Cross-platform income data closed-loop analysis method

Abstract

The invention discloses a cross-platform income data closed-loop analysis method, and particularly relates to the technical field of data closed-loop analysis. The method comprises the following steps: collecting original multi-platform interaction data, and generating a cross-platform interaction data set; executing multi-layer anonymous identifier mapping, and outputting a unified user link identifier sequence and event sequence mapping table; calculating an observability matrix of each platform event chain to obtain a cross-platform observability score set; a cross-platform implicit path sequence set is mined based on a probabilistic graph reasoning method, and a backflow path complete sequence is constructed in combination with a cross-platform observability score set; performing income event association calculation and link credibility weight normalization, and outputting a closed-loop income attribution result; the closed-loop income attribution result is fed back to the cross-platform putting strategy database, closed-loop optimization is completed, accurate reduction of the real conversion path of the user and closed-loop quantitative analysis of multi-platform income contribution are achieved, and the optimization effect and the actual income performance of the cross-platform putting strategy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data closed-loop analysis, and more specifically, to a cross-platform revenue data closed-loop analysis method. Background Art

[0002] Existing cross-platform revenue data analysis methods mainly rely on the built-in data tracking mechanism and embedding technology of a single platform to identify user behavior paths and attribute revenue.

[0003] In actual business scenarios, users' decision-making and conversion paths often span multiple platforms of different natures, making it difficult to effectively restore and penetrate the user's complete cross-platform return behavior path. As a result, existing analysis methods have broken links in the revenue attribution path or inaccurate attribution, which in turn causes deviations in the optimization of cross-platform delivery strategies, restricting the improvement of cross-platform delivery strategies and revenue optimization capabilities. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a cross-platform revenue data closed-loop analysis method to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A closed-loop analysis method for cross-platform revenue data includes the following steps: S1: Collect original multi-platform interaction data and generate a cross-platform interaction dataset based on a unified timestamp and cross-platform business identifier; S2: Based on the cross-platform interaction dataset, a multi-layer anonymous identity mapping algorithm is executed to output a unified user link identity sequence and event sequence mapping table; S3: Calculate the observability matrix of each platform's event chain based on the unified user link identification sequence and output a cross-platform observability score set; S4: Based on the unified user link identification sequence, a probabilistic graph reasoning method is used to output a set of cross-platform implicit path sequences; S5: Fuse the cross-platform observability score set and the cross-platform implicit path sequence set to construct a complete sequence of reflux paths; S6: Perform revenue event correlation calculation and link credibility weight normalization on the complete sequence of the return path, and output the closed-loop revenue attribution result; S7: Feedback the closed-loop revenue attribution results to the cross-platform delivery strategy database to complete closed-loop optimization.

[0006] In a preferred embodiment, S1 is specifically: Collect raw multi-platform interaction data including ad click event data, product browsing event data, and order settlement event data; According to the cross-platform unified timestamp standard, the original multi-platform interaction data is uniformly timestamped; Based on the unified cross-platform business identification rules, the original multi-platform interaction data is marked with business identification; The original multi-platform interaction data is matched with the time series and business logic according to the unified timestamp tag and business identification tag to generate a cross-platform interaction dataset.

[0007] In a preferred embodiment, the multi-layer anonymous identity mapping algorithm includes device-level anonymous identity mapping, account-level anonymous identity mapping, and session-level anonymous identity mapping.

[0008] In a preferred embodiment, S2 is specifically: Device-level anonymous identification mapping uses the device identification data of each platform in the cross-platform interaction dataset to perform unified mapping and output a device-level anonymous identification sequence; Account-level anonymous identifier mapping uses the account identifier data of each platform in the cross-platform interaction dataset to perform unified mapping and output an account-level anonymous identifier sequence; The session-level anonymous identifier mapping uses the session identifier data of each platform in the cross-platform interaction dataset to perform unified mapping and output a session-level anonymous identifier sequence; All identification sequences are sorted based on the cross-platform unified timestamp tag to generate a unified user link identification sequence; Establish an event sequence mapping table between a unified user link identification sequence and a cross-platform interaction dataset.

[0009] In a preferred embodiment, S3 is specifically: From the cross-platform interaction dataset, event sequence groups are extracted based on the unified user link identifier sequence and platform dimension. For each platform’s event sequence group, an event integrity matching method is applied to determine the event integrity score; For each platform's event sequence group, apply data visibility assessment rules to determine a data visibility score; Based on the event completeness score and data visibility score, an event chain observability score is generated according to the platform order; The observability scores of each platform's event chain are organized according to the platform dimension to form an event chain observability matrix; Extract matrix elements from the event chain observability matrix and aggregate them according to the unified user link identifier to output a cross-platform observability score set.

[0010] In a preferred embodiment, S4 is specifically: Based on the unified user link identification sequence and cross-platform interaction dataset, an initial directed graph of cross-platform user behavior paths is constructed; The user behavior nodes recorded in the initial directed graph of the cross-platform user behavior path are defined as explicit nodes; The platform nodes that do not record user behavior data in the initial directed graph of cross-platform user behavior paths are defined as hidden nodes; Based on the event context and event link structure between explicit and implicit nodes, the probabilistic graph reasoning method is used to infer the existence and location of implicit nodes and generate an implicit node inference sequence. Determine the set of cross-platform implicit path sequences based on implicit node inference sequences.

[0011] In a preferred embodiment, S5 is specifically: According to the observability score of each platform event chain in the cross-platform observability score set, a credibility value is assigned to each cross-platform implicit path sequence in the cross-platform implicit path sequence set; After credibility is assigned, the paths of each cross-platform implicit path sequence are sorted according to the unified user link identification sequence; The path sorting results are merged with the explicit nodes recorded in the cross-platform interaction dataset to form a complete sequence of reflux paths containing explicit nodes and implicit nodes.

[0012] In a preferred embodiment, S6 is specifically: Determine the association relationship between the explicit nodes and implicit nodes in the complete sequence of the return path and the order settlement event data according to the unified user link identification sequence; According to the association relationship, the initial benefit weight values ​​are assigned to the explicit nodes and implicit nodes in the complete sequence of the return path; Based on the observability score of the event chain, the initial benefit weight values ​​of the explicit nodes and implicit nodes are respectively modified for credibility; Normalize the benefit weight value after credibility correction to generate the normalized link credibility weight result; The closed-loop benefit attribution result is determined based on the normalized result of the link credibility weight.

[0013] In a preferred embodiment, S7 is specifically: Import closed-loop revenue attribution results into the cross-platform delivery strategy database; Based on the unified user link identification sequence, the closed-loop revenue attribution results are matched with the cross-platform revenue data in the cross-platform delivery strategy database; Based on the data matching results, the cross-platform advertising data and product browsing data delivery strategy are adjusted through the cross-platform delivery strategy database; Update the cross-platform delivery strategy database based on the adjusted delivery strategy to complete closed-loop optimization.

