Method for automatic clearing and cross-database reconciliation of heterogeneous system data for tourism and travel business scenarios

Abstract

The application provides a heterogeneous system data automatic clearing and cross-database reconciliation method for a travel business scene, comprising: receiving a travel business request, performing operational risk assessment through a reconciliation risk prediction model, and performing joint scheduling control through adaptive traffic peak clipping; based on a pre-constructed travel field ontology, the Schema alignment problem between heterogeneous systems is converted into a constraint satisfaction problem on a factor graph, and the field mapping relationship between the heterogeneous systems is automatically derived; based on the field mapping relationship, the operation causal dependence of the travel business request is tracked using a vector clock causal consistency protocol without a coordinator, and the target heterogeneous system operation is written through direct exchange of difference vector clocks between nodes; the written operation data is subjected to multi-dimensional reconciliation consistency proof, and the actual settlement or transfer of funds is driven, and the automatic clearing is completed. The application realizes cross-system real-time reconciliation and risk prediction, and ensures the strong consistency and auditability of global order data.

Description

Technical Field

[0001] This invention relates to the field of distributed system data consistency and heterogeneous data integration technology, and in particular to a method for automated clearing and cross-database reconciliation of heterogeneous system data for cultural tourism business scenarios. Background Technology

[0002] With the deepening of digital transformation in the cultural and tourism industry, large-scale cultural and tourism platforms typically need to integrate multiple heterogeneous business systems such as ticketing systems, hotel systems, catering systems, financial systems, and merchant systems. However, these heterogeneous business subsystems are geographically unevenly distributed and differ in data models, storage engines, and interface protocols. This not only leads to data fragmentation but also causes semantic gaps, preventing large-scale cultural and tourism platforms from achieving automated data clearing and real-time reconciliation across systems. Consequently, they cannot guarantee business continuity and financial security, thus facing severe technical challenges.

[0003] Currently, when handling high-concurrency data clearing and cross-database reconciliation, existing technologies typically employ a two-phase commit architecture centered on a coordinator to achieve transaction consistency control. This presents a single point of failure risk, as the coordinator node can easily become a system performance bottleneck, failing to support the concurrent processing requirements of high-concurrency scenarios. Simultaneously, network round-trip latency exceeds 200ms, failing to meet the one-second response time requirements of business scenarios, resulting in low system processing efficiency and inability to adapt to the real-time demands of high-concurrency businesses. While using Raft / Paxos consensus to achieve distributed data consistency through leader election and log replication mechanisms, service unavailability occurs during leader election, compromising business continuity. Furthermore, write operations require confirmation from more than half of the nodes, and in cross-regional deployment scenarios, uncontrollable data synchronization latency between nodes easily leads to data consistency conflicts, failing to meet the stability requirements of cross-regional business deployments. While standardizing data through a business middle platform and forcing upstream systems to modify their schema formats to achieve data unification has proven costly due to the rapid changes in the cultural tourism sector and the need to adapt to numerous legacy systems, freezing or modifying the schema is prohibitively expensive. This approach fails to accommodate the rapid iteration and frequent changes in requirements within the cultural tourism industry. Frozen schemas cannot flexibly respond to business adjustments, resulting in low data adaptation efficiency and poor system scalability. When using rule engines for mapping, manual configuration of field correspondences is relied upon. However, as the number of business systems increases and system scale grows exponentially (e.g., 4 heterogeneous systems require 6 mapping rules, 6 systems require 15, and 8 systems require a staggering 28), manual maintenance costs are high, and rule expansion is limited, making it unsuitable for complex business scenarios involving multiple systems and rules. Furthermore, using T+1 batch reconciliation typically relies on post-event discrepancy correction for fund verification, resulting in significant financial risk exposure and a lack of real-time risk warnings. In scenic areas, tourists leave immediately after making purchases; by the time discrepancies are discovered during post-event reconciliation, the tourists have already departed, making recovery difficult and increasing the risk of financial loss.

[0004] The aforementioned shortcomings of existing technologies make it impossible for the system to balance processing efficiency, data consistency, cost control, and risk prevention in scenarios with high concurrency, cross-regional operations, and rapid business iteration. Therefore, there is an urgent need for a new method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural and tourism business scenarios to at least solve some of the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an automated data clearing and cross-database reconciliation method for heterogeneous systems in cultural tourism business scenarios. This method solves the problems of multi-terminal access and data fragmentation, realizes real-time cross-system reconciliation and risk prediction, ensures strong consistency and auditability of order data across the entire domain, and guarantees high availability of business operations.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios, comprising: Receive requests for cultural and tourism services; The operational risk assessment of cultural and tourism business requests is carried out by the reconciliation risk prediction model to obtain the reconciliation risk score. Based on the reconciliation risk score, joint scheduling control is carried out through adaptive traffic peak shaving to obtain the cultural and tourism business requests after adaptive traffic peak shaving scheduling. For cultural and tourism business requests after adaptive traffic peak shaping scheduling, based on a pre-built cultural and tourism domain ontology, the schema alignment problem between heterogeneous systems is transformed into a constraint satisfaction problem on a factor graph. By constructing a constraint propagation network, the field mapping relationship between heterogeneous systems is automatically derived. Based on field mapping relationships, a coordinator-free vector clock causal consistency protocol is used to track the operational causal dependencies of the cultural and tourism business requests, and the operation data corresponding to the cultural and tourism business requests is written in parallel to multiple target heterogeneous systems through direct exchange of differential vector clocks between nodes. Based on the operational data written in the target heterogeneous system, multi-dimensional reconciliation consistency proof is performed, and the actual settlement or allocation of funds is driven according to the reconciliation consistency proof results, thus completing automated clearing.

[0007] Preferably, when the reconciliation risk prediction model assesses the operational risk of requests for cultural and tourism business, the dimensions of analysis and calculation include: the number of target heterogeneous systems, the amount of operation, the merchant's historical dispute rate, and time sensitivity.

