boundary.

[0029] The first association record with different advertising features encountered along the end of the feature association sequence is a conflicting association record, and the association record adjacent to the conflicting association record in the stable association segment is a boundary association record.

[0030] Extract scene feature values ​​from conflict association records and boundary association records, and then perform character comparison between the two scene feature values.

[0031] When the scene feature value of a conflict-related record is the same as the scene feature value of a boundary-related record, the scene features are determined to be the same. At this time, the conflict-related record and the boundary-related record have the same scene features but different advertising features. The connection position between the conflict-related record and the boundary-related record is marked as the conflict boundary.

[0032] In one feasible implementation, the marking of conflict boundaries is achieved by recording the index position of the conflict association record in the feature association sequence.

[0033] Specifically, the sequence number of each associated record in the feature association sequence is read. The sequence number starts from 1 and increments. The sequence number of the boundary association record is 79, and the sequence number of the conflict association record is 78. The position between the sequence number 78 and the sequence number 79 is recorded as the conflict boundary, and the sequence numbers 78 and 79 corresponding to the conflict boundary are written into the boundary mark table.

[0034] In another feasible implementation, the marking of conflict boundaries is achieved by recording pointer offsets.

[0035] Specifically, counting starts from the beginning of the feature association sequence. The count is incremented by 1 for each association record read. When a boundary association record is read, the counter value is 79. When a conflict association record is read, the counter value is 78. The interval between the counter values ​​78 and 79 is recorded as the conflict boundary. The interval start value 78 and the interval end value 79 are stored as boundary position parameters.

[0036] Continuous associated records located in front of the conflict boundary along the direction forward from the end are assigned to the stable side, while associated records located behind the conflict boundary are assigned to the conflict side.

[0037] The direction forward from the end is the direction of moving from the end of the characteristic association sequence toward the starting end. In this direction, the association records before the conflict boundary are all association records with a sequence number greater than 79, and these association records are assigned to the stable side. The association records after the conflict boundary are all association records with a sequence number less than or equal to 78, and these association records are assigned to the conflict side.

[0038] The stable side contains 18 related records from record 80 to record 97, while the conflict side contains 78 related records from record 1 to record 78.

[0039] S3: Retain the stable association segment before the conflict boundary, and the decision association state is the retained state. Remove the association between user features and advertising features after the conflict boundary, and the decision association state is the excluded state. Use the user feature at the end of the stable association segment as the reconnection starting point to form a reconnection chain to be reconnected. Maintain the original arrangement order and connection relationship of each associated record in the stable side. The stable side contains 18 associated records from record 80 to record 97. The associated records are connected sequentially according to their arrangement numbers. Record 80 is connected to record 81, record 81 is connected to record 82, and so on until record 97. Mark the decision association status of each associated record in the stable side as a reserved status.

[0040] Read the user feature field and advertising feature field of each associated record from the associated records on the conflict side, and identify the correspondence between user features and advertising features in the same associated record as an association relationship. When an associated record contains both user feature data and advertising feature data, there is an association relationship between the user feature and the advertising feature.

[0041] For each related record on the conflict side, user feature data and advertising feature data are extracted from the record, and the relationship between user features and advertising features is identified as the conflict relationship.

[0042] Remove the advertising feature data from the associated record corresponding to the conflict relationship. At this time, only the user feature data and the scene feature data are retained in the record. There is no longer a correspondence between the user features and the advertising features. Mark the corresponding decision association state as the exclusion state. At the same time, keep the association relationship between the user features and the scene features in the associated record unchanged. The user features and the scene features still exist in the same record and remain bound.

[0043] Read the associated record at the end of the stable association segment. The stable association segment contains 19 associated records from the 79th to the 97th. The associated record at the end is the 97th associated record. Extract the user feature data stored in the user feature field from the 97th associated record as the reconnection starting point.