[0014] The technical effects and advantages of the cross-platform closed-loop revenue data analysis method of the present invention are as follows: 1. By unifying timestamps and cross-platform business identifiers, raw interaction data scattered across different platforms is aggregated into comparable datasets, breaking down data silos. A multi-layer anonymous identifier mapping algorithm accurately links devices, accounts, and sessions while protecting user privacy, achieving cross-platform user identity unification. The event chain observability matrix quantifies the data integrity and visibility of each platform, providing a reliability assessment for path reasoning. Probabilistic graph reasoning methods can penetrate unauthorized or missing nodes, complete hidden user behavior paths, and avoid attribution chain breaks.

[0015] 2. By constructing a complete reflux path sequence, we achieve full-link behavior tracking. Based on this complete reflux path sequence, we correlate revenue events and normalize credibility weights to achieve more precise revenue contribution allocation. By feeding closed-loop revenue attribution results back into the delivery strategy database, we can adjust cross-platform delivery strategies, thereby establishing a closed-loop optimization mechanism. While maintaining privacy, security, and compliance, this approach overcomes data blind spots across multiple platforms, providing refined and actionable decision-making for cross-platform delivery strategies and significantly improving advertising ROI and user conversion efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of a closed-loop analysis method for cross-platform revenue data according to the present invention. DETAILED DESCRIPTION

[0017] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] Example Figure 1 The present invention provides a closed-loop analysis method for cross-platform revenue data, which includes the following steps: S1: Collect original multi-platform interaction data and generate a cross-platform interaction dataset based on a unified timestamp and cross-platform business identifier; S2: Based on the cross-platform interaction dataset, a multi-layer anonymous identity mapping algorithm is executed to output a unified user link identity sequence and event sequence mapping table; S3: Calculate the observability matrix of each platform's event chain based on the unified user link identification sequence and output a cross-platform observability score set; S4: Based on the unified user link identification sequence, a probabilistic graph reasoning method is used to output a set of cross-platform implicit path sequences; S5: Fuse the cross-platform observability score set and the cross-platform implicit path sequence set to construct a complete sequence of reflux paths; S6: Perform revenue event correlation calculation and link credibility weight normalization on the complete sequence of the return path, and output the closed-loop revenue attribution result; S7: Feedback the closed-loop revenue attribution results to the cross-platform delivery strategy database to complete closed-loop optimization.

[0019] S1: Collect original multi-platform interaction data and generate a cross-platform interaction dataset based on a unified timestamp and cross-platform business identifier, including: Collect raw multi-platform interaction data including ad click event data, product browsing event data, and order settlement event data; Specifically, ad click event data refers to data generated by users clicking on ads on advertising platforms. For example, when a user sees an e-commerce ad on a search engine or social media and clicks on it, the platform automatically records the time the user clicked on the ad, the user's device type, the ad content, and the ad delivery channel identifier. Product browsing event data refers to data generated when users browse products on e-commerce trading platforms, including the time the user visited the product page, the duration of the user's stay on the product page, user page interaction records (such as clicks, saves, and adding to cart), user device type, and browsing environment information. Order settlement event data refers to data generated after users complete the actual purchase and payment of products on the order management platform, including order submission time, order confirmation time, order payment completion time, user payment method, order amount, order number, item purchase quantity, and item type. Ad click event data, product browsing event data, and order settlement event data originate from different platforms. For example, ad click event data primarily originates from online advertising platforms, product browsing event data primarily originates from e-commerce trading platforms, and order settlement event data originates from internal or third-party order management platforms.

[0020] According to the cross-platform unified timestamp standard, the original multi-platform interaction data is uniformly timestamped; Specifically, a cross-platform unified timestamp standard uses a standardized time format to mark raw data from different platforms, allowing for unified data alignment. Standard time, such as Coordinated Universal Time or Beijing Time, is used, down to the second or millisecond level, and expressed in a fixed format, such as "year-month-day hour:minute:second:millisecond," providing a consistent time reference for cross-platform data merging, sorting, and analysis.

[0021] Based on the unified cross-platform business identification rules, the original multi-platform interaction data is marked with business identification; Specifically, the unified cross-platform business identification rule means that all original interactive data on each platform uses a unified and standardized business identifier to uniformly identify the same business scenario occurring across platforms. The unified cross-platform business identification rule includes defining a unified business type identifier and a unified business scenario identifier. For example, all advertising click event data is uniformly identified as "ADV_CLICK", product browsing event data is uniformly identified as "ITEM_BROWSE", and order settlement event data is uniformly identified as "ORDER_SETTLEMENT". During promotional activities, the data participating in the promotion on all platforms is uniformly identified as "PROMO2024_01", thereby ensuring that cross-platform data can be accurately matched in business scenarios.

[0022] Match the time series and business logic of the original multi-platform interaction data based on the unified timestamp and business identification tags to generate a cross-platform interaction dataset; Specifically, matching time series with business logic involves combining data from different platforms based on time and business dimensions, and organizing them on a user-by-user basis to form a cross-platform interaction dataset. For example, after a user clicks an ad on an advertising platform, enters an e-commerce trading platform to browse products, and ultimately completes the purchase and settlement of an order, then, under the premise of unified timestamp marking and unified business identification, the user's cross-platform interaction dataset will contain ad click events, product browsing events, and order settlement events sorted in that order. Ad click events, product browsing events, and order settlement events are arranged in a unified timestamp order, and a unified business identification is used as the basis for event identification, thus reflecting the user's complete interaction chain across platforms and events.

[0023] S2: Based on the cross-platform interaction dataset, a multi-layer anonymous identity mapping algorithm is executed to output a unified user link identity sequence and event sequence mapping table, including: The multi-layer anonymous identity mapping algorithm includes device-level anonymous identity mapping, account-level anonymous identity mapping, and session-level anonymous identity mapping; Specifically, in the closed-loop analysis of cross-platform revenue data, the application of a multi-layer anonymous identification mapping algorithm is to process user data from multiple different platforms through multi-level, multi-dimensional anonymous identification processing, so that the associations between user data on different platforms can be effectively identified and analyzed. The multi-layer anonymous identification mapping algorithm includes device-level anonymous identification mapping, account-level anonymous identification mapping, and session-level anonymous identification mapping, which correspond to the device layer, user account layer, and user access session layer in the cross-platform data closed-loop analysis process, respectively. For example, a user clicks on an ad on an advertising platform, browses products on an e-commerce trading platform, and completes an order payment on an order management platform. During the cross-platform interaction process, the user involves device information, account information, and specific access behavior (i.e., session information). In order to conduct unified analysis and attribution of user data across multiple platforms, it is necessary to map the user's anonymous identity through a multi-layer anonymous identification mapping method to ensure the accuracy of data association analysis and the protection of user privacy.

