Hyper-fusion feature calculation method and system based on big data risk control scene

Abstract

The invention discloses a hyper-fusion feature calculation method and system based on a big data risk control scene. The method comprises the following steps: acquiring streaming data of a big data risk control scene to obtain event data; determining feature dimension attributes corresponding to the event data and whether grouping is carried out based on time slices, evaluating whether features support incremental calculation and selecting a data processing structure, and processing streaming data according to the selected structure and calculation logic to obtain hyper-fusion feature data; storing the hyper-fusion feature data; performing aggregation calculation on the hyper-fusion feature data to obtain a hyper-fusion feature final value; and outputting the final value of the hyper-fusion feature. By implementing the method provided by the invention, the calculation efficiency can be improved, the resource consumption can be reduced, the response speed and the data processing capability of the system can be enhanced, the defects of the existing technical scheme can be overcome, and particularly, the cost can be reduced and the response speed and the data processing capability of the system can be enhanced while the flexibility and the expansibility are improved.

Description

Technical Field

[0001] This invention relates to big data, and more specifically to a hyper-converged feature calculation method and system based on big data risk control scenarios. Background Technology

[0002] In the field of financial risk control, feature computation is a crucial means of risk assessment, and its efficiency and accuracy are paramount. Traditional feature computation methods primarily rely on the expertise of developers and are typically implemented using Python or Spark engines. The core of this approach lies in scanning the entire historical dataset and grouping and aggregating it according to business-defined main dimensions such as transaction card numbers and merchant numbers, before applying statistical functions to generate feature results. However, this method lacks a standardized process, demands high levels of professional skills, and requires the construction of multiple sets of logic when processing multi-source data. This leads to complex management and heavy maintenance workload in medium to large-scale systems, while also presenting data management and resource reuse issues.

[0003] With the development of big data technology, feature computation patterns based on the Lambda architecture have emerged. This pattern combines the capabilities of batch processing and stream processing, aiming to provide high fault tolerance, flexibility, and scalability. However, the Lambda architecture requires maintaining two independent sets of code logic and technology stacks, which increases system complexity and cost, while also introducing inconsistencies between online and offline systems and the risk of difficulty in guaranteeing data consistency. Furthermore, the long data flow not only slows down processing speed but also affects system response time, limiting its application in scenarios with high real-time requirements.

[0004] Another solution is a feature computation model based on ETL technology. It builds a feature computation pipeline through an ETL toolchain, covering key stages such as data extraction, transformation, and loading. While this method can handle multi-source heterogeneous data, its complex software architecture and high deployment costs limit its efficiency and cost-effectiveness. The long data transfer links in the ETL process also lead to performance degradation, further limiting the system's response speed and processing capacity.

[0005] Therefore, it is necessary to design a new method to improve computing efficiency, reduce resource consumption, enhance system response speed and data processing capabilities, overcome the shortcomings of existing technical solutions, and especially reduce costs while improving flexibility and scalability, and enhancing system response speed and data processing capabilities. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a hyper-converged feature calculation method and system based on big data risk control scenarios.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a hyper-converged feature calculation method based on big data risk control scenarios, comprising: Acquire streaming data from big data risk control scenarios to obtain event data; Define the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluate whether the features support incremental calculation and select a data processing structure, and process the streaming data according to the selected structure and calculation logic to obtain hyper-converged feature data; Store the hyper-converged feature data; The hyperconverged feature data is aggregated and calculated to obtain the final value of the hyperconverged feature. Output the final value of the hyperfusion feature.

[0008] The further technical solution is as follows: the event data also includes batch data formed by persisting streaming data to a big data warehouse or RDMS.

[0009] The further technical solution is as follows: The feature dimension attributes corresponding to the event data and whether they are grouped based on time slices are clearly defined; the feature is evaluated to determine whether it supports incremental computation and a data processing structure is selected; the streaming data is processed according to the selected structure and computational logic to obtain hyper-converged feature data, including: The features corresponding to the event data are classified, the dimensional attributes, time slice characteristics and calculation logic are clarified, and it is evaluated whether incremental calculation is suitable to obtain the analysis results. Based on the analysis results, the data processing structure is determined according to whether the features depend on event details. The streaming data is then processed according to the selected structure and computational logic to obtain hyper-converged feature data.