[0008] Preferably, joint scheduling control based on reconciliation risk scoring and adaptive traffic shaving includes: Real-time acquisition of the current system load and reconciliation queue depth of heterogeneous systems; The token generation rate of the token bucket is dynamically adjusted based on the current system load and reconciliation queue depth. For high-risk operations with a reconciliation risk score greater than the set emergency threshold, control their priority to borrow future tokens. Low-risk operations with a reconciliation risk score below the set delay threshold are controlled to enter the delay queue.

[0009] Preferably, the field mapping relationship between heterogeneous systems is automatically derived by constructing a constraint propagation network, including: Initialize the belief in field similarity between target heterogeneous systems; Beliefs are propagated through a constraint propagation network, and belief values ​​are updated iteratively. After the beliefs converge, the maximum a posteriori mapping is extracted as the final field mapping relationship.

[0010] Preferably, the field similarity is calculated using a pre-trained graph neural network encoder, including: Extract contextual features of fields for the target heterogeneous system; The contextual features are input into the graph neural network encoder to generate the context-aware feature vector of the current field; Cosine similarity is calculated based on the context-aware feature vector of the current field to determine the semantic similarity score between fields.

[0011] Preferably, the differential vector clock is a 64-bit compressed vector clock, the structure of which includes: the high 16 bits for recording the node ID, the middle 16 bits for recording the logic clock, and the low 32 bits for recording the verification.

[0012] Preferably, when directly exchanging differential vector clocks between nodes, vector clock merging is performed, the difference between the reference clock of the current node and the clock of the node to be merged is calculated, and only the differential portion containing the difference is transmitted during network transmission.

[0013] Preferably, when using a coordinatorless vector clock causal consistency protocol to trace the operational causal dependencies of the cultural and tourism business requests, conflict resolution is also performed, including: Detect and analyze whether there are causal conflicts between operations, and identify the types of causal conflicts obtained; The target cultural and tourism business adjudication rules are invoked based on the type of causal conflict, and an automatic adjudication is performed through the target cultural and tourism business adjudication rules. After the automatic adjudication is completed, a resolution certificate is generated, which includes the operation identifier, the adjudication rules for the target cultural and tourism business, and the processing timestamp.

[0014] Preferably, when performing multi-dimensional reconciliation consistency verification, it includes: A three-tiered reconciliation verification system is established based on causal verification, numerical conservation, and anomaly detection. A three-tiered reconciliation verification system is used to prove the consistency of reconciliation from multiple dimensions and to generate a reconciliation consistency verification report.

[0015] Preferably, a three-layer reconciliation verification system is used to prove the consistency of reconciliation from multiple dimensions, including: Extract all business operations within a preset time period, construct a causal chain based on vector clocks, and verify whether there are orphan write operations without causal correlation. The operation is hashed to construct a Merkle tree, and the numerical balance relationship between the capital inflow, capital outflow and balance change of all target heterogeneous systems is verified. Finally, the Merkle tree root hash is output. The time series features of the operational data are extracted, and the isolated forest algorithm is used to detect abnormal patterns, identifying complex abnormal patterns where the amount of funds is correct but the time series is abnormal.

[0016] The present invention has achieved the following beneficial effects: This invention balances processing efficiency, data consistency, cost control, and risk prevention in high-concurrency, cross-regional, and rapidly iterating business scenarios. Through adaptive traffic shaping, a coordinatorless vector clock causal consistency protocol, and schema alignment, it not only avoids the impact of instantaneous peaks on heterogeneous systems, ensuring rapid response and business continuity in high-concurrency scenarios, but also eliminates the system's dependence on a central node, achieving true decentralized high availability. In cross-regional deployment environments, it reduces data write latency and ensures causal consistency in distributed operations. Simultaneously, it solves the problems of extremely high costs associated with forcibly modifying the schema of old systems, poor system scalability, and the exponential growth of manually configured rules with the number of systems, leading to high maintenance costs and difficulty in adapting to rapidly iterating business. It achieves zero-intrusion integration of heterogeneous systems and automated generation of mapping rules, greatly reducing manual intervention costs and improving data adaptation efficiency and the flexible scalability of the system architecture.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the application.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a scenario topology diagram of the method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios as described in this invention. Figure 2 This is a schematic diagram illustrating the steps of the method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios as described in this invention. Figure 3 This is a schematic diagram of some steps in S2 of the method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios described in this invention. Figure 4 This is a schematic diagram of some steps in S3 of the method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios as described in this invention. Figure 5 This is a schematic diagram of some steps in S4 of the method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios described in this invention. Figure 6 This is a schematic diagram of some steps in S5 of the method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios described in this invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] like Figure 1 and Figure 2 As shown, this embodiment of the invention provides a method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios. The specific steps for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios include: S1, Receive requests for cultural and tourism services.

[0022] In this step, the cultural and tourism business requests are issued by the client terminal, which covers smart ticketing touchpoints such as mini-programs, OTA platforms, turnstiles, and self-service ticketing machines, realizing multi-terminal aggregation, high-concurrency entry, and elastic traffic access, and supporting business scenarios such as sudden ticket grabbing / flash sales, with peak traffic reaching 50,000 QPS.

[0023] S2. The reconciliation risk prediction model is used to assess the operational risk of cultural and tourism business requests, and a reconciliation risk score is obtained. Based on the reconciliation risk score, joint scheduling control is carried out through adaptive traffic shaving to obtain cultural and tourism business requests after adaptive traffic shaving scheduling.

[0024] In this step, load awareness is performed on the received requests for cultural and tourism services, and CPU / connection count is monitored in real time to achieve elastic scaling and reduction threshold control. The joint scheduling control refers to simultaneously considering system load and queue backlog.

[0025] S3. For cultural and tourism business requests after adaptive traffic peak shaving scheduling, based on the pre-built cultural and tourism domain ontology, the schema alignment problem between heterogeneous systems is transformed into a constraint satisfaction problem on the factor graph. By constructing a constraint propagation network, the field mapping relationship between heterogeneous systems is automatically derived.