[0024] Device-level anonymous identification mapping uses the device identification data of each platform in the cross-platform interaction dataset to perform unified mapping and output a device-level anonymous identification sequence; Specifically, this approach utilizes device-level data from cross-platform interaction data collection, including user terminal device information recorded across different platforms, such as mobile device IDs, browser cookie identifiers, terminal hardware identifiers (e.g., IMEI numbers), and network identifiers (e.g., IP addresses). This data is then anonymized using specific mapping rules to generate a unified, cross-platform device-level anonymous identifier sequence. For example, the advertising platform records the user's mobile device IMEI number as "867891234567890," the e-commerce transaction platform records the user's browser cookie identifier as "cookie_xyz12345," and the order management platform records the user's IP address as "123.45.67.89." Using a device-level anonymous identifier mapping algorithm, the different device identifiers from multiple platforms are mapped using hash encryption mapping rules or other device identifier anonymization algorithms to ultimately generate a unified device-level anonymous identifier sequence, such as "DEVICE_ANON_001." This anonymous identifier sequence represents the same terminal device across platforms, effectively establishing cross-platform user data association at the device level.

[0025] Account-level anonymous identifier mapping uses the account identifier data of each platform in the cross-platform interaction dataset to perform unified mapping and output an account-level anonymous identifier sequence; Specifically, anonymous mapping is performed using the user account information recorded on each platform in the cross-platform interaction dataset. Account identification data includes the account ID, username, membership ID, or platform user number used by users to register or log in on each platform. For example, the user account ID is recorded as "ad_user_567" on the advertising platform, "buyer_user_890" on the e-commerce trading platform, and "order_user_123" on the order management platform. Account identification data is mapped using cross-platform account-level anonymous mapping rules, such as secondary mapping based on the device-level anonymous identification sequence, to generate a unified account-level anonymous identification sequence, such as "ACCOUNT_ANON_001," thereby achieving unified cross-platform data association and identification at the account level.

[0026] The session-level anonymous identifier mapping uses the session identifier data of each platform in the cross-platform interaction dataset to perform unified mapping and output a session-level anonymous identifier sequence; Specifically, on the basis of account-level anonymous identification mapping, the user session data recorded by each platform in the cross-platform interaction data set is used for mapping. Session identification data includes independent session identifications formed during a user's continuous visit to a single platform, such as session access tokens, session numbers, or temporary access identifications. For example, when a user clicks an ad on an advertising platform, the session number generated by the platform is "ad_session_345", when a user visits an e-commerce trading platform once, the session identification recorded by the platform is "shop_session_789", and when a user completes a payment on an order management platform, the session identification is "order_session_456". A cross-platform session-level anonymous mapping algorithm is used, based on the account-level anonymous identification sequence, to uniformly map the session-level identification and generate a session-level anonymous identification sequence, such as "SESSION_ANON_001", so that user session data in a cross-platform environment can be associated and analyzed across platforms using a unified standard.

[0027] All identification sequences are sorted based on the cross-platform unified timestamp tag to generate a unified user link identification sequence; Specifically, the device-level anonymous identification sequence, account-level anonymous identification sequence, and session-level anonymous identification sequence are sorted in combination with the cross-platform unified timestamp tag, and the sorting rules are arranged in order from earliest to latest according to the unified timestamp. For example, a user first clicks on an ad on the advertising platform using the device-level anonymous identification sequence "DEVICE_ANON_001", then browses products on the e-commerce trading platform using the account-level anonymous identification sequence "ACCOUNT_ANON_001", and finally completes the order settlement on the order management platform using the session-level anonymous identification sequence "SESSION_ANON_001". The unified user link identification sequence records the order of the user's cross-platform anonymous identification data in the above order, presenting the chronological relationship of the cross-platform user behavior link.

[0028] Establish an event sequence mapping table between a unified user link identification sequence and a cross-platform interaction dataset; Specifically, the unified user link identifier sequence already reflects the temporal relationship of cross-platform user behavior. The unified user link identifier sequence is combined with the cross-platform interaction dataset to construct an event sequence mapping table, that is, each user link identifier in the unified user link identifier sequence corresponds to a specific event in the cross-platform interaction dataset. For example, "DEVICE_ANON_001" corresponds to an ad click event, "ACCOUNT_ANON_001" corresponds to a product browsing event, and "SESSION_ANON_001" corresponds to an order settlement event. The above events are mapped to user link identifiers to form a one-to-one event sequence mapping relationship table, ensuring that user cross-platform behavior data can be attributed and uniformly analyzed through mapping relationships.

[0029] S3: Calculate the observability matrix of each platform’s event chain based on the unified user link identifier sequence, and output a cross-platform observability score set, including: From the cross-platform interaction dataset, event sequence groups are extracted based on the unified user link identifier sequence and platform dimension. Specifically, to perform platform-specific data analysis, we need to extract each user's event data on different platforms from the cross-platform interaction dataset based on the unified user link identifier sequence, forming event sequence groups for each platform dimension. For example, if the unified user link identifier sequence contains an anonymous identifier "USER_LINK_0001," the corresponding event sequence groups for that user are: ad click event data on the advertising platform (e.g., "2024-01-15 10:00:01 user clicked ad event"); product browsing event data on the e-commerce trading platform (e.g., "2024-01-15 10:03:12 user browsed product details page event"); and order settlement event data on the order management platform (e.g., "2024-01-15 10:15:32 user submitted order payment event"). For all users in the cross-platform interaction dataset, we extract event sequence groups based on this logic.

[0030] For each platform’s event sequence group, an event integrity matching method is applied to determine the event integrity score; Specifically, the event integrity matching method quantitatively analyzes and evaluates the data integrity of event sequence groups from various platforms to reflect whether the event data is missing, omitted, or incomplete during the data recording process. The implementation of the event integrity matching method includes setting several event integrity evaluation indicators, such as data missing rate, data continuity, and data coverage. The data missing rate refers to the ratio of the number of missing events in the platform event data records to the total number of events that should be recorded. Data continuity refers to the continuity of the platform event sequence group data in the time dimension. Data coverage refers to the coverage of the event data recorded by the platform relative to the expected recorded data range. For example, in an e-commerce trading platform's event sequence group, a user with the unified user link identifier "USER_LINK_ANON_1001" should record 20 product browsing events, but only 18 are actually recorded. Therefore, the data missing rate is (20-18) / 20 × 100% = 10%. If there are time interval gaps in the data recorded within a certain period of time in the platform's event sequence group, the data continuity index decreases. If the recorded data of the user's access behavior covers all expected user access scenarios, the data coverage is high. The event completeness score is calculated and weighted based on the data missing rate, data continuity, and data coverage. Specifically, the data missing rate score is first determined using the formula: Data missing rate score = (1-data missing rate) × 100. For example, when the data missing rate is 10%, the data missing rate score is (1-10%) × 100 = 90. Next, the data continuity score is determined. The data continuity score is determined based on the temporal continuity of the platform's event sequence group using the formula: Data continuity index score = (number of continuously recorded events / total number of events that should be recorded) × 100. For example, if the event sequence group of an e-commerce trading platform should record 20 events continuously, but only 18 events are actually recorded continuously, the data continuity index score is (18 / 20) × 100 = 90 points.