[0010] The further technical solution is as follows: The features corresponding to the event data are classified, the dimensional attributes, time slice characteristics, and calculation logic are clarified, and the suitability for incremental calculation is evaluated to obtain the analysis results, including: The features corresponding to the event data are systematically classified, and the dimensional attributes of each feature, whether it has time slice characteristics, and its calculation logic are identified to obtain the classification results. Based on the classification results, features suitable for incremental calculation are determined; for features that support incremental calculation, the required intermediate process variables are further identified; for features that do not support incremental calculation, they are processed based on event details to obtain the analysis results.

[0011] Its further technical solution is as follows: based on the analysis results, the data processing structure is determined according to whether the features depend on event details, and the streaming data is processed according to the selected structure and calculation logic to obtain hyper-converged feature data, including: Based on the analysis results, it is confirmed whether the feature depends on the event details for incremental calculation; if the feature can be incrementally calculated, the time slice structure is determined to be the data processing structure; if the feature cannot be incrementally calculated but depends on the event details, the event details structure is determined to be the data processing structure. The streaming data is processed according to the selected structure and computational logic to obtain hyperconverged feature data.

[0012] The further technical solution is as follows: The time slice structure divides continuous event data into different time slice units according to time and stores statistical indicators in the form of a dimension table to achieve incremental calculation of features.

[0013] The further technical solution is as follows: the event detail structure records the necessary original attributes of each transaction to support feature calculations that depend on details and are time-sensitive, and realizes flexible processing and extended calculation of features through custom aggregation logic and a unified data access system.

[0014] This invention also provides a hyper-converged feature computing system based on big data risk control scenarios, including: The acquisition unit is used to acquire streaming data from big data risk control scenarios to obtain event data; The hyperconverged unit is used to determine the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluate whether the features support incremental calculation and select a data processing structure, and process the streaming data according to the selected structure and calculation logic to obtain hyperconverged feature data. Storage unit for storing the hyperconverged feature data; A computing unit is used to perform aggregation calculations on the hyperconverged feature data to obtain the final value of the hyperconverged feature; The output unit is used to output the final value of the hyper-fusion feature.

[0015] The present invention also provides a computer device, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the above-described method.

[0016] The present invention also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0017] The advantages of this invention compared to existing technologies are as follows: By acquiring streaming data and clarifying the feature dimensions and attributes corresponding to event data, as well as whether time-slice grouping is necessary, this invention assesses the possibility of features supporting incremental computation, selects the optimal data processing structure and computational logic to process the data, and achieves efficient storage and fast access. This method not only reduces resource consumption but also enhances the system's response speed and data processing capabilities, especially in terms of flexibility and scalability. Compared to existing technical solutions, this invention can reduce costs while improving the system's adaptability and response efficiency to different business needs, ensuring high-performance performance in application scenarios such as real-time decision-making and batch decision-making, effectively overcoming the limitations and shortcomings of traditional data processing solutions. This method makes data processing more intelligent and efficient, greatly improving the overall processing capabilities and response speed in big data risk control scenarios.

[0018] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic diagram illustrating an application scenario of the hyperconverged feature calculation method based on big data risk control provided in this embodiment of the invention; Figure 2 A flowchart illustrating the hyperconverged feature calculation method based on big data risk control scenarios provided in this embodiment of the invention; Figure 3 This is a schematic diagram of an example process for a hyper-converged feature calculation method based on big data risk control scenarios provided in an embodiment of the present invention; Figure 4 A schematic block diagram of a hyperconverged feature computing system based on big data risk control scenarios provided in an embodiment of the present invention; Figure 5 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0023] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0025] Please see Figure 1 and Figure 2 , Figure 1 This is a schematic diagram illustrating an application scenario of the hyperconverged feature calculation method based on big data risk control provided in this embodiment of the invention. Figure 2 This is a schematic flowchart illustrating a hyperconverged feature calculation method for big data risk control scenarios provided in this invention. This method is applied in a server. The server interacts with the terminal, systematically classifies event data features, clarifies dimensional attributes and time-slice characteristics, assesses whether incremental computation is supported, selects a suitable data processing structure (time-slice or event detail structure), processes the streaming data to obtain hyperconverged feature data, and finally aggregates and calculates the hyperconverged feature values. This method not only achieves effective incremental feature calculation and reduces resource consumption, but also enhances the system's flexibility and scalability by using an event detail structure to process features that do not support incremental computation. This method significantly improves computational efficiency and response speed, while optimizing data processing capabilities, overcoming the shortcomings of existing technologies in terms of cost, efficiency, and flexibility, thus providing a more efficient, economical, and powerful data processing solution.