[0044] The connection relationships between the records in the stable association segment are preserved according to the original association order. In the stable association segment, record 79 connects to record 80, record 80 connects to record 81, and so on up to record 97. The reconnection start point is taken as the starting access point of the association chain to be reconnected. The user characteristics in the 97th association record are taken as the connection port that can be connected outward at the starting access point, thus obtaining the association chain to be reconnected. The association chain to be reconnected contains 19 association records from record 79 to record 97, as well as the reconnection start point located at the end of the 97th association record.

[0045] In one feasible implementation, the labeling of decision-related states is achieved by performing state determination on all related records of the stable side and the conflict side.

[0046] Specifically, all 18 associated records (80 to 97) on the stable side and 78 associated records (1 to 78) on the conflict side are used as status determination records. For each associated record in the status determination records, it is verified whether both the user feature field and the scene feature field in the associated record contain non-empty data. When both fields are non-empty, it is determined that the association between the user feature and the scene feature is maintained. When either field is empty, it is determined that the association between the user feature and the scene feature is not maintained. At the same time, the advertising feature field of the associated record is read, and the permutation number of the associated record in the feature association sequence is read. The advertising feature field of the adjacent associated record is taken out and compared with the advertising feature field of the current associated record. When the character sequences of the two advertising feature fields are completely consistent, it is determined that the advertising features are consistent. When the character sequences are inconsistent, it is determined that the advertising features are inconsistent.

[0047] For each associated record in the status determination record, obtain the attribution information of the associated record, the association maintenance determination result between user features and scene features, and the consistency determination result of advertising features. When the associated record belongs to the stable side, the association between user features and scene features is maintained, and the advertising features are consistent, the status verification result of the associated record is determined to be a retained state. When the associated record belongs to the conflict side or the association between user features and advertising features is terminated, the status verification result of the associated record is determined to be an excluded state.

[0048] In another feasible implementation, the labeling of the decision association state is achieved by traversing the feature association sequence one by one and writing the state according to the record position.

[0049] Specifically, starting from the first associated record in the feature association sequence, for each read associated record, the permutation number of the associated record is obtained. The permutation number is compared with the permutation numbers 78 and 79 corresponding to the conflict boundary. When the permutation number is greater than 79, the associated record is determined to belong to the stable side, and the decision association status of the associated record is written as the reserved state. When the permutation number is less than or equal to 78, the associated record is determined to belong to the conflict side. Then, it is verified whether the advertising feature field of the associated record has been removed. When the advertising feature field has been removed, the association between the user feature and the advertising feature is determined to be terminated, and the decision association status of the associated record is written as the excluded state. The data content written to the reserved state or excluded state is stored in the status field corresponding to the associated record to obtain the decision association status of the associated record, until the 97th associated record is traversed.

[0050] S4: Filter out advertisements that do not conform to the preset advertisement constraints, maintain the association between the reconnection starting point and the scene features, and connect the advertisements that have not been filtered out under the same scene identifier to the reconnection starting point of the association chain to be reconnected, forming a candidate decision chain; Extract all advertisements under the same scene identifier one by one, read the advertisement attribute data carried by each advertisement, and compare each field in the advertisement attribute data with the corresponding constraint field in the preset advertisement constraints item by item.

[0051] The preset advertising constraints are stored as a list of constraints, which contains a total of 5 constraints: the advertising review status is approved, the advertising industry category belongs to the allowed placement set, the advertising placement time covers the current time, the advertising budget balance is greater than 0 yuan, and the advertising placement frequency does not exceed the daily limit of 5 times.

[0052] For each advertisement, verify in sequence whether the advertisement's review status field is equal to "approved", whether the industry category field exists in the allowed placement set, whether the placement time field includes the current time, whether the budget balance field is greater than 0, and whether the number of placements for the day is less than 5.

[0053] When all five constraints of an advertisement are met, the advertisement is marked as meeting the preset advertising constraints and is included as an advertisement to be accessed. When any one of the constraints of an advertisement is not met, the advertisement is marked as not meeting the preset advertising constraints and is removed from subsequent processing.

[0054] In one feasible implementation, the comparison of preset advertising constraints is achieved by sequentially verifying each constraint condition.