[0031] Then determine the data coverage index score. The data coverage index score represents the ratio of the user behavior scenario range covered by the actual recorded data to the expected user behavior scenario range. The calculation formula is: Data coverage index score = (actual number of recorded behavior scenarios / total number of expected behavior scenarios) × 100. For example, if a user actually recorded 18 behavior scenarios and the total number of expected behavior scenarios was 20, the data coverage index score would be (18 / 20 × 100) = 90 points.

[0032] Finally, based on the scores of the above three indicators, the event integrity score is weighted and calculated using the following weight coefficients: Event completeness score = data missing rate score × weight coefficient α + data continuity score × weight coefficient β + data coverage score × weight coefficient γ. α, β, and γ are 0.4, 0.3, and 0.3, respectively, and the sum of the weight coefficients is 1.

[0033] The event completeness score for the platform event sequence group is: Event Completeness Score = 90 × 0.4 + 90 × 0.3 + 90 × 0.3 = 36 + 27 + 27 = 90 points. A higher event completeness score indicates a more complete event record, while a lower score indicates an incomplete event record.

[0034] For each platform's event sequence group, apply data visibility assessment rules to determine a data visibility score; Specifically, data visibility assessment rules include evaluation indicators such as the platform's data access authorization level, data openness scope, and data transparency. For example, the platform's data access authorization level refers to the level of control over data access rights in cross-platform data analysis methods. Different platforms may have different degrees of data access restrictions, with some platforms allowing full access to user data and others allowing only partial access. The data openness scope refers to the range of data that the platform allows external systems to access or call. Data transparency refers to whether the platform's data provides sufficiently detailed records to facilitate cross-platform analysis. The data visibility score is determined through quantitative calculations based on the above evaluation indicators. Specifically, the data access authorization level indicator adopts a three-level scoring system: high, medium, and low, with a high score of 100 points, a medium score of 70 points, and a low score of 40 points. The data openness scope indicator is divided into three levels: fully open, partially open, and restricted open, with scores of 100 points, 60 points, and 30 points, respectively. The data transparency indicator is divided into three levels: detailed records, general records, and rough records, with scores of 100 points, 75 points, and 50 points, respectively. Finally, the data visibility score is calculated according to the following formula: Data Visibility Score = (Data Access Authorization Level Score × Weight Coefficient a) + (Data Openness Scope Score × Weight Coefficient b) + (Data Transparency Score × Weight Coefficient c). The weight coefficients a, b, and c are 0.4, 0.3, and 0.3, respectively, and the sum of the weight coefficients is 1.

[0035] For example, if an advertising platform's data access authorization level is high (100 points), its data openness is fully open (100 points), and its data transparency is detailed records (100 points), then the data visibility score is: Data Visibility Score = 100 × 0.4 + 100 × 0.3 + 100 × 0.3 = 100 points. A higher data visibility score means better platform data visibility and greater usability for data analysis.

[0036] Based on the event completeness score and data visibility score, an event chain observability score is generated according to the platform order; Specifically, the event chain observability score is a single score calculated by combining the event integrity score and the data visibility score. The calculation method is: first determine the weight coefficients of the event integrity score and the data visibility score of each platform, for example, the event integrity score weight coefficient is 0.6, and the data visibility score weight coefficient is 0.4; calculate according to the following formula, for example, event chain observability score = event integrity score × 0.6 + data visibility score × 0.4. For example, if the event integrity score of an advertising platform is 80 and the data visibility score is 100, then the observability score of the advertising platform's event chain is calculated as 80 × 0.6 + 100 × 0.4 = 88. If the event integrity score of an e-commerce trading platform is 100 and the data visibility score is 70, then the observability score of the e-commerce trading platform's event chain is calculated as 100 × 0.6 + 70 × 0.4 = 88. If the event integrity score of an order management platform is 90 and the data visibility score is 80, then the observability score of the order management platform's event chain is 90 × 0.6 + 80 × 0.4 = 86. Using the above method, we can generate the observability score of each platform's event chain, facilitating cross-platform analysis and comparison.

[0037] The observability scores of each platform's event chain are organized according to the platform dimension to form an event chain observability matrix; Specifically, the event chain observability matrix is ​​a two-dimensional data structure organized according to the platform dimension. The rows of the matrix represent different unified user link identifier sequences, and the columns represent different platforms, namely, the advertising platform, the e-commerce trading platform, and the order management platform. Each cell in the matrix represents the calculated event chain observability score for each user on each platform. For example, if the unified user link identifier sequence "USER_LINK_0001" has an event chain observability score of 88 on the advertising platform, 88 on the e-commerce trading platform, and 86 on the order management platform, the data elements in the event chain observability matrix, corresponding to the user, are populated with (88, 88, 86) in sequence. Similarly, the platform event chain observability scores corresponding to other users in the matrix are also populated according to the above rules, forming the event chain observability matrix.

[0038] Extract matrix elements from the event chain observability matrix and aggregate them according to the unified user link identifier to output a cross-platform observability score set; Specifically, each row of the event chain observability matrix (i.e., each user's event chain observability score across platforms) is aggregated using a unified user link identifier as an index to form a cross-platform observability score set. For example, the cross-platform observability score set corresponding to the unified user link identifier sequence "USER_LINK_0001" is (Advertising Platform: 88 points, E-commerce Platform: 88 points, Order Management Platform: 86 points). All user data is aggregated into a cross-platform observability score set using the same logic, thus enabling input data preparation for closed-loop cross-platform data analysis.