[0026] In this step, to address the issue of inconsistent data formats across multiple heterogeneous systems in the cultural and tourism sector, automatic alignment is achieved based on cultural and tourism ontology, constraint propagation, and semantic similarity technologies. Specifically, the cultural and tourism ontology defines business concept models for ticketing, hotels, and catering based on OWL2. The heterogeneous systems include various business systems such as ticketing systems (MySQL, order schema), hotel systems (Oracle, reservation schema), catering systems (MongoDB, ordering schema), financial systems (TiDB, settlement schema), and merchant systems (PostgreSQL, revenue sharing schema).

[0027] When constructing the ontology for the cultural tourism domain, a core concept set is predefined, including passenger orders, ticketing entities, and hotel entities. Each concept is then configured with attribute sets and business logic constraints. Inheritance and dependency relationships between core concepts are established, forming a domain ontology knowledge base oriented towards cultural tourism business. Conceptual equivalence mapping relationships between different heterogeneous business systems are defined within the domain ontology knowledge base. Based on the logical constraints and mapping relationships in the domain ontology knowledge base, a constraint propagation algorithm is used to automatically derive the field alignment results between the source and target heterogeneous systems, thereby determining the cultural tourism domain ontology. The business logic constraints include: numerical range constraints, time sequence constraints, and state enumeration constraints. For example, numerical range constraints require the total order amount to be greater than zero; time sequence constraints require the travel date to be greater than or equal to the booking date; and state enumeration constraints require the payment status to be within a preset range of unpaid, paid, refunded, or partially refunded values. Conceptual equivalence mapping relationships include: defining semantic equivalence between the order status field in the ticketing system and the transaction status field in the financial system, guiding data consistency verification during cross-database reconciliation. The domain ontology is digitally represented using the OWL2 format and serves as prior knowledge for the graph neural network encoder, assisting in calculating the semantic similarity of heterogeneous fields.

[0028] S4. Based on the field mapping relationship, the operation causal dependency of the cultural and tourism business request is tracked using a coordinatorless vector clock causal consistency protocol, and the operation data corresponding to the cultural and tourism business request is written in parallel to multiple target heterogeneous systems by directly exchanging differential vector clocks between nodes.

[0029] In this step, the coordinatorless vector clock causal consistency protocol, based on Vector Clock, causal tracing, and semantic resolution technologies, achieves consistency assurance in a distributed system without centralized coordination. The target heterogeneous systems include one or more of the following business systems: ticketing systems (MySQL, order schema), hotel systems (Oracle, reservation schema), catering systems (MongoDB, ordering schema), financial systems (TiDB, settlement schema), and merchant systems (PostgreSQL, revenue sharing schema).

[0030] When using a coordinatorless vector clock causal consistency protocol to trace the causal dependencies of cultural and tourism business requests, the operation type identifier and user identifier of the cultural and tourism business request are identified. Based on the operation type identifier, historical business operations associated with the user identifier and belonging to a preset preceding type are queried within a preset time window. If a historical business operation that meets the conditions is found, the vector clock corresponding to the historical business operation is extracted and added to the dependency list of the current business operation. The dependency list is then bound to the write request of the current business operation, so that the business operation carries a complete distributed causal dependency relationship when executed in heterogeneous systems. The dependency list contains the vector clock of at least one preceding business operation, establishing a logical execution order conforming to the Happens-Before principle among the distributed nodes without a coordinator. When binding the dependency list to the write request of the current business operation to carry a complete distributed causal dependency relationship when the business operation is executed in heterogeneous systems, distributed consistency order tracing across multiple heterogeneous databases is achieved by carrying the causal dependency chain in the request message, eliminating the need for a centralized coordinator's timing adjudication.

[0031] Furthermore, when a temporary system outage or prolonged failure leads to write failure, operations with causal dependencies are automatically retried by attaching them to a distributed persistent message queue. If the number of retries exceeds a threshold, the vector clock dependency of the operation is extracted, automatically triggering a reverse compensation transaction. Orphan write detection at the reconciliation consistency proof layer ensures that the system state during the failure period is eventually restored. Here, a reverse compensation transaction is, for example, executing a refund instruction for an order that has already been paid.

[0032] S5. Based on the operation data written in the target heterogeneous system, perform multi-dimensional reconciliation consistency proof, and drive the actual settlement or transfer of funds according to the reconciliation consistency proof results to complete automated clearing.

[0033] In this step, end-to-end reconciliation consistency is ensured based on causal verification, numerical conservation, and anomaly detection technologies. When the actual settlement or transfer of funds is driven by the reconciliation consistency proof result, the actual settlement or transfer of funds is carried out when the reconciliation consistency proof result is deemed satisfactory, while automated clearing is also performed.

[0034] The above technical solution can balance processing efficiency, data consistency, cost control, and risk prevention in scenarios with high concurrency, cross-regional operations, and rapid business iteration. It avoids the impact of instantaneous peaks on heterogeneous systems, ensures the system's rapid response and business continuity in high-concurrency scenarios, reduces data write latency in cross-regional deployment environments, ensures causal consistency of operations in distributed environments, enables the timely detection and interception of financial risks, minimizes risk exposure, and protects financial security.

[0035] The above technical solution uses adaptive traffic shaping to prevent the coordinator node from easily becoming a system performance bottleneck when there are concurrent processing needs in high-concurrency scenarios. While ensuring that the underlying heterogeneous system and database are not overwhelmed by instantaneous peak traffic, it smooths the system load, fundamentally eliminates single-point performance bottlenecks, and ensures that the system can still meet extremely high response time requirements during peak business periods, greatly improving the robustness and high concurrency carrying capacity of the system.