[0026] Figure 2 This is a flowchart illustrating the hyper-converged feature calculation method based on big data risk control scenarios provided in this embodiment of the invention. Figure 2 As shown, the method includes the following steps S110 to S150.

[0027] S110. Obtain streaming data from big data risk control scenarios to obtain event data.

[0028] In this embodiment, the event data also includes batch data formed by persisting streaming data to a big data warehouse or RDMS.

[0029] In an embodiment of the present invention, step S110 mainly involves acquiring streaming data from a big data risk control scenario and converting it into event data that can be used for subsequent processing and analysis. This process is the foundation of the entire hyperconverged feature calculation method, ensuring the real-time nature and accuracy of the data.

[0030] Event data refers to records of transactions or operations that occur at a specific point in time or within a specific time period. These records contain key information such as card number, transaction amount, and transaction time. In this invention, event data includes not only data received directly from streaming data sources (such as HTTP, Kafka, etc.), but also batch data formed by persisting this streaming data to a big data warehouse or relational database management system (RDMS). This means that event data includes both the latest data flowing into the system in real time and historical data, thus enabling a more comprehensive risk assessment.

[0031] Streaming data refers to data received through real-time pipelines (such as HTTP requests, Kafka message queues, etc.). This type of data is characterized by its high real-time performance, enabling it to reflect the latest business dynamics in a timely manner.

[0032] Batch data refers to datasets created by storing streaming data in a big data warehouse or RDMS. Batch data is primarily used to support offline analysis and complex computations, providing detailed information over a longer timeframe than streaming data.

[0033] Specifically, first, the system needs to be configured with the appropriate interfaces or connectors to receive streaming data from different sources. This typically involves support for transport protocols such as HTTP and Kafka.

[0034] Streaming data entering the system is generally in JSON format and needs to be parsed and transformed to meet internal processing requirements. During this process, the raw data is broken down into multiple dimensions of information, such as transaction serial number, card number, transaction amount, and transaction time. To ensure data traceability and support more complex analytical needs, the streaming data is simultaneously written to a big data warehouse or RDS, forming batch data. This approach offers the advantages of batch processing for in-depth analysis and facilitates future data backtracking needs. Based on the data processed through these steps, the system can extract specific event data, which will form the basis for subsequent feature calculations, aggregations, and service provision.

[0035] Through this series of operations, step S110 ensures that the system can effectively obtain necessary information from big data risk control scenarios, laying a solid foundation for subsequent feature calculations, risk assessments, and other tasks. Furthermore, by combining streaming and batch data, the method of this invention significantly improves the flexibility and efficiency of data processing, contributing to the construction of a more accurate and efficient risk control system.

[0036] S120. Determine the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluate whether the features support incremental calculation and select a data processing structure, and process the streaming data according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0037] In this embodiment, hyper-fusion feature data refers to a comprehensive feature representation that can reflect real-time dynamics and integrate historical information by identifying the dimensional attributes and time slice characteristics of features, evaluating whether they support incremental computation, and selecting an appropriate data structure (such as a time slice structure or an event detail structure) to process streaming data.

[0038] Step S120 aims to conduct in-depth analysis of the acquired event data and design corresponding data processing structures based on its characteristics, thereby providing a foundation for subsequent feature calculations. This process involves not only data classification and parsing, but also the evaluation of whether the features are suitable for incremental calculations, and the selection of appropriate hyper-converged data structures for data processing.

[0039] In one embodiment, step S120 described above may include steps S121 to S122.

[0040] S121. Classify the features corresponding to the event data, clarify the dimensional attributes, time slice characteristics and calculation logic, and evaluate whether it is suitable for incremental calculation in order to obtain the analysis results.

[0041] In this embodiment, the analysis result refers to the conclusive information that reflects the potential patterns, trends, or insights of the data, obtained by processing the collected data using specific algorithms and models.

[0042] In one embodiment, step S121 described above may include steps S121 to S122.