[0039] S4: Based on the unified user link identifier sequence, a probabilistic graph reasoning method is used to output a set of cross-platform implicit path sequences, including: Based on the unified user link identification sequence and cross-platform interaction dataset, an initial directed graph of cross-platform user behavior paths is constructed; Specifically, each user link identifier in the unified user link identifier sequence is first used as a node in the initial directed graph of the cross-platform user behavior path, and the specific events occurring in the cross-platform interaction dataset corresponding to each user link identifier (such as advertising click events, product browsing events, order settlement events) are used as event nodes in the initial directed graph of the cross-platform user behavior path, and the behavior transfer relationship of users between different platforms or within the same platform is used as the directed connection relationship between nodes. For example, the unified user link identifier sequence contains the identifier "USER_LINK_10001." The events recorded in the cross-platform interaction dataset corresponding to the user link identifier "USER_LINK_10001" include: the "user clicked on ad A at 2024-02-01 09:10:00" event recorded on the advertising platform; the "user browsed the details page of product B at 2024-02-01 09:12:30" event recorded on the e-commerce trading platform; and the "user completed payment for order C at 2024-02-01 09:20:15" event recorded on the order management platform. By establishing directional connections between these three event nodes in the order in which they actually occurred (i.e., ad click - product browsing - order payment), an initial directed graph of the cross-platform user behavior path of user "USER_LINK_10001" is formed. In this way, a corresponding initial directed graph is established for all user data in the cross-platform interaction dataset.

[0040] The user behavior nodes recorded in the initial directed graph of the cross-platform user behavior path are defined as explicit nodes; Specifically, user behavior nodes recorded in the initial directed graph of cross-platform user behavior paths are nodes in the cross-platform interaction dataset that record the time, location, and specific content of the event. For example, in the initial directed graph of cross-platform user behavior paths for user "USER_LINK_10001," the ad click node records "clicked on Ad A at 2024-02-01 09:10:00," the product browsing node records "browsed the details page of Product B at 2024-02-01 09:12:30," and the order payment node records "paid Order C at 2024-02-01 09:20:15." All three nodes have complete and detailed records in the cross-platform interaction dataset, so these nodes are defined as explicit nodes. In the initial directed graph of cross-platform user behavior paths, explicit nodes are typically recorded by the platform and provide detailed event information, which can be used for path analysis and attribution calculations.

[0041] The platform nodes that do not record user behavior data in the initial directed graph of cross-platform user behavior paths are defined as hidden nodes; Specifically, hidden nodes in the initial directed graph of cross-platform user behavior paths refer to platform nodes that fail to record specific event data in detail or lack sufficient detailed information in the cross-platform interaction dataset. For example, the cross-platform interaction dataset of user "USER_LINK_10001" records the user's ad-clicking behavior on the advertising platform and order payment behavior on the order management platform, but lacks information on the user's product browsing behavior on the e-commerce trading platform, or there is only access to the e-commerce trading platform but no detailed records of the product browsing behavior details. In this case, the node corresponding to the e-commerce trading platform is defined as a hidden node because it does not record complete user behavior data. Hidden nodes in the initial directed graph of cross-platform user behavior paths often appear as nodes where platform event records are missing or specific data is not provided. Specific inference methods are needed to complete the data in order to restore the integrity of the user's cross-platform behavior path.

[0042] Based on the event context and event link structure between explicit and implicit nodes, the probabilistic graph reasoning method is used to infer the existence and location of implicit nodes and generate an implicit node inference sequence. Specifically, probabilistic graph reasoning methods infer unknown nodes based on known node information and the probabilities of directed paths between them. This method is suitable for inferring the existence and specific locations of hidden nodes based on the event link structure and sequential occurrence relationship between explicit and implicit nodes in the initial directed graph of cross-platform user behavior paths. First, the start and end event data of the user's cross-platform behavior path are extracted from the explicit nodes. For example, the explicit nodes for user "USER_LINK_10001" are the ad click node (starting node) and the order payment node (ending node), while the intermediate implicit nodes are the product browsing behavior nodes on the e-commerce trading platform. The probabilistic graph reasoning method uses historical data on user cross-platform behavior paths to determine the probability of a user clicking an ad on an advertising platform, then browsing products on an e-commerce platform, and then completing payment on an order management platform. For example, historical data shows a 90% probability for the "ad click-product browsing-order payment" path, and a 10% probability for the "ad click-direct order payment" path. Based on these path probabilities, it can be inferred that user "USER_LINK_10001" has a hidden node (the e-commerce platform browsing behavior node) located between the ad click and order payment nodes. This method is used to infer all user path data with hidden nodes, generating an inferred sequence of hidden nodes that includes their locations.

[0043] Determine a set of cross-platform implicit path sequences based on implicit node inference sequences; Specifically, the implicit node inference sequence is generated by inferring the position and existence of implicit nodes in the initial directed graph of cross-platform user behavior paths using probabilistic graph reasoning methods. The cross-platform implicit path sequence set is the complete set of cross-platform user behavior paths determined based on the implicit node inference sequence. The cross-platform implicit path sequence is formed based on the implicit node positions indicated in the implicit node inference sequence and combined with the explicit nodes. For example, the implicit node inference sequence for user "USER_LINK_10001" indicates the presence of the implicit node "e-commerce platform product browsing event" between the user's ad click event and order payment event. Therefore, the user's cross-platform implicit path sequence is: "advertising platform ad click event - e-commerce platform product browsing event - order management platform order payment event." The same process is performed on the implicit node inference sequences corresponding to all unified user link identifier sequences, ultimately forming a set of cross-platform implicit path sequences that encompasses all users' complete cross-platform paths.

[0044] S5: Fuse the cross-platform observability score set with the cross-platform implicit path sequence set to construct a complete sequence of return paths, including: According to the observability score of each platform event chain in the cross-platform observability score set, a credibility value is assigned to each cross-platform implicit path sequence in the cross-platform implicit path sequence set; Specifically, for each cross-platform implicit path sequence in the set of cross-platform implicit path sequences, the platforms involved in the path are identified; the corresponding event chain observability scores of the platforms involved in the path are extracted from the cross-platform observability score set; and the overall credibility weighted score of the path is calculated based on the platform's event chain observability score. The credibility weighted score calculation formula is as follows: Cross-platform implicit path sequence credibility assignment = Platform A event chain observability score × weight coefficient W_A + Platform B event chain observability score × weight coefficient W_B + Platform C event chain observability score × weight coefficient W_C. The weight coefficients W_A, W_B, and W_C represent the credibility weights of the corresponding platforms in the path, and the sum of the weight coefficients is 1. The weight coefficient is calculated as follows: Based on statistical analysis of historical data, evaluate the average weight of the contribution of user behavior on different platforms to the final revenue. For example, historical data shows that the average weight of the contribution of user behavior on the advertising platform to the final order conversion is 0.3, the average weight of the contribution of user behavior on the e-commerce transaction platform is 0.4, and the average weight of the contribution of behavior on the order management platform is 0.3. Then the weight coefficient W_A is 0.3, W_B is 0.4, and W_C is 0.3. For example, for the cross-platform implicit path sequence of a user "USER_LINK_1001" in the unified user link identification sequence: "advertising platform ad click event-e-commerce transaction platform product browsing event-order management platform order payment event", assuming that the observability score of the advertising platform event chain is 90 points, the observability score of the e-commerce transaction platform event chain is 85 points, and the observability score of the order management platform event chain is 95 points, then the credibility assignment of the cross-platform implicit path sequence is: 90 points × 0.3 + 85 points × 0.4 + 95 points × 0.3 = 27 + 34 + 28.5 = 89.5 points.