[0055] Specifically, the first ad is read from the list of stored ad data. The review status, industry category, time period, budget balance, and frequency of ad placement are verified in the order listed in the constraint condition list. When the second industry category is verified, it is found that the industry category of the ad is not in the allowed placement set. The subsequent verification of the ad is stopped and the ad is removed. The next ad is read and the verification starts from the beginning. There are a total of 24 ads to be verified. After verification, 9 ads meet all 5 constraints. These 9 ads are selected as ads to be accessed.

[0056] In another feasible implementation, the comparison of preset advertising constraints is achieved by first screening and then confirming each item.

[0057] Specifically, from all 24 ads, ads with an approved status are first selected, resulting in 21 ads. From these 21 ads, ads whose industry category belongs to the allowed placement set are selected, resulting in 17 ads. From these 17 ads, ads whose placement time covers the current time are selected, resulting in 13 ads. From these 13 ads, ads with a budget balance greater than 0 yuan are selected, resulting in 11 ads. From these 11 ads, ads that have been placed less than 5 times that day are selected, resulting in 9 ads. The final 9 ads are selected as ads to be integrated.

[0058] Read the association record corresponding to the reconnection starting point from the association chain to be reconnected, namely the 97th association record. This association record contains user feature data and scene feature data. The user feature data and scene feature data are stored in the same record and the association relationship between the two is maintained. Extract the scene feature data from this association record and keep the scene feature data and the user feature data of the reconnection starting point in the same association record without separating them.

[0059] Ads to be accessed are arranged according to a preset access order, which is the order of ad priority from high to low. Each ad carries a priority value, which ranges from 1 to 100. The larger the value, the higher the priority. The nine ads to be accessed carry priority values ​​of 85, 72, 68, 65, 59, 53, 47, 41, and 36 respectively. These nine ads to be accessed are arranged in descending order of priority value as the first ad to be accessed, to the ninth ad to be accessed.

[0060] Read the advertising features and their decision association states corresponding to the historical association records from the decision association states of the feature association sequence to obtain the decision association state correspondence. The correspondence records three types of historical advertising features that are marked as excluded: advertising feature A, advertising feature B, and advertising feature C.

[0061] For each sorted advertisement to be accessed, read the advertisement features of the advertisement to be accessed, and compare the advertisement features with the three advertisement features marked as excluded in the decision association status correspondence one by one.

[0062] When the ad feature of the ad to be accessed is completely consistent with the character sequence of any one of ad feature A, ad feature B, or ad feature C, the ad to be accessed is associated with the exclusion state. When the ad feature of the ad to be accessed is inconsistent with all three of the ad features A, B, and C, the ad to be accessed is associated with the retention state.

[0063] The advertising features of the 9 pending advertisements are advertising feature D, advertising feature E, advertising feature A, advertising feature F, advertising feature G, advertising feature B, advertising feature H, advertising feature I, and advertising feature C. Among them, advertising features A, advertising feature B, and advertising feature C correspond to the first pending advertisement, the sixth pending advertisement, and the ninth pending advertisement, respectively. These 3 pending advertisements are associated with the exclusion state, and the remaining 6 pending advertisements are associated with the retention state.

[0064] For each advertisement to be accessed, the advertisement feature data of the advertisement to be accessed is combined with the corresponding decision association status into a data record. Then, the data record is associated with the reconnection starting point, that is, a correspondence is established between the data record and the user feature data corresponding to the reconnection starting point to obtain a candidate association record.

[0065] Nine candidate association records are generated for each of the nine advertisements to be integrated. Each candidate association record contains the advertisement characteristics of the advertisement to be integrated, the corresponding decision association status, and the correspondence between the advertisement and the user characteristics of the reconnection starting point.

[0066] Each candidate association record is connected to the reconnection start point of the association chain to be reconnected. That is, after the candidate association record is connected to the reconnection start point at the end of the 97th association record, a complete decision chain is formed. After the 9 candidate association records are connected, 9 candidate decision chains are formed. Each candidate decision chain contains the 79th to 97th association records and a candidate association record connected to the end of the chain.