[0036] The above technical solution transforms the schema alignment problem between heterogeneous systems into a constraint satisfaction problem on a factor graph by using a pre-built ontology for the cultural and tourism sector. It also automatically derives field mapping relationships between heterogeneous systems by constructing a constraint propagation network. This enables the integration of heterogeneous systems to adapt to the rapid iteration and frequent changes in requirements of cultural and tourism business with extremely low maintenance costs when dealing with data fragmentation and semantic gap issues. This greatly improves system scalability and data adaptation efficiency. When a new heterogeneous system is added, it automatically derives mappings of dozens of fields with ticketing and financial systems. At the same time, manual review is only required for abnormal items, without freezing or forcibly modifying the old data model, completely eliminating the dependence on high-frequency manual configuration of mapping rules.

[0037] The above technical solution completely abandons the traditional transaction model that strongly relies on a central node by adopting a coordinatorless vector clock causal consistency protocol and directly exchanging differential vector clocks. It eliminates the single point of failure and centralized latency of traditional 2PC, as well as the service unavailability and uncontrollable cross-regional synchronization latency during the election of consensus algorithms such as Raft and Paxos. While completely eliminating the single point of failure and downtime election time, it significantly reduces the latency loss of cross-regional network communication. This enables large-scale cultural tourism platforms to achieve causal consistency of operations with extremely low latency when deployed in a cross-domain distributed manner, meeting the real-time and stability requirements of high-concurrency businesses.

[0038] The above technical solution transforms post-event discrepancy repair into real-time operational-level verification during business operations by proving consistency through multi-dimensional reconciliation. It accurately captures data anomalies and fund discrepancies during business processes, triggering immediate interception and early warning. This enables the immediate detection and interception of fund risks, avoids the unique problem of fund loss due to staff turnover in cultural and tourism scenic areas, minimizes fund risk exposure, and provides strong technical protection for the fund security of cultural and tourism platforms.

[0039] In one embodiment of the present invention, when the reconciliation risk prediction model assesses the operational risk of requests for cultural and tourism business, the dimensions of analysis and calculation include: the number of target heterogeneous systems, the amount of operation, the merchant's historical dispute rate, and time sensitivity.

[0040] In the above technical solution, the reconciliation risk prediction model, when assessing the operational risk of requests for cultural and tourism business, includes: A risk assessment function is constructed based on the analytical and computational dimensions. When constructing the risk assessment function, dimensional indicators are determined according to the analytical and computational dimensions, and risk weights are determined for each dimensional indicator. Then, the risk weights and corresponding dimensional indicator data are comprehensively summed to obtain the risk assessment function. The dimensional indicators include: cross-system dimension, funding dimension, credit dimension, and time dimension. When determining the relative risk weights for each dimensional indicator, the assessment method is defined and the weight is determined according to the analytical and computational dimensions. For example, the cross-system dimension calculates risk based on the number of heterogeneous systems involved in the operation (i.e., the number of target heterogeneous systems), with a weight of 20. The funding dimension calculates risk based on the amount of the operation, with a weight of 30. The credit dimension is obtained by retrieving the merchant's historical dispute rate, with a weight of 25. The time dimension determines risk based on time sensitivity for near-expiration business (such as orders for next-day travel), with a weight of 25.

[0041] Obtain the attribute characteristics of the business operations to be processed, and conduct dimensional indicator risk assessment based on the attribute characteristics according to the analysis and calculation dimensions through the reconciliation risk prediction model to obtain dimensional indicator risk data.

[0042] The reconciliation risk score is calculated based on the risk assessment function combined with dimensional indicators and risk data analysis.

[0043] The aforementioned technical solution, when assessing operational risks for cultural and tourism business requests using the reconciliation risk prediction model, achieves multi-dimensional comprehensive analysis by considering the number of target heterogeneous systems, operation amounts, merchant historical dispute rates, and time sensitivity. This avoids the mistaken killing of high-value or time-sensitive critical business processes due to reliance on a single system load indicator. It enables subsequent adaptive traffic shaping and joint scheduling control to possess business semantic-level insights. Under extremely high-concurrency and high-pressure environments, it intelligently balances data consistency requirements with system processing performance. Thus, while ensuring high-quality fund security and zero-lag core experience, it completes the automated clearing and cross-database reconciliation of massive heterogeneous data with optimal resource utilization.

[0044] In one embodiment provided by the present invention, such as Figure 3 As shown, joint scheduling control based on reconciliation risk scoring and adaptive traffic shaving includes: S21. Obtain the current system load and reconciliation queue depth of heterogeneous systems in real time.

[0045] In this step, when obtaining the current system load of the heterogeneous system in real time, the real-time load of the current system (current_load) is obtained through load_monitor, and the reconciliation task backlog is obtained to determine the reconciliation queue depth.

[0046] S22. Dynamically adjust the token generation rate of the token bucket based on the current system load and reconciliation queue depth.

[0047] In this step, the token generation rate is related not only to the current system load but also to the reconciliation queue depth. The token generation rate is calculated using the following formula: V = v × α × β; In the above formula, V is the real-time token generation rate; v is the token generation baseline rate; α is the system load coefficient, determined by α=1-a based on the current system load, where a is the current system load. The higher the current system load a, the smaller the system load coefficient α, and the lower the final token generation rate, thus actively tightening traffic and protecting the system from being overloaded; β is the queue backlog factor, determined by β=1 / (1+b / 1000) based on the reconciliation queue depth, where b is the reconciliation queue depth. The more reconciliation tasks are backlogged, the greater the reconciliation queue depth, the smaller the queue backlog factor, and the lower the final token generation rate, thus tightening token issuance and preventing background tasks from piling up and request avalanche.

[0048] S23. For high-risk operations with a reconciliation risk score greater than the set emergency threshold, control their priority to borrow future tokens.

[0049] In this step, the emergency threshold is dynamic, calculated based on the current system load. The higher the load, the lower the emergency threshold. For example, when the current dynamic system load is 90%, the emergency threshold is 28. When a high-risk operation with a reconciliation risk score greater than the set emergency threshold is performed, it is allowed to use a pre-borrowed token for processing.

[0050] S24. Low-risk operations with a reconciliation risk score less than the set delay threshold are controlled to enter the delay queue.