[0045] Through the above method, credibility is assigned to all paths in the cross-platform implicit path sequence set, thereby generating a credibility score for each path, which is used for path sorting and explicit node fusion.

[0046] After credibility is assigned, the paths of each cross-platform implicit path sequence are sorted according to the unified user link identification sequence; Specifically, the cross-platform implicit path sequences, after credibility has been assigned, are aggregated and classified at the user level based on the unified user link identifier sequence to ensure that each user's corresponding path is clearly attributed. The multiple cross-platform implicit path sequences corresponding to each user are then sorted in descending order based on the path credibility assigned, with the path with the highest credibility assigned first, the path with the second highest credibility assigned second, and so on. After the path sorting is completed, the order relationship between the cross-platform implicit path sequences corresponding to each user is determined. For example, for user "USER_LINK_1001" in the unified user link identifier sequence, suppose there are two cross-platform implicit path sequences. The first path, "advertisement click event - product browsing event - order payment event," has a credibility assigned of 89.5 points, and the second path, "advertisement click event - order payment event" (which does not explicitly record product browsing), has a credibility assigned of 75.5 points. After sorting, the first path is ranked first, and the second path is ranked last, thus determining the user path order. All user paths are processed using this sorting method to form a cross-platform implicit path sequence.

[0047] The path sorting results are merged with the explicit nodes recorded in the cross-platform interaction dataset to form a complete sequence of reflux paths containing explicit nodes and implicit nodes; Specifically, for each sorted cross-platform implicit path sequence, the implicit and explicit node positions contained therein are identified. The explicit node event data recorded in the cross-platform interaction dataset is then matched one-to-one with the explicit node positions in the corresponding path. Based on the implicit node positions, the inferred implicit node event data is inserted between the explicit nodes according to the path sorting results, forming a cross-platform behavioral path sequence, i.e., a complete reflux path sequence. For example, for user "USER_LINK_1001," the sorted cross-platform implicit path sequence is: "Ad click event (explicit node) - Product browsing event (implicit node) - Order payment event (explicit node)." The cross-platform interaction dataset records detailed information (time, specific operation, platform identifier, etc.) for the ad click and order payment events. The product browsing event is only an inferred implicit node event because platform data is missing or undisclosed. Therefore, the merging process is as follows: insert the complete data of the explicit ad click node and the explicit order payment node. Then, based on the implicit node inference results, insert the inferred product browsing event information (such as the event platform and inferred occurrence time) between the explicit nodes. The final result is the complete reflux path sequence for user "USER_LINK_1001": "2024-02-01 09:10:00 ad click event on the advertising platform - 2024-02-01 09:12:30 (inferred) product browsing event on the e-commerce trading platform - 2024-02-01 09:20:15 order payment event on the order management platform." All user path sequences in the cross-platform implicit path sequence set are merged with the explicit nodes in this way, ultimately forming a complete reflux path sequence set containing both explicit and implicit nodes.

[0048] S6: Perform revenue event correlation calculation and link credibility weight normalization on the complete sequence of the return path, and output closed-loop revenue attribution results, including: Determine the association relationship between the explicit nodes and implicit nodes in the complete sequence of the return path and the order settlement event data according to the unified user link identification sequence; Specifically, the association relationship determination process is as follows: based on the unified user link identification sequence, the corresponding complete sequence of the return path and order settlement event data are extracted for each user respectively; each complete sequence of the return path is analyzed to identify the specific event content and timestamps of the explicit nodes and implicit nodes in the sequence; for each order settlement event data, the event timestamps and event content of each node in the path are compared respectively to determine the causal relationship and temporal relationship between each explicit node and implicit node and the specific order settlement event data. Take a user in the unified user link identification sequence as an example: the complete sequence of the reflux path corresponding to the user includes the following event nodes: the ad click event on the advertising platform occurred at 10:0:00 on February 1, 2024, the product browsing event on the e-commerce trading platform (hidden node, the event occurrence time is determined by probabilistic inference) occurred at 10:03:30 on February 1, 2024, and the order payment event on the order management platform occurred at 10:10:00 on February 1, 2024. The order settlement event data corresponding to the user is the order number "ORD20240201A", the order generation time is 10:08 on February 1, 2024, and the payment completion time is 10:10:00 on February 1, 2024. The order payment event node in the complete sequence of the user return path is exactly the same as the payment completion time in the order settlement event data, so there is a direct temporal correlation. The product browsing event occurred at 10:03:30 on February 1, 2024, before the order generation time of 10:08:00 on February 1, 2024, so there is a causal relationship. The ad click event occurred even earlier at 10:00:00 on February 1, 2024, and is also causally related to the order. The above method is applicable to the association analysis and processing of all user data in the unified user link identification sequence, and forms the association relationship between the explicit and implicit nodes in each complete sequence of the return path and the order settlement event data.

[0049] According to the association relationship, the initial benefit weight values ​​are assigned to the explicit nodes and implicit nodes in the complete sequence of the return path; Specifically, the initial revenue weight represents the contribution each node makes to the final order conversion revenue. Initial revenue weights are assigned by statistically analyzing the average contribution ratio of each node to order conversion based on historical data, generating historical contribution rate data. For example, taking the ad click node as an example, historical statistics show that the average conversion rate after an ad click event is 20%, so the initial revenue weight for the ad click node is set to 0.2. For the product browsing node on an e-commerce platform, the average conversion contribution ratio is 30%, so the initial revenue weight is 0.3. The order payment node, due to its direct connection to order conversion, has an even higher contribution ratio of 0.5. For the complete sequence of user return paths within a unified user link identifier sequence, the initial revenue weight assignment results are as follows: the ad click event node is assigned an initial revenue weight of 0.2, the product browsing node is assigned an initial revenue weight of 0.3, and the order payment node is assigned an initial revenue weight of 0.5. This method is used to assign initial revenue weights to all user path nodes.