[0067] S5: When the decision association state of the candidate decision chain is in the reserved state, the candidate decision chain is continued to the stable association segment; Read the associated record that is directly connected to the reconnection starting point from the candidate decision chain, and use it as the candidate associated record. For each candidate decision chain, the end position of the candidate decision chain is the position of the candidate associated record. One end of the candidate associated record is connected to the 97th associated record corresponding to the reconnection starting point, and the other end is the end point of the candidate decision chain.

[0068] When the value of the decision association status field of a candidate association record is "reserved", the candidate association record is determined to be in the "reserved" state. At this time, the association relationship between the candidate association record and the reconnection starting point remains unchanged, that is, the candidate association record is still connected to the end of the 97th association record.

[0069] When the value of the decision association status field of a candidate association record is "exclusion status", the candidate association record is determined to be in the exclusion status and no continuation operation is performed.

[0070] In one feasible implementation, the extraction of candidate association records is accomplished by traversing the end nodes of the candidate decision chain.

[0071] Specifically, each of the nine candidate decision chains contains the same stable association segment, which contains 19 association records from the 79th to the 97th. The difference between the nine candidate decision chains lies in the candidate association records connected after the 97th record. For each candidate decision chain, starting from the 79th record at the beginning of the chain, each record is read sequentially along the connection direction until the association record with the largest sequence number in the chain is read. This record is the candidate association record connected to the reconnection starting point. The end records of the nine candidate decision chains are read sequentially to obtain the nine candidate association records.

[0072] For each of the nine candidate association records, the decision association status field is read one by one. The decision association status of the first candidate association record is excluded, the second is retained, the third is excluded, the fourth is retained, the fifth is retained, the sixth is excluded, the seventh is retained, the eighth is retained, and the ninth is excluded.

[0073] The candidate decision chains are checked sequentially starting from the first one. The decision association status of the first candidate association record is excluded, which does not meet the retention condition. The candidate decision chain is skipped and the second one is checked. The decision association status of the second candidate association record is retained, which meets the retention condition. The association between the second candidate association record and the reconnection starting point is maintained.

[0074] Candidate association records are appended to the end of the stable association segment according to the arrangement direction of the feature association sequence, which is from the 1st to the 97th. The end of the stable association segment is the location of the 97th association record. The candidate association record is appended to the 97th association record, making the candidate association record a new end record of the stable association segment.

[0075] In one feasible implementation, the continuation operation is completed by pointing the forward pointer of the candidate associated record to the 97th associated record.

[0076] Specifically, each associated record carries a forward pointer field and a backward pointer field. The forward pointer field stores the sequence number of the record preceding the current record, and the backward pointer field stores the sequence number of the record following the current record. The backward pointer field of the 97th associated record is currently empty. The sequence number of the 2nd candidate associated record is read as 98. The 98th sequence number is written into the backward pointer field of the 97th associated record, and the 97th sequence number is written into the forward pointer field of the 98th candidate associated record. At this time, the 98th candidate associated record has been continued after the 97th associated record.

[0077] Update the 98th candidate association record to the end of the stable association segment. The original end association record was the 97th record. After the update, the end of the stable association segment becomes the 98th candidate association record. At this time, the stable association segment contains 20 association records from the 79th to the 98th record.

[0078] Stop connecting subsequent candidate decision chains, that is, no further connection operations will be performed on candidate decision chains 3 to 9, and these candidate decision chains will remain unconnected.

[0079] The entire decision chain that has completed the continuation operation of the second candidate decision chain is taken as the candidate decision chain that has completed the continuation. The candidate decision chain that has completed the continuation includes the 79th to 97th stable association segments and the 98th candidate association record that follows the 97th record, for a total of 20 association records.

[0080] In another feasible implementation, the continuation operation is completed by updating the list of first and last pointers of the stable associated segment.

[0081] Specifically, the stable association segment is stored as a list of association records. The start pointer of this list points to the 79th association record, and the end pointer points to the 97th association record. When the 2nd candidate association record is in a reserved state, the storage address of the candidate association record is read and written to the end pointer of the stable association segment, so that the end pointer changes from pointing to the 97th record to pointing to the 98th candidate association record. At the same time, the total number of records in the stable association segment is updated from 19 to 20. After the above update operation is completed, a complete candidate decision chain is formed.