[0051] In this step, when a low-risk operation has a reconciliation risk score below a set delay threshold, a downgrade operation of delayed queuing or circuit breaker rejection is performed based on its risk score. Specifically, the higher the reconciliation risk score, the higher the scheduling priority of the cultural tourism business request when system resources are limited, to ensure real-time consistency of high-value or highly sensitive business operations.

[0052] The aforementioned technical solution abandons the traditional gateway layer's blind rate limiting model that relies solely on QPS (queries per second). It achieves elastic contraction of the total traffic volume through a dynamic token bucket. Simultaneously, through a two-way scheduling strategy of high-risk future-limited allocation and low-risk delayed queuing, it reorganizes and staggers traffic flow. This ensures that when facing unpredictable, massive concurrency surges in cultural and tourism scenarios, it not only maintains the availability baseline of heterogeneous systems and prevents system downtime, but also improves the success rate at the business level. It ensures that high-priority services are not degraded and low-priority services are not lost, adapting to the complex demands of cultural and tourism businesses for high concurrency, high reliability, and refined user experience. By acquiring the current system load and reconciliation queue depth of heterogeneous systems in real time and dynamically adjusting the token generation rate of the token bucket, a state-aware dynamic feedback loop is achieved. This enables on-demand token generation, maximizing system throughput in real time while ensuring the absolute security of heterogeneous systems. By controlling high-risk operations with reconciliation risk scores exceeding a set emergency threshold, priority is given to borrowing future tokens. This prevents core businesses with high time sensitivity or high financial risk from failing due to lack of tokens, reducing the risk of business disruptions and financial disputes. It ensures that even under extreme traffic surges, the most critical and indispensable cultural tourism businesses remain uninterrupted, achieving a leap from extensive traffic throttling to priority protection based on business value. Conversely, by controlling low-risk operations with reconciliation risk scores below a set delay threshold, they are placed in a delay queue. This prevents low-priority tasks from competing with high-priority tasks for valuable system connections and computing resources during peak periods, thus avoiding exacerbating system congestion. This achieves flexible temporal and spatial shifting of business traffic, freeing up critical resources for high-priority requests while ensuring the final completion rate of low-priority businesses.

[0053] In one embodiment provided by the present invention, such as Figure 4 As shown, by constructing a constraint propagation network, the field mapping relationships between heterogeneous systems are automatically derived, including: S31. Initialize the belief on field similarity between target heterogeneous systems.

[0054] In this step, a pre-set graph neural network encoder is used to perform context-aware encoding on the source and target schema fields, and the semantic similarity between each field pair is calculated as the initial belief value. The pre-set graph neural network encoder refers to a pre-trained graph neural network encoder. When training the pre-set graph neural network encoder, the training data usually consists of 1000 manually labeled cross-system field pairs (positive examples: equivalent fields, negative examples: irrelevant fields). The loss function usually uses contrastive loss to bring positive examples closer and push away negative examples. The optimizer uses Adam with a learning rate of 0.001. The evaluation metrics are usually set to AUC-ROC=0.94 and Top-5 accuracy=89%. This optimizes the model parameters of the graph neural network encoder, determines the pre-trained graph neural network encoder, and encodes the schema fields of the newly accessed system in real time using the pre-trained graph neural network encoder, calculates the semantic similarity matrix with the existing system, and drives constraint propagation alignment. Here, belief refers to the confidence level of field equivalence mapping. The belief of field similarity between target heterogeneous systems is initialized by analyzing and calculating the confidence level of the initial field similarity between target heterogeneous systems.

[0055] Furthermore, the specific steps for analyzing and calculating using a pre-trained graph neural network encoder are as follows: Contextual features of fields are extracted for the target heterogeneous system. These contextual features include the current field, fields in the same table, foreign key related fields, co-occurrence relationships between fields, foreign key constraint relationships, and naming similarity relationships.

[0056] The context features are input into a graph neural network encoder to generate a context-aware feature vector for the current field. Specifically, this involves: defining the current target field, fields in the same table, and foreign key associations as nodes based on the context features; defining co-occurrence relationships, foreign key constraints, and naming similarity relationships between fields as edges to construct a field context graph; mapping the field name labels and data types of each node in the field context graph to name embedding vectors and type embedding vectors of preset dimensions, respectively, and then concatenating the features. The name embedding vector has a dimension of 64, and the data type embedding vector has a dimension of 32. The concatenated initial feature vector serves as the input to the first-layer graph attention network; performing graph convolution processing on the concatenated feature vector using a multi-layer graph attention network, aggregating the context features of adjacent nodes through an attention mechanism. The multi-layer graph attention network includes a first-layer graph attention network and a second-layer graph attention network. The first-layer graph attention network performs feature aggregation and generates intermediate-layer features through a non-linear activation function; the second-layer graph attention network further performs semantic refinement on the intermediate-layer features to generate a globally enhanced node feature matrix; and extracting a context-aware field semantic representation vector from the graph processing result based on the node index corresponding to the target field in the graph.

[0057] The cosine similarity is calculated based on the context-aware feature vector of the current field, and the cosine similarity between the two context vectors is obtained. The cosine similarity is then used as the semantic similarity score between the fields.

[0058] S32. Propagate beliefs through a constraint propagation network and iteratively update belief values.

[0059] In this step, the constraint propagation network is used to build a factor graph containing field consistency constraints based on the domain ontology. When propagating beliefs through the constraint propagation network, the messages between each field pair are iteratively updated on the factor graph through the constraint network. This keeps the computational complexity of schema mapping within a range that increases linearly with the number of fields. Logical constraints defined by the domain ontology are used to filter and correct field mapping relationships, ensuring that the derived alignment results conform to business rules. The belief value is then updated based on this message until the algorithm converges.

[0060] S33. After the belief converges, extract the maximum a posteriori mapping as the final field mapping relationship.

[0061] In this step, when extracting the maximum a posteriori mapping as the final field mapping relationship, the maximum a posteriori mapping estimate is extracted from the converged belief values ​​to generate an automatic alignment result between the source schema and the target schema.