[0043] S121. Systematically classify the features corresponding to the event data, identify the dimensional attributes of each feature, whether it has time slice characteristics, and determine its calculation logic to obtain the classification result.

[0044] In this embodiment, the classification result refers to the final determination of dividing the data into different categories according to certain standards or features after analyzing the items in the dataset using a specific algorithm.

[0045] First, it's necessary to identify the dimensional attributes of each feature. Dimensional attributes are key information used for global grouping during feature computation, such as card number and transaction type. These attributes help us divide event data into different subsets for more accurate feature computation.

[0046] Next, it is necessary to determine whether the feature exhibits time-slicing characteristics. Time-slicing refers to dividing continuous event data into specific time units (such as seconds, minutes, hours, days, etc.) to facilitate incremental calculations and real-time updates. For features exhibiting time-slicing characteristics, it is also necessary to further determine their specific time units.

[0047] Finally, based on the above analysis, the specific calculation logic for each feature is clarified. This includes, but is not limited to, operations such as frequency counting, summation of amounts, and average calculation. Establishing the calculation logic can guide subsequent data processing procedures.

[0048] S122. Based on the classification results, determine the features suitable for incremental calculation; wherein, for features that support incremental calculation, further clarify the required intermediate process variables; for features that do not support incremental calculation, process them based on event details to obtain the analysis results.

[0049] Based on the classification results obtained from S121, we evaluate which features are suitable for incremental computation. Incremental computation allows us to update existing results only based on newly incoming data, without recalculating the entire historical dataset, greatly improving computational efficiency.

[0050] For features that support incremental calculation, the intermediate variables required in the calculation process should be further clarified. These variables may include cumulative counts, cumulative amounts, etc., which are the basis for implementing incremental calculation.

[0051] For features that do not support incremental calculations, an event detail approach must be used. This approach requires recording detailed information for each transaction so that complex feature calculations can be performed based on this raw data when needed.

[0052] Through the above steps, the most suitable processing method can be selected for each feature, and hyper-fusion feature data can be constructed accordingly. Specifically: Time-slice structure: Suitable for features that can be grouped and incrementally computed using time-slice units. This type of structure utilizes the KKV (Key-Key-Value) form, where the first K represents the dimension and value, the second K represents the time slice, and V represents the result or state within that time slice.

[0053] Event detail structure: Used when features cannot be easily grouped or incrementally calculated by time slices. This structure adopts the KV (Key-Value) form, where K represents the dimension and value, and V contains detailed records of all related events under that dimension.

[0054] The resulting hyper-converged feature data not only covers the latest event information but also takes into account the impact of historical data, ensuring the comprehensiveness and accuracy of feature calculation. Furthermore, by adopting differentiated processing strategies for different features, the method of this invention significantly improves the system's flexibility and scalability, providing financial institutions with more accurate risk assessment capabilities.

[0055] S122. Based on the analysis results, determine the data processing structure according to whether the features depend on event details, and process the streaming data according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0056] In this embodiment, step S122 involves determining whether feature computation depends on event details based on previous analysis results, and selecting an appropriate data processing structure accordingly. This step is the core of hyperconverged feature data processing, ensuring that streaming data can be effectively transformed into feature data that can be used for decision support.

[0057] In one embodiment, step S122 described above may include steps S1221 to S1222.

[0058] S1221. Based on the analysis results, confirm whether the feature depends on the event details for incremental calculation; if the feature can be incrementally calculated, then determine the time slice structure as the data processing structure; if the feature cannot be incrementally calculated but depends on the event details, then determine the event details structure as the data processing structure.

[0059] In this embodiment, the data structure type (time slice structure or event detail structure) required for feature calculation is determined.

[0060] Based on the preliminary analysis of feature dimension attributes, time slice requirements, and incremental computing capabilities, it was determined whether feature computation could be performed directly through a time slice structure.

[0061] If features can be incrementally calculated directly without relying on event details, then a time-slice structure is suitable.

[0062] If feature calculations need to consider the original attributes of each transaction, especially those features that are time-sensitive or cannot be simply described by time slices, then an event detail structure should be used.

[0063] Based on the above analysis, the decision is made to use either a time-slice structure or an event detail structure.

[0064] The time-slice structure is suitable for features that can be directly grouped and statistically analyzed based on time, such as the number of transactions and the total amount.