[0050] Based on the observability score of the event chain, the initial benefit weight values ​​of the explicit nodes and implicit nodes are respectively modified for credibility; Specifically, the initial revenue weight is determined based on the average contribution ratio of historical data. However, historical statistical data does not consider the impact of platform data observability. Therefore, in the actual revenue attribution calculation, the initial revenue weight needs to be adjusted for credibility based on the observability score of each platform's event chain. The credibility adjustment method is: the initial revenue weight is multiplied by the normalization coefficient of the corresponding platform's event chain observability score. The normalization coefficient is calculated as: the platform's event chain observability score divided by the sum of all platform event chain observability scores. Taking the ad click node as an example, if the observability score of the ad delivery platform's event chain is 90, the observability score of the e-commerce transaction platform's event chain is 80, and the observability score of the order management platform's event chain is 85, then the total score is 90 + 80 + 85 = 255. The normalization coefficients for the ad delivery platform are 90 ÷ 255 ≈ 0.353, 80 ÷ 255 ≈ 0.314, and 85 ÷ 255 ≈ 0.333, respectively. The initial revenue weight values ​​are then adjusted for credibility using the normalization coefficients. For example, the adjusted revenue weight for the ad click node is 0.2 × 0.353 = 0.0706, the adjusted revenue weight for the product browsing node is 0.3 × 0.314 = 0.0942, and the adjusted revenue weight for the order payment node is 0.5 × 0.333 = 0.1665. The above method applies to the credibility adjustment of revenue weight values ​​for all nodes.

[0051] Normalize the benefit weight value after credibility correction to generate the normalized link credibility weight result; Specifically, the sum of the credibility-corrected revenue weights of all nodes is used as the normalization benchmark, and the credibility-corrected revenue weights of individual nodes are proportionally adjusted so that their sum equals 1. For example, the sum of the credibility-corrected revenue weights is 0.0706 + 0.0942 + 0.1665 = 0.3313. This is used as the benchmark for normalization. The normalized revenue weight of the ad click node is 0.0706 ÷ 0.3313 ≈ 0.213, the normalized revenue weight of the product browsing node is 0.0942 ÷ 0.3313 ≈ 0.284, and the normalized revenue weight of the order payment node is 0.1665 ÷ 0.3313 ≈ 0.503. Finally, the normalized results of the link credibility weights are obtained.

[0052] Determine the closed-loop benefit attribution results based on the normalized link credibility weights. The normalized link credibility weights directly represent the exact proportion of each node's contribution to the final order revenue. The closed-loop revenue attribution result is the sum of the product of each node's revenue contribution ratio and the order amount. For example, if the order amount is 100 RMB, the revenue contribution of the ad click node is 100 × 0.213 = 21.3 RMB, the revenue contribution of the product browsing node is 100 × 0.284 = 28.4 RMB, and the revenue contribution of the order payment node is 100 × 0.503 = 50.3 RMB. The sum of the revenue of each node is exactly 100 RMB, forming the closed-loop revenue attribution result. Revenue attribution is calculated for all user data to obtain the closed-loop revenue attribution result for all users.

[0053] S7: Feedback closed-loop revenue attribution results to the cross-platform delivery strategy database to complete closed-loop optimization, including: Import closed-loop revenue attribution results into the cross-platform delivery strategy database; Specifically, the cross-platform delivery strategy database contains cross-platform advertising delivery data, product browsing data, order settlement data, and cross-platform revenue data organized according to a unified user link identification sequence; the cross-platform delivery strategy database is a data storage structure organized with a unified user link identification sequence as an index, and contains various cross-platform data. Cross-platform advertising delivery data includes the number of ad exposures, ad clicks, delivery channel identification, ad material type, and click-through conversion cost for each user on the advertising delivery platform; cross-platform product browsing data includes the number of times users browse product pages on e-commerce trading platforms, page dwell time, click behavior, and the number of times users add items to favorites or add to shopping carts; cross-platform order settlement data includes the order number, order amount, order generation time, order payment completion time, and user payment method information generated by users on the order management platform; cross-platform revenue data is a record of comprehensive revenue data generated by user behavior across different platforms in the cross-platform delivery strategy database, and includes correlation information between user behavior and revenue based on statistical analysis of historical data. For example, for user "USER_LINK_1001" in the unified user link identification sequence, the closed-loop revenue attribution results are: RMB 21.3 in revenue from the advertising click node, RMB 28.4 in revenue from the product browsing node, and RMB 50.3 in revenue from the order payment node; the closed-loop revenue attribution results are stored in the cross-platform delivery strategy database through the database import method, with the unified user link identification sequence as the index, and correspond to the user's originally stored advertising delivery data, product browsing data, order settlement data, and historical cross-platform revenue data, forming a unified data storage structure.

[0054] Based on the unified user link identification sequence, the closed-loop revenue attribution results are matched with the cross-platform revenue data in the cross-platform delivery strategy database; Specifically, based on the unified user link identifier sequence, the existing cross-platform revenue data in the database is matched user by user and item by item. For example, the historical cross-platform revenue data for user "USER_LINK_1001" in the database includes: historical revenue contribution from advertising platforms of 20 yuan, historical revenue contribution from product browsing of 25 yuan, and historical revenue contribution from order payments of 55 yuan. After matching these with the newly imported closed-loop revenue attribution results (ad clicks of 21.3 yuan, product browsing of 28.4 yuan, and order payments of 50.3 yuan), the user's cross-platform revenue data is updated to 21.3 yuan for ad clicks, 28.4 yuan for product browsing, and 50.3 yuan for order payments. This process completes the matching of all user data corresponding to the unified user link identifier sequence, ensuring that the data records in the cross-platform delivery strategy database are always consistent with the actual revenue attribution analysis results.

[0055] Based on the data matching results, the cross-platform advertising data and product browsing data delivery strategy are adjusted through the cross-platform delivery strategy database; Specifically, the delivery strategy adjustment involves adjusting the cross-platform advertising budget allocation ratio and the cross-platform product browsing page push frequency in the cross-platform delivery strategy database. The cross-platform delivery strategy database uses the correlation between user behavior and revenue to adjust the advertising budget and product page push strategy. The cross-platform advertising budget allocation ratio adjustment method involves adjusting the advertising budget allocation ratio across channels based on the revenue contribution ratio of each node in the closed-loop revenue attribution results. For example, the closed-loop revenue attribution results for user "USER_LINK_1001" indicate that ad clicks account for 21.3% of revenue, lower than product browsing (28.4%) and order payment (50.3%). To optimize budget effectiveness, the budget allocation ratio for advertising channels is appropriately reduced, with more budget allocated to improving the browsing page user experience and page push frequency, thereby increasing overall conversion rates. The advertising budget allocation ratio is adjusted from 30% to 20%, and the resulting budget savings (10%) are allocated to product browsing page recommendations on the e-commerce platform. The method for adjusting the push frequency of cross-platform product browsing pages is to analyze the revenue contribution ratio generated by users at the product browsing node through the cross-platform delivery strategy database to determine a reasonable page push frequency; for example, for the user "USER_LINK_1001", due to the high revenue contribution of the product browsing node, the push frequency of the product page should be appropriately increased; the page push frequency is adjusted to increase the number of times users browse the product page per day from 10 times to 12 times, in order to increase the user's stay and interaction opportunities, thereby improving the order conversion rate.