[0062] The above technical solution enables automatic derivation of field mapping relationships, eliminating the need for manual configuration in the rule engine and reducing manual maintenance costs from O(N²) to near O(1). When adding heterogeneous systems, efficient alignment can be achieved through self-learning semantic and structural features without the need for manually writing hundreds of mapping rules. By constructing a constraint propagation network, dynamic adaptive capabilities are improved, enabling automatic recalculation of beliefs and derivation of new relationships when interface changes or schema updates occur in business systems such as hotels and ticketing in the cultural tourism scenario. This adapts to the needs of the cultural tourism industry, avoids schema freezing under rapid iteration of cultural tourism businesses, and, with the adoption of context-aware features and structural constraint verification, the derived mapping relationships are more objective and comprehensive than those configured manually. This provides a high-quality data foundation for subsequent automated clearing, reducing reconciliation discrepancies caused by semantic understanding biases from the source and ensuring the security of fund clearing for large cultural tourism platforms.

[0063] In one embodiment of the present invention, the differential vector clock is a 64-bit compressed vector clock, the structure of which includes: the high 16 bits for recording the node ID, the middle 16 bits for recording the logic clock, and the low 32 bits for recording the verification.

[0064] The high 16 bits are used to record the node ID, which can cover the heterogeneous subsystem requirements of ultra-large cultural tourism parks, enabling extremely fast node identity retrieval and spatial hard specification, and providing a standardized index foundation for subsequent rapid addressing and data aggregation. The node ID adopts a hierarchical encoding mechanism. When the number of nodes approaches the 16-bit (65,536) limit, the merchant nodes are divided into multiple autonomous domains according to geographical or logical business. Within each autonomous domain, the 16-bit ID is dynamically recycled and reused. Inter-domain communication is carried out through a globally unique identifier mapping at the gateway layer, thereby ensuring the long-term scalability of the 64-bit compressed structure.

[0065] The 16 bits are used to record the logical clock, limiting the logical clock to 16 bits. Combined with the differential switching mechanism, this compresses the size of causal dependency metadata, enabling the reduction of the load of a single request and the reduction of network latency in cross-regional network transmission, thus meeting the requirement of response time within 1 second.

[0066] The lower 32 bits are used for verification, providing a strong data self-checking capability. This allows the system to determine the reliability of the operation as soon as the clock is read or received, preventing invalid or damaged clock data from entering the conflict resolution process and ensuring the absolute underlying correctness of clearing and reconciliation.

[0067] In the above technical solution, the clock state of the current node is compressed into a 64-bit binary bit sequence by using the high 16 bits to record the node ID, the middle 16 bits to record the logical clock, and the low 32 bits to record the differential vector clock for verification. This avoids excessive network transmission overhead for the vector clock. Here, in the bit sequence, the high 16 bits are typically used to represent the node ID, the middle 16 bits to represent the logical clock, and the low 32 bits to represent the checksum.

[0068] The above technical solution uses a 64-bit compressed vector clock, which can tightly compress the node ID, logical clock, and checksum into a standard 64-bit integer. This reduces the complex vector clock merging and state updates between heterogeneous systems to low-level machine instruction-level bit operations. As a result, the CAS (Compare-And-Swap) primitive at the bottom layer of modern CPUs can be used to perform single-instruction-level atomic replacement of the complete clock state. This eliminates the business layer locking mechanism in high-concurrency environments and improves the causal tracing processing capability of a single node without increasing hardware resources.

[0069] Furthermore, when directly exchanging differential vector clocks between nodes, the node identifiers and corresponding timestamps in the target clock are traversed, and vector clock merging is performed. If the node identifier does not exist in the local reference clock pool, or the target timestamp is greater than the local reference timestamp, the difference between the two is calculated to determine the difference between the reference clock of the current node and the clock of the node to be merged. During network transmission, only the differential part containing the difference is transmitted. The difference and its corresponding node identifier are stored in the differential dataset and returned to achieve incremental synchronization.

[0070] The above technical solution achieves distributed transaction sequence tracking in high-concurrency scenarios by reducing the size of clock metadata carried by a single data write request. Through the differential mechanism, the transmission overhead of the vector clock is reduced from O(N) to a constant level close to O(1), so that the communication bandwidth pressure will not increase exponentially when the cultural tourism platform continuously integrates new heterogeneous subsystems, improving horizontal scalability. At the same time, it can also shorten the serialization, transmission and deserialization time of messages between heterogeneous systems, enabling responses in a very short time, ensuring the real-time performance and high throughput of cultural tourism business. In addition, it can also ensure that the most core causal dependencies can reach consensus among target heterogeneous systems at the lowest cost, providing a timely, accurate and lightweight logical basis for subsequent automated clearing and cross-database reconciliation.

[0071] In one embodiment provided by the present invention, such as Figure 5 As shown, when using a coordinatorless vector clock causal consistency protocol to trace the operational causal dependencies of the cultural and tourism business requests, conflict resolution is also performed, including: S41. Detect whether there is a causal conflict between the analysis operations, and identify the type of causal conflict.

[0072] In this step, when detecting whether there is a causal conflict between operations, it is achieved by comparing the compressed vector clock carried in the operation request. When the vector clock is found to be incomparable, a causal conflict is determined to exist; otherwise, no causal conflict exists. The type of causal conflict is determined according to the business attributes involved in the conflicting operation, such as: duplicate booking conflict, inventory overselling conflict, and price inconsistency conflict.

[0073] S42. Invoke the target cultural tourism business adjudication rule according to the type of causal conflict, and make an automatic adjudication through the target cultural tourism business adjudication rule.