[0082] S6: When the decision association state of the candidate decision chain is excluded, the association between the candidate decision chain and the reconnection starting point is removed, the next candidate decision chain is connected, the candidate decision chain that has been connected is taken as the target decision chain, and the corresponding advertisement is output.

[0083] Extract candidate decision chains whose decision association status is excluded, as well as candidate association records in the candidate decision chain that are connected to the reconnection start point.

[0084] There are a total of 9 candidate decision chains in the candidate decision chain set, which are arranged in the preset access order as the first candidate decision chain to the ninth candidate decision chain. Among them, the decision association status of the first, third, sixth and ninth candidate decision chains is excluded, while the decision association status of the second, fourth, fifth, seventh and eighth candidate decision chains is retained.

[0085] Read the candidate association record connected to the reconnection starting point in the first candidate decision chain, namely the 98th candidate association record, which is connected to the end of the 97th association record.

[0086] Clear the forward pointer field pointing to the 97th associated record in the 98th candidate associated record, and at the same time clear the backward pointer field pointing to the 98th candidate associated record in the 97th associated record. At this time, the association between the 98th candidate associated record and the reconnection starting point is broken.

[0087] The first candidate decision chain is marked as an excluded association chain, and this mark is stored in the status mark field of the candidate decision chain.

[0088] According to the preset access order of the candidate decision chain set, the candidate decision chain located after the excluded association chain is selected. The excluded association chain is the first candidate decision chain, and the one following it is the second candidate decision chain. The end candidate association record of the second candidate decision chain, namely the 99th candidate association record, is read.

[0089] The 99th candidate association record in the second candidate decision chain is associated with the reconnection starting point. Specifically, the backward pointer field of the 97th association record is written to the 99th permutation number, and the forward pointer field of the 99th candidate association record is written to the 97th permutation number.

[0090] Verify the decision association status corresponding to the second candidate decision chain. If the decision association status of the candidate association record at the end of the second candidate decision chain is a reserved status, the continuation condition is met, and the second candidate decision chain is continued to the stable association segment.

[0091] In one feasible implementation, when the decision association status of a candidate decision chain is verified to be excluded, the association of the candidate decision chain is further removed and the next candidate decision chain is selected for replacement.

[0092] Specifically, if the first candidate decision chain in the preset access order is in an excluded state, its association with the reconnection starting point is removed and it is marked as an excluded association chain. Then, the second candidate decision chain is selected. If the decision association state of the second candidate decision chain is still in an excluded state, the association between the second candidate decision chain and the reconnection starting point is removed and the second candidate decision chain is also marked as an excluded association chain. Then, the third candidate decision chain is selected for verification. This process is repeated until a candidate decision chain with a decision association state in a reserved state is found.

[0093] When the decision association status of the second candidate decision chain is in the reserved state, the second candidate decision chain is continued to the stable association segment. The continuation operation is to connect the 99th candidate association record in the second candidate decision chain to the end of the stable association segment, i.e., after the 97th association record, and update the 99th candidate association record to the end association record of the stable association segment. At this time, the stable association segment contains 20 association records, including records 79 to 97 and record 99.

[0094] The second candidate decision chain that completes the continuation is taken as the target decision chain, which includes the 79th to 97th stable association segments and the 99th candidate association record that continues at the end.

[0095] Read the advertisement corresponding to the end-of-target decision chain record. The end-of-target decision chain record is candidate record number 99. Read the advertisement feature field from this record and use the advertisement corresponding to the stored advertisement feature data as the target advertisement.

[0096] In one feasible implementation, the output of the target advertisement is accomplished by reading the advertisement identifier field of the associated record at the end of the target decision chain.

[0097] Specifically, in addition to the advertising feature field, candidate association record No. 99 also includes an advertising identifier field. The advertising identifier field stores the unique number of the advertisement in the advertising database. The advertising material file corresponding to the unique number is read from the advertising database and sent to the advertising delivery port.