[0065] Event detail structures are used for complex feature calculations that require access to specific transaction details, such as risk assessment metrics that only take effect under certain conditions.

[0066] S1222. Process the streaming data according to the selected structure and computational logic to obtain hyper-converged feature data.

[0067] In this embodiment, the time slice structure divides continuous event data into different time slice units according to time and stores statistical indicators in the form of a dimension table to achieve incremental calculation of features.

[0068] The event detail structure records the necessary original attributes of each transaction to support feature calculations that depend on details and are time-sensitive. It enables flexible processing and extended calculation of features through custom aggregation logic and a unified data access system.

[0069] The steps in this embodiment extract and process hyperconverged feature data from streaming data based on the selected data structure and corresponding computational logic.

[0070] For different types of features, data processing is performed according to the requirements of time slice structure or event detail structure.

[0071] For time-slice structures, continuous events are divided into different time-slice units, and statistical operations such as counting, summing, and averaging are performed within each unit.

[0072] For the event details structure, record the key attributes of each transaction so that more complex feature calculations can be performed later based on this detailed information.

[0073] After completing the above processing, the system will output a series of hyperconverged feature data, which includes both basic statistical information and advanced analysis results based on detailed data, providing a solid foundation for subsequent risk assessment.

[0074] In summary, step S122, by accurately identifying the needs for feature calculation and selecting the most suitable data processing structure accordingly, makes the conversion process from streaming data to hyperconverged feature data more efficient and accurate, greatly improving the performance and applicability of the risk control model.

[0075] S130. Store the hyper-fusion feature data.

[0076] In this embodiment, the processed hyper-fusion feature data is persistently stored to ensure that the data can be accessed and expanded efficiently.

[0077] Specifically, high-speed caching databases such as Aerospike or Redis can be chosen to store hyperconverged feature data. These databases offer fast read / write capabilities and good scalability, making them ideal for scenarios requiring real-time processing and frequent access.

[0078] Considering the unique characteristics of hyperconverged feature data structures (such as time-slice structures and event detail structures), a reasonable storage scheme needs to be designed to support efficient query and update operations. For example, for time-slice structured data, it can be stored by partitioning according to the time dimension; for event detail structures, the integrity and traceability of the original transaction records should be emphasized.

[0079] The hyperconverged feature data of the intermediate or final results obtained after calculation are stored in the selected database according to the above design scheme to ensure the security and reliability of the data.

[0080] S140. Perform aggregation calculations on the hyperconverged feature data to obtain the final value of the hyperconverged feature.

[0081] In this embodiment, the final value of hyperconverged features refers to a comprehensive feature index that can be used for actual business decision-making, obtained by aggregating and calculating the stored hyperconverged feature data.

[0082] By further aggregating and calculating the stored hyperconverged feature data, final feature values ​​that can be used for actual business decisions are generated.

[0083] Based on specific business needs and feature definitions, determine the specific logic for aggregation calculations. This may include, but is not limited to, summarizing data within a time slice (such as summation, averaging, maximum / minimum values, etc.), merging data across time slices, and complex calculations based on event details. Use appropriate calculation methods or tools (such as custom Lua scripts) to aggregate the hyperconverged feature data in the storage layer. This process may involve cleaning up expired data (based on TTL and time slice range) and dynamically adjusting aggregation strategies to adapt to constantly changing data flows. After completing the aggregation calculations, the system will output a series of final values ​​for the hyperconverged features, which can be directly used in application scenarios such as risk assessment and decision-making.

[0084] S150, Output the final value of the hyper-fusion feature.

[0085] In this embodiment, the calculated final value of the hyperconverged feature is provided to external systems or applications for their use in actual business scenarios.

[0086] Design and implement unified interfaces (such as HTTP API, Dubbo services, asynchronous message queues, etc.) to facilitate other systems in easily obtaining the required feature data. Through these interfaces, the final values ​​of the hyper-converged features are passed to the requester or other applications that rely on this data. Ensure the security and efficiency of data transmission, while considering performance optimization under potential high concurrency scenarios. The final values ​​are integrated into various business scenarios, such as real-time decision engines, batch analysis platforms, or machine learning model training processes, providing strong support for risk management in financial institutions.