[0074] In this step, automatic adjudication based on the target tourism business adjudication rules involves automatically correcting or retaining data status locally according to these rules. Different types of causal conflicts correspond to different tourism business adjudication rules. When invoking the target tourism business adjudication rule based on the type of causal conflict, the corresponding rule is selected to perform the corresponding automatic adjudication. For example: if the conflict type is a duplicate booking conflict, the operation request with the earlier payment timestamp is retained, and an automatic refund process is triggered for the other operation request; if the conflict type is an inventory overselling conflict, adjudication is based on the order of inventory reservation operations, retaining the operation request with the highest priority; if the conflict type is a price inconsistency conflict, the lower price strategy is implemented, retaining the lower price value and having the platform bear the resulting price difference.

[0075] S43. After completing the automatic adjudication, generate a resolution certificate containing the operation identifier, the adjudication rules for the target cultural and tourism business, and the processing timestamp.

[0076] In this step, after automatic adjudication, if the causal conflict is resolved, a resolution proof is generated, containing the operation identifier, the target cultural tourism business adjudication rule, and the processing timestamp. This resolution proof serves as input data for the multi-dimensional reconciliation consistency verification step, verifying the logical integrity and auditability of the causal conflict resolution process. This reduces the manual intervention costs for financial auditing and dispute resolution, providing solid technical support for the compliance requirements of large-scale cultural tourism platforms. If the causal conflict remains unresolved, the corresponding operation is suspended, and a detailed causal conflict report is generated requesting manual intervention.

[0077] When faced with high-concurrency requests across regions and systems, the aforementioned technical solution can autonomously complete logical alignment and conflict resolution even in network partitioning or extreme conflict situations. It resolves complex business conflicts within the protocol layer, ensuring that data output to the target heterogeneous system is both causally consistent and conforms to the cultural tourism business logic. This improves the success rate of automated clearing and provides an automated defense barrier to ensure business continuity and financial security for the cultural tourism platform. By detecting and analyzing causal conflicts between operations and using vector clocks to trace causal dependencies, it achieves millisecond-level accurate determination of concurrent conflicts, providing a guarantee for subsequent refined processing. This ensures intervention when semantic contradictions arise, improving the concurrent processing efficiency of the distributed system. Automatic adjudication is performed by calling the target cultural tourism business adjudication rules based on the type of causal conflict. Business knowledge of the cultural tourism industry is embedded into the adjudication rule base to achieve intelligent conflict resolution based on business logic. This ensures deep unification of accounting smoothness and business compliance without relying on a centralized coordinator, avoiding bad debts caused by technical conflicts.

[0078] In one embodiment provided by the present invention, such as Figure 6 As shown, multi-dimensional reconciliation consistency verification includes: S51. Establish a three-layer reconciliation verification system based on causal verification, numerical conservation, and anomaly detection.

[0079] In this step, the three-layer reconciliation verification system is used to prove the consistency of reconciliation from multiple dimensions, ensure the continuity of data logic and the overall conservation of funds, and can also detect abnormal situations in a timely manner, avoiding the difficulties in fund tracing in the cultural and tourism scenario under the T+1 reconciliation model, reducing reconciliation delays and ensuring fund security.

[0080] S52. A three-layer reconciliation verification system is used to prove the consistency of reconciliation from multiple dimensions, and a reconciliation consistency proof report is generated.

[0081] In this step, when performing multi-dimensional reconciliation consistency proof through a three-layer reconciliation verification system, causal integrity verification, numerical conservation proof, and abnormal pattern detection are performed respectively. Then, a reconciliation consistency proof report is generated based on the results of causal integrity verification, numerical conservation proof, and abnormal pattern detection.

[0082] This involves a three-tiered reconciliation verification system to demonstrate consistency across multiple dimensions, including: Extract all business operations within a preset time period, construct a causal chain based on vector clocks, and verify whether there are orphan write operations without causal origin.

[0083] In this step, the compressed vector clock status of each node in the heterogeneous system is obtained; the full-link business topology is restored according to the "Happens-Before" principle of the vector clock; operations that are not associated with any business front-end dependency chain are identified as logical anomalies, thereby determining whether there are orphan write operations without causal relationship, and ensuring the continuity of data logic.

[0084] The operation is hashed to construct a Merkle tree, and the numerical balance relationship between the capital inflows, capital outflows and balance changes of all target heterogeneous systems is verified. Finally, the Merkle tree root hash is output.

[0085] In this step, the operations are hashed, and the hash values ​​of all business transaction details are calculated as leaf nodes of a Merkle tree to construct the Merkle tree. Combined with a preset precision threshold, the numerical balance relationship is verified based on the fund inflows, outflows, and balance changes of the target heterogeneous system using the formula |Fund Inflow - Fund Outflow - Balance Change| < preset precision threshold. If the numerical balance relationship is valid, the hash values ​​are aggregated layer by layer in the Merkle tree until a unique root hash is generated. The construction of the Merkle tree reduces the time complexity of traditional step-by-step verification from O(N) to O(logN), where N is the total number of operation records. By aggregating hash values ​​layer by layer to form a tree structure, when a numerical imbalance is detected, it is only necessary to trace down along the imbalanced branch to locate the single or a small number of operation records causing the inconsistency within logarithmic time (logarithmic time complexity), significantly improving the efficiency of reconciliation and location under massive cultural tourism transaction data.

[0086] The time series features of the operational data are extracted, and the isolated forest algorithm is used to detect abnormal patterns, identifying complex abnormal patterns where the amount of funds is correct but the time series is abnormal.

[0087] In this step, the isolated forest algorithm is used for anomaly pattern detection to identify complex anomalies such as operations with correct amounts but conflicting timing, providing a basis for root cause analysis. It also identifies complex logical risks where the amount of funds is correct but the execution timing is abnormal, providing semantically aware real-time auditing capabilities for cultural and tourism businesses.

[0088] The above technical solution avoids missed judgments caused by logical inversion through causal verification, enabling the detection of refund instructions preceding payment instructions and ensuring the legitimacy of business logic. By reusing upstream vector clocks to build a causal chain, it can accurately trace the logical source of each business action and identify orphan write operations without causal connections. This fundamentally eliminates ghost transactions caused by network replay attacks, system vulnerabilities, or out-of-order execution, achieving an upgrade from state consistency to end-to-end operational logic consistency and ensuring the absolute correctness of business timing in a decentralized concurrent environment.