[0056] Update the cross-platform delivery strategy database based on the adjusted delivery strategy to complete closed-loop optimization; Specifically, the update process is to update the latest adjustments to the advertising budget ratio and the frequency of product browsing page push back to the database, and form a comparative record with the original data to ensure that the database records the specific parameters of each adjustment to support future long-term data analysis and strategy optimization. For example, the latest advertising budget ratio of 20% and the daily product page push frequency of 12 times for user "USER_LINK_1001" are updated to the cross-platform delivery strategy database; the update process covers all user records in the cross-platform delivery strategy database, and ultimately achieves closed-loop optimization of the delivery strategy based on actual revenue performance. The closed-loop optimization process fully combines cross-platform revenue analysis, user behavior correlation calculation, delivery budget and page recommendation strategy optimization, etc. to form a complete closed-loop system. The system can automatically monitor the actual revenue effect of subsequent cycles and continuously perform iterative optimization.

[0057] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0058] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in the embodiments of this application are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0059] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0060] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0061] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0062] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0063] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0064] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0065] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0066] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A closed-loop analysis method for cross-platform revenue data, characterized in that: The steps include: S1: Collects raw multi-platform interaction data and generates a cross-platform interaction dataset based on a unified timestamp and cross-platform business identifier. The raw multi-platform interaction data includes ad click event data, product browsing event data, and order settlement event data. S2: Based on the cross-platform interaction dataset, a multi-layer anonymous identity mapping algorithm is executed to output a unified user link identity sequence and event sequence mapping table; S3: Calculate the observability matrix of each platform's event chain based on the unified user link identification sequence and output a cross-platform observability score set; S4: Based on the unified user link identification sequence, a probabilistic graph reasoning method is used to output a set of cross-platform implicit path sequences; S5: Fuse the cross-platform observability score set and the cross-platform implicit path sequence set to construct a complete sequence of reflux paths; S6: Perform revenue event correlation calculation and link credibility weight normalization on the complete sequence of the return path, and output the closed-loop revenue attribution result; S7: Feedback the closed-loop revenue attribution results to the cross-platform delivery strategy database to complete closed-loop optimization.

2. A cross-platform revenue data closed-loop analysis method according to claim 1, characterized in that: S1, specifically: Collect raw multi-platform interaction data including ad click event data, product browsing event data, and order settlement event data; According to the cross-platform unified timestamp standard, the original multi-platform interaction data is uniformly timestamped; Based on the unified cross-platform business identification rules, the original multi-platform interaction data is marked with business identification; The original multi-platform interaction data is matched with the time series and business logic according to the unified timestamp tag and business identification tag to generate a cross-platform interaction dataset.

3. A cross-platform revenue data closed-loop analysis method according to claim 2, characterized in that: The multi-layer anonymous identity mapping algorithm includes device-level anonymous identity mapping, account-level anonymous identity mapping, and session-level anonymous identity mapping.

4. A closed-loop analysis method for cross-platform revenue data according to claim 3, characterized in that: S2, specifically: Device-level anonymous identification mapping uses the device identification data of each platform in the cross-platform interaction dataset to perform unified mapping and output a device-level anonymous identification sequence; Account-level anonymous identifier mapping uses the account identifier data of each platform in the cross-platform interaction dataset to perform unified mapping and output an account-level anonymous identifier sequence; The session-level anonymous identifier mapping uses the session identifier data of each platform in the cross-platform interaction dataset to perform unified mapping and output a session-level anonymous identifier sequence; All identification sequences are sorted based on the cross-platform unified timestamp tag to generate a unified user link identification sequence; Establish an event sequence mapping table between a unified user link identification sequence and a cross-platform interaction dataset.

5. A closed-loop analysis method for cross-platform revenue data according to claim 4, characterized in that: S3, specifically: From the cross-platform interaction dataset, event sequence groups are extracted based on the unified user link identifier sequence and platform dimension. For each platform’s event sequence group, an event integrity matching method is applied to determine the event integrity score; For each platform's event sequence group, apply data visibility assessment rules to determine a data visibility score; Based on the event completeness score and data visibility score, an event chain observability score is generated according to the platform order; The observability scores of each platform's event chain are organized according to the platform dimension to form an event chain observability matrix; Extract matrix elements from the event chain observability matrix and aggregate them according to the unified user link identifier to output a cross-platform observability score set.

6. A cross-platform revenue data closed-loop analysis method according to claim 5, characterized in that: S4, specifically: Based on the unified user link identification sequence and cross-platform interaction dataset, an initial directed graph of cross-platform user behavior paths is constructed; The user behavior nodes recorded in the initial directed graph of the cross-platform user behavior path are defined as explicit nodes; The platform nodes that do not record user behavior data in the initial directed graph of cross-platform user behavior paths are defined as hidden nodes; Based on the event context and event link structure between explicit and implicit nodes, the probabilistic graph reasoning method is used to infer the existence and location of implicit nodes and generate an implicit node inference sequence. Determine the set of cross-platform implicit path sequences based on implicit node inference sequences.

7. A closed-loop analysis method for cross-platform revenue data according to claim 6, characterized in that: S5, specifically: According to the observability score of each platform event chain in the cross-platform observability score set, a credibility value is assigned to each cross-platform implicit path sequence in the cross-platform implicit path sequence set; After credibility is assigned, the paths of each cross-platform implicit path sequence are sorted according to the unified user link identification sequence; The path sorting results are merged with the explicit nodes recorded in the cross-platform interaction dataset to form a complete sequence of reflux paths containing explicit nodes and implicit nodes.

8. A closed-loop analysis method for cross-platform revenue data according to claim 7, characterized in that: S6, specifically: Determine the association relationship between the explicit nodes and implicit nodes in the complete sequence of the return path and the order settlement event data according to the unified user link identification sequence; According to the association relationship, the initial benefit weight values ​​are assigned to the explicit nodes and implicit nodes in the complete sequence of the return path; Based on the observability score of the event chain, the initial benefit weight values ​​of the explicit nodes and implicit nodes are respectively modified for credibility; Normalize the benefit weight value after credibility correction to generate the normalized link credibility weight result; The closed-loop benefit attribution result is determined based on the normalized result of the link credibility weight.

9. A closed-loop analysis method for cross-platform revenue data according to claim 8, characterized in that: S7, specifically: Import closed-loop revenue attribution results into the cross-platform delivery strategy database; Based on the unified user link identification sequence, the closed-loop revenue attribution results are matched with the cross-platform revenue data in the cross-platform delivery strategy database; Based on the data matching results, the cross-platform advertising data and product browsing data delivery strategy are adjusted through the cross-platform delivery strategy database; Update the cross-platform delivery strategy database based on the adjusted delivery strategy to complete closed-loop optimization.

Images