[0087] In summary, steps S130 to S150 constitute a complete chain from storing hyperconverged feature data to aggregation calculation and finally outputting the final value, making feature calculation in big data risk control scenarios more flexible and efficient, and able to meet the requirements of accurate risk assessment under different business needs.

[0088] In this embodiment, the method is particularly applicable to big data risk control in the financial sector. It primarily achieves effective integration and computation of feature information from different data sources, application scenarios, and language environments through a multi-layered hyper-converged feature model. This overcomes the limitations of traditional risk control systems in feature processing, significantly enhancing system compatibility and scalability, and improving real-time analysis capabilities. Specifically, by introducing a unified data structure standard, incremental feature computation is supported, enabling the system to flexibly respond to diverse feature computation needs and operators. In particular, a time-slice data structure called KKV (KeyKeyValue) is used to implement hyper-converged computation, where the first K represents the dimension and its corresponding value, the second K represents the time slice, and V stores the final computation result. This design not only improves data processing efficiency but also provides strong support for real-time, streaming, and batch feature computation, thereby providing financial institutions with more comprehensive and accurate risk assessment services.

[0089] An example of a time-slice data structure for hyperconverged computing could be as follows: First, define dimensions and their corresponding values ​​as the first-level keys. Then, set the second-level keys according to specific time periods. Finally, store the calculated results in the value. This approach ensures efficient and accurate data processing and analysis capabilities even in complex and ever-changing risk control scenarios.

[0090] Table 1 shows an example of the hyperconverged computing time slice data structure, and Table 2 shows the feature calculation logic.

[0091] Table 1. Number of transactions by card number in the last 7 days serial number card number Trading hours Hyperconverged architecture COUNT E000001 C0001 2025-06-01 08:01:12 {"card_no=C0001":{"20250601":1}} E000002 C0001 2025-06-01 14:04:53 {"card_no=C0001":{"20250601":2}} E000003 C0001 2025-06-05 20:45:19 {"card_no=C0001":{"20250601":2,"20250605":1}} E000004 C0001 2025-06-10 10:38:27 {"card_no=C0001":{"20250605":1,"20250610":1}} The Value of the frequency calculation indicator structure is an integer, representing the number of transactions within the current time slice.

[0092] In the example above, when sequence event E000001 occurs, the characteristic value (number of transactions for the card number in the last 7 days) is 1; When sequence event E000002 occurs, the characteristic value (number of transactions for the card number in the last 7 days) is 2; When sequence event E000003 occurs, the characteristic value (number of transactions for the card number in the last 7 days) is 3; When the sequence event E000004 occurs, the characteristic value (the number of transactions for the card number in the last 7 days) will increase over time, and some transactions will exceed the 7-day period. Therefore, the final characteristic value is 2, and the two transactions E000004 and E000003 are calculated respectively.

[0093] Table 2. Average Transaction Amount by Card Number in the Last 7 Days serial number card number Amount Trading hours Hyperconverged AVG E000001 C0001 80 2025-06-01 08:01:12 {"card_no=C0001":{"20250601":[80,1]}} E000002 C0001 68 2025-06-01 14:04:53 {"card_no=C0001":{"20250601":[148,2]}} E000003 C0001 34 2025-06-05 20:45:19 {"card_no=C0001":{"20250601":[148,2],"20250605":[34,1]}} E000004 C0001 67 2025-06-10 10:38:27 {"card_no=C0001":{"20250605":[34,1],"20250610":[67,1]}} For the average value calculation indicator structure, the Value is represented by an array of length 2. The first element of the array records the sum of all transaction amounts within the current time slice, while the second element records the number of transactions within that time slice.

[0094] Take a specific time-series event as an example: When E000001 occurs, based on the transaction data of the most recent 7 days, the characteristic value, i.e. the average transaction amount of the transaction card number, is calculated as 80 (total amount) divided by 1 (number of transactions), and the result is 80.

[0095] When E000002 occurs, the average value is updated to 148 (cumulative amount) divided by 2 (cumulative number of times), resulting in 74.

[0096] With the occurrence of E000003, this average value was further adjusted to (148+34) (latest cumulative amount) divided by (2+1) (latest cumulative number of times), and the new average value was calculated to be 60.67.

[0097] When E000004 occurred, considering that data exceeding 7 days would no longer be included in the statistics, the final calculated characteristic value was (34+67) (sum of valid amounts) divided by (1+1) (number of valid occurrences), which is 50.5.