[0089] The above technical solution introduces Merkle tree technology from the blockchain field into cross-database reconciliation in the cultural tourism industry through numerical conservation. This not only provides cryptographic-level tamper-proof proof for the hash calculation of fund flow and balance changes, but also enables the rapid location of specific nodes or specific transactions with O(logN) time complexity when faced with massive transaction records. This breaks through the technical bottlenecks of difficult and time-consuming error detection, and improves reconciliation efficiency and the authority of fund verification.

[0090] The above technical solution introduces the isolated forest algorithm through anomaly detection and extracts time series features to achieve an intelligent leap in risk control mode. This enables advanced risk control capabilities that can capture financial balance but abnormal behavior, break through the limitations of static mathematical rules, and identify hidden malicious attacks or gray and black market behaviors that bypass traditional verification rules from the dynamic time series dimension. This refines the granularity of the risk control defense network to the behavioral pattern level.

[0091] Those skilled in the art should understand that the terms "first" and "second" in this invention merely refer to different application stages.

[0092] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0093] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for automated data clearing and cross-database reconciliation in heterogeneous systems for cultural tourism business scenarios, characterized in that, include: Receive requests for cultural and tourism business; The operational risk assessment of cultural and tourism business requests is carried out by the reconciliation risk prediction model to obtain the reconciliation risk score. Based on the reconciliation risk score, joint scheduling control is carried out through adaptive traffic peak shaving to obtain the cultural and tourism business requests after adaptive traffic peak shaving scheduling. For cultural and tourism business requests after adaptive traffic peak shaping scheduling, based on a pre-built cultural and tourism domain ontology, the schema alignment problem between heterogeneous systems is transformed into a constraint satisfaction problem on a factor graph. By constructing a constraint propagation network, the field mapping relationship between heterogeneous systems is automatically derived. Based on field mapping relationships, a coordinator-free vector clock causal consistency protocol is used to track the operational causal dependencies of the cultural and tourism business requests, and the operation data corresponding to the cultural and tourism business requests is written in parallel to multiple target heterogeneous systems through direct exchange of differential vector clocks between nodes. Based on the operational data written in the target heterogeneous system, multi-dimensional reconciliation consistency proof is performed, and the actual settlement or allocation of funds is driven according to the reconciliation consistency proof results, thus completing automated clearing.

2. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 1, characterized in that, When assessing operational risks for requests related to cultural and tourism businesses, the reconciliation risk prediction model analyzes and calculates the following dimensions: the number of heterogeneous target systems, the amount of the operation, the merchant's historical dispute rate, and time sensitivity.

3. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 1, characterized in that, Joint scheduling control based on reconciliation risk scoring and adaptive traffic shaping includes: Real-time acquisition of the current system load and reconciliation queue depth of heterogeneous systems; The token generation rate of the token bucket is dynamically adjusted based on the current system load and reconciliation queue depth. For high-risk operations with a reconciliation risk score greater than the set emergency threshold, control their priority to borrow future tokens. Low-risk operations with a reconciliation risk score below the set delay threshold are controlled to enter the delay queue.

4. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 1, characterized in that, Automatically derive field mapping relationships between heterogeneous systems by constructing a constraint propagation network, including: Initialize the belief in field similarity between target heterogeneous systems; Beliefs are propagated through a constraint propagation network, and belief values ​​are updated iteratively. After the beliefs converge, the maximum a posteriori mapping is extracted as the final field mapping relationship.

5. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 4, characterized in that, The field similarity is calculated using a pre-trained graph neural network encoder, including: Extract contextual features of fields for the target heterogeneous system; The contextual features are input into the graph neural network encoder to generate the context-aware feature vector of the current field; Cosine similarity is calculated based on the context-aware feature vector of the current field to determine the semantic similarity score between fields.

6. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 1, characterized in that, The differential vector clock is a 64-bit compressed vector clock, and its structure includes: the high 16 bits for recording the node ID, the middle 16 bits for recording the logic clock, and the low 32 bits for recording the verification.

7. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 6, characterized in that, When directly exchanging differential vector clocks between nodes, vector clock merging is performed. The difference between the reference clock of the current node and the clock of the node to be merged is calculated, and only the differential portion containing the difference is transmitted during network transmission.

8. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 1, characterized in that, When tracing the operational causal dependencies of the cultural and tourism business requests using a coordinatorless vector clock causal consistency protocol, conflict resolution is also performed, including: Detect and analyze whether there are causal conflicts between operations, and identify the types of causal conflicts obtained; The target cultural and tourism business adjudication rules are invoked based on the type of causal conflict, and an automatic adjudication is performed through the target cultural and tourism business adjudication rules. After the automatic adjudication is completed, a resolution certificate is generated, which includes the operation identifier, the adjudication rules for the target cultural and tourism business, and the processing timestamp.

9. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 1, characterized in that, When performing multi-dimensional reconciliation consistency verification, it includes: A three-tiered reconciliation verification system is established based on causal verification, numerical conservation, and anomaly detection. A three-tiered reconciliation verification system is used to prove the consistency of reconciliation from multiple dimensions and to generate a reconciliation consistency verification report.

10. The method for automated data clearing and cross-database reconciliation of heterogeneous systems for cultural tourism business scenarios according to claim 9, characterized in that, A three-tiered reconciliation verification system is used to prove the consistency of reconciliation from multiple dimensions, including: Extract all business operations within a preset time period, construct a causal chain based on vector clocks, and verify whether there are orphan write operations without causal correlation. The operation is hashed to construct a Merkle tree, and the numerical balance relationship between the capital inflow, capital outflow and balance change of all target heterogeneous systems is verified. Finally, the Merkle tree root hash is output. The time series features of the operational data are extracted, and the isolated forest algorithm is used to detect abnormal patterns, identifying complex abnormal patterns where the amount of funds is correct but the time series is abnormal.

Images