[0098] In another feasible implementation, the output of the target advertisement is accomplished by directly reading the advertisement content field of the associated record at the end of the target decision chain.

[0099] Specifically, the 99th candidate association record directly stores the display content data of the advertisement. The display content data includes the advertisement title text, the storage path of the advertisement image, and the advertisement landing page link. The advertisement title text, the storage path of the advertisement image, and the advertisement landing page link are read one by one from this record. The three items read are combined into a complete advertisement output data packet, and the advertisement output data packet is sent to the advertisement display terminal.

[0100] like Figure 2 The diagram shown is a functional block diagram of an advertising intelligent decision-making system based on feature matching provided in an embodiment of the present invention.

[0101] The feature-matching-based intelligent advertising decision-making system 100 described in this invention can be installed in an electronic device. Depending on the functions implemented, the feature-matching-based intelligent advertising decision-making system 100 may include a feature association module 101, a stable boundary module 102, an association reconnection module 103, a candidate construction module 104, a retention continuation module 105, and a replacement decision module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0102] In this embodiment, the functions of each module / unit are as follows: The feature association module 101 is used to associate user features, scene features and advertising features with the same scene identifier to obtain a feature association sequence; The stable boundary module 102 is used to compare the associated records with adjacent records in the feature association sequence along the end of the feature association sequence, integrate the associated records with the same scene features and advertising features into a stable association segment, and take the position of the first different advertising feature as the conflict boundary. The association reconnection module 103 is used to retain the stable association segment before the conflict boundary and decide the association state as the retention state; and to remove the association between the user feature and the advertising feature after the conflict boundary and decide the association state as the exclusion state; and to form an association chain to be reconnected by taking the user feature at the end of the stable association segment as the reconnection starting point. The candidate construction module 104 is used to filter out advertisements that do not meet the preset advertisement constraints, maintain the association between the reconnection starting point and the scene features, and connect the advertisements that have not been filtered out under the same scene identifier to the reconnection starting point of the association chain to be reconnected, forming a candidate decision chain. The retention continuation module 105 is used to continue the candidate decision chain to the stable association segment when the decision association state of the candidate decision chain is in the retention state. The replacement decision module 106 is used to, when the decision association state of the candidate decision chain is in the exclusion state, disconnect the candidate decision chain from the reconnection starting point, connect to the next candidate decision chain, take the completed candidate decision chain as the target decision chain, and output the corresponding advertisement.

[0103] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0104] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0105] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0106] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0107] The embodiments of this application can acquire and process relevant data based on an artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A feature-matching-based intelligent advertising decision-making method, characterized in that, The method includes: S1: Associate user features, scene features, and advertising features that have the same scene identifier to obtain a feature association sequence; S2: Compare the associated records with the adjacent associated records in the feature association sequence forward from the end of the feature association sequence, integrate the associated records with the same scene features and advertising features into a stable association segment, and take the position of the first different advertising feature as the conflict boundary; S3: Retain the stable association segment before the conflict boundary, and the decision association state is the retained state. Remove the association between user features and advertising features after the conflict boundary, and the decision association state is the excluded state. Use the user feature at the end of the stable association segment as the reconnection starting point to form a reconnection chain to be reconnected. S4: Filter out advertisements that do not conform to the preset advertisement constraints, maintain the association between the reconnection starting point and the scene features, and connect the advertisements that have not been filtered out under the same scene identifier to the reconnection starting point of the association chain to be reconnected, forming a candidate decision chain; S5: When the decision association state of the candidate decision chain is in the reserved state, the candidate decision chain is continued to the stable association segment; S6: When the decision association state of the candidate decision chain is excluded, the association between the candidate decision chain and the reconnection starting point is removed, the next candidate decision chain is connected, the candidate decision chain that has been connected is taken as the target decision chain, and the corresponding advertisement is output.