[0098] The examples mentioned here demonstrate how to perform statistical calculations based on transaction amount and frequency, both of which rely on the KKV (KeyKeyValue) time-slice data structure. This structure is not only applicable to the operators in the above examples, but also plays an important role in many other feature calculation operators, such as the number of transactions, average, maximum and minimum values, variance, standard deviation, moving speed, and correlation analysis within the most recent time period.

[0099] Furthermore, in the field of risk control, there are also needs for feature calculations that are sensitive to time series or cannot be described solely by time-slice structures. For this, a method called hyperconverged computing event detail data structure can be adopted. It is based on a KV (Key-Value) structure and is referred to as an event detail structure. In this structure, the first K represents the dimension and its corresponding value, while V represents the detailed event information under the corresponding dimension. This method provides an effective solution for handling complex situations that require meticulous consideration of time sequence or cannot be simply categorized into specific time slices.

[0100] Specifically, the method in this embodiment aims to provide efficient data analysis and decision support through streaming and batch data processing. The specific process is as follows: First, streaming data from channels such as HTTP or Kafka is received and imported into the real-time computing module for initial processing. Simultaneously, this data is also stored in a big data warehouse or relational database management system (RDMS) to ensure data persistence. Batch data, on the other hand, goes directly into the offline computing module.

[0101] Whether it's streaming or batch data events, they all pass through the hyperconverged computing module, where the event data is extracted in a structured manner to meet subsequent processing needs.

[0102] The hyperconverged data storage layer is responsible for the unified storage and management of processed data, ensuring data consistency and integrity.

[0103] After data storage, the hyperconverged data aggregation module performs further aggregation calculations on these data to complete the processing and calculation of feature values.

[0104] Ultimately, the processed and aggregated feature results can be used to provide decision support information to various business scenarios through a unified feature service platform.

[0105] At this stage, the focus is on classifying different features according to their implementation methods and clarifying their computational logic. This includes determining the feature's dimensional attributes, whether to use time-slice partitioning, and the specific computation method. Dimensional attributes help determine the global grouping rules in feature computation, while time-slice units are used for secondary grouping. Furthermore, it is necessary to evaluate whether the feature is suitable for incremental computation; if so, the necessary intermediate variables need to be defined; otherwise, an event detail structure should be used for processing.

[0106] This embodiment involves designing a feature incremental calculation structure that integrates streaming, batch, and combined data processing mechanisms. Based on business requirements, it first determines whether the feature can be incrementally calculated without relying on event details. If so, a time-slice structure is used; otherwise, an event-detailed structure is chosen. For example, in the risk control field, commonly used time-slice structures include dividing continuous event data into time units such as seconds, minutes, hours, days, or months, and storing statistical indicators such as the number of transactions and average transaction amount. For features that require detailed event information for calculation, or those that are sensitive to the order of events, an event-detailed structure is more suitable.

[0107] This design not only improves system performance but also provides flexible scalability for features that cannot be incrementally computed, making custom feature aggregation processing based on detailed data possible. Simultaneously, it supports the construction of a unified data access system to facilitate feature processing and computation.

[0108] Next, using the previously designed data structure, computations are performed on event data from multiple sources and formats to generate intermediate process data. This data can be obtained through real-time requests, message queues, or historical data, and the structure of multi-source data is unified through various technical means. The hyperconverged infrastructure is designed to be compatible with various types of data input, including but not limited to structured, unstructured, and semi-structured data (such as XML, JSON, etc.), thereby supporting the computation of feature-rich intermediate process data.

[0109] Persist in storing the results calculated in the previous step to ensure efficient data storage, fast access, and elastic scalability. It is recommended to use a high-speed caching database, such as Aerospike or Redis, to store these calculation results to meet high-performance requirements.

[0110] Based on the intermediate feature data obtained from the previous steps, further aggregation is performed to derive the final feature results. Depending on the business configuration, it can be selected whether to include data from the current event in the calculation. Due to the use of a hyperconverged infrastructure, the aggregation process can be executed by the storage engine, typically through a custom Lua script. Furthermore, during the aggregation process, expired historical data is also cleaned up according to TTL (Time To Live) and time slice range.

[0111] After the feature results are aggregated and calculated, the final value of the feature can be obtained, providing decision support services