2. The intelligent advertising decision-making method based on feature matching as described in claim 1, characterized in that, The association of user features, scene features, and advertising features with the same scene identifier to obtain a feature association sequence includes: Based on the source identifiers corresponding to user characteristics, scenario characteristics, and advertising characteristics, user characteristics, scenario characteristics, and advertising characteristics are grouped and integrated into a placement association group. Placement association groups in which the source identifiers of each advertising placement record are consistent are designated as source verification groups. Check whether the correspondence between user features, scene features and advertising features in the source verification group is complete, and check whether adjacent feature combinations are continuous according to the order of advertising placement. The source verification group that passes the integrity verification and continuity verification is regarded as a trusted association group. Establish a correspondence between user features, scene features and advertising features with the same scene identifier in the trusted association group to form an association record, and arrange the association record according to the order of advertising placement to obtain a feature association sequence.

3. The intelligent advertising decision-making method based on feature matching as described in claim 1, characterized in that, The process of comparing adjacent associated records along the end of the feature association sequence forward, integrating associated records with the same scene features and advertising features into a stable association segment, and using the position of the first differing advertising feature as the conflict boundary, includes: The associated record located at the end of the feature association sequence is taken as the reference associated record, and the associated record adjacent to the reference associated record is selected forward along the end of the feature association sequence as the record to be compared; When the scene features and advertising features of the record to be compared are the same as those of the benchmark associated record, the record to be compared is merged into the stable associated segment and the record to be compared is used as the new benchmark associated record. Continue selecting associated records adjacent to the new benchmark associated record until associated records with different advertising characteristics appear, thus obtaining a stable associated segment, and using the associated records with different advertising characteristics as conflict boundaries.

4. The intelligent advertising decision-making method based on feature matching as described in claim 3, characterized in that, The step of using the associated records with different advertising features as conflict boundaries includes: The first association record with different advertising features along the end of the feature association sequence is designated as a conflict association record, and the association record adjacent to the conflict association record in the stable association segment is designated as a boundary association record. Verify whether the conflict association record and the boundary association record have the same scene characteristics. When the scene characteristics are the same but the advertising characteristics are different, mark the connection position between the conflict association record and the boundary association record as a conflict boundary. The continuous associated records located in front of the conflict boundary in the direction forward from the end are assigned to the stable side, and the associated records located behind the conflict boundary are assigned to the conflict side.

5. The intelligent advertising decision-making method based on feature matching as described in claim 4, characterized in that, The stable association segment before the conflict boundary is retained, and the association state is retained. After the conflict boundary, the association between user features and advertising features is removed, and the association state is excluded. The user feature at the end of the stable association segment is used as the reconnection starting point to form a reconnection chain, including: Maintain the association relationships between the associated records on the stable side, and mark the decision association status corresponding to the stable side as a reserved status; Identify conflicting associations between user features and advertising features from the associated records on the conflicting side; The conflict association is resolved, the corresponding decision association state is marked as excluded, and the association between the corresponding user features and scene features is maintained; The user characteristics in the associated record at the end of the stable association segment are selected as the reconnection starting point of the stable association segment; The connection relationship between the stable association segment and the reconnection starting point is preserved according to the original association order to obtain the association chain to be reconnected.

6. The intelligent advertising decision-making method based on feature matching as described in claim 5, characterized in that, The decision-related state also includes: The associated records in the stable side and the conflict side are used as state determination records. The association between the user characteristics and scene characteristics of each associated record in the state determination record is verified to see if it is maintained. The advertising characteristics of adjacent associated records are compared to see if they are consistent, and the state verification result is obtained. The state verification result when the associated record is located on the stable side, the association between user features and scene features is maintained, and the advertising features are consistent is taken as the retained state; the state verification result when the associated record is located on the conflict side or the association between user features and advertising features is broken is taken as the excluded state. Write the retained state or the excluded state into the corresponding associated record to obtain the decision association state of the associated record.

7. The intelligent advertising decision-making method based on feature matching as described in claim 1, characterized in that, The process of filtering out ads that do not meet the preset ad constraints, maintaining the association between the reconnection starting point and the scene features, and connecting the ads that have not been filtered out under the same scene identifier to the reconnection starting point of the association chain to be reconnected, forming a candidate decision chain, includes: Each advertisement under the same scene identifier is compared with the preset advertisement constraints. Advertisements that do not meet the preset advertisement constraints are filtered out, and advertisements that meet the preset advertisement constraints are selected as advertisements to be accessed. Maintain the association between the reconnection starting point and the scene features, and arrange the advertisements to be accessed according to the preset access order; The advertising features of each ad to be accessed are compared with the correspondence of the decision association states. Ads to be accessed that have the same advertising features as the ad to be excluded are associated with the exclusion state, and the remaining ads to be accessed are associated with the retention state. Each advertisement to be accessed and its corresponding decision association status are associated with the reconnection starting point to obtain candidate association records. Each candidate association record is then connected to the reconnection association chain to obtain a candidate decision chain corresponding to each advertisement to be accessed.

8. The intelligent advertising decision-making method based on feature matching as described in claim 1, characterized in that, When the decision association state of the candidate decision chain is in a reserved state, the candidate decision chain is continued to the stable association segment, including: Extract the associated records connected to the reconnection starting point from the candidate decision chain, and use them as candidate associated records; Verify whether the decision association status corresponding to the candidate association record is a reserved status, and if the decision association status is a reserved status, maintain the association between the candidate association record and the reconnection starting point; According to the arrangement direction of the feature association sequence, the candidate association record is continued to the end of the stable association segment; The candidate association record is updated to the end association record of the stable association segment, and the access of subsequent candidate decision chains is stopped, thus forming a complete candidate decision chain.

9. The intelligent advertising decision-making method based on feature matching as described in claim 1, characterized in that, When the decision association state of the candidate decision chain is in the exclusion state, the association between the candidate decision chain and the reconnection starting point is removed, the next candidate decision chain is connected, the completed candidate decision chain is taken as the target decision chain, and the corresponding advertisement is output, including: Extract candidate decision chains whose decision association status is excluded, and candidate association records in the candidate decision chains that are connected to the reconnection starting point; The association between the candidate association record and the reconnection start point is removed, and the candidate decision chain is marked as an excluded association chain. Then, according to the preset access order of the candidate decision chain set, the candidate decision chain located after the excluded association chain is selected. The selected candidate decision chain is associated with the reconnection starting point, and the corresponding decision association status is verified; when the corresponding decision association status is excluded, the association is terminated and the next candidate decision chain is selected. When the corresponding decision association state is a reserved state, the corresponding candidate decision link is continued to the stable association segment, and the completed candidate decision chain is taken as the target decision chain. Read the advertisement corresponding to the end of the target decision chain and output the advertisement as the target advertisement.

10. An intelligent advertising decision-making system based on feature matching, characterized in that, The system is used to implement the feature-matching-based intelligent advertising decision-making method according to claim 1, the system comprising: The feature association module is used to associate user features, scene features, and advertising features that have the same scene identifier to obtain a feature association sequence; The stable boundary module is used to compare the associated records with adjacent records in the feature association sequence along the end of the feature association sequence, integrate the associated records with the same scene features and advertising features into a stable association segment, and take the position of the first different advertising feature as the conflict boundary. The association reconnection module is used to retain the stable association segment before the conflict boundary, and the decision association state is to retain it. After the conflict boundary, the association between user features and advertising features is to be removed, and the decision association state is to exclude it. The user feature at the end of the stable association segment is used as the reconnection starting point to form the association chain to be reconnected. The candidate construction module is used to filter out advertisements that do not meet the preset advertising constraints, maintain the association between the reconnection starting point and the scene features, and connect the advertisements that have not been filtered out under the same scene identifier to the reconnection starting point of the association chain to be reconnected, forming a candidate decision chain. The retention continuation module is used to continue the candidate decision chain to the stable association segment when the decision association status of the candidate decision chain is in the retention state; The replacement decision module is used to remove the association between the candidate decision chain and the reconnection starting point when the decision association status of the candidate decision chain is in the exclusion state, connect to the next candidate decision chain, take the completed candidate decision chain as the target decision chain, and output the corresponding advertisement.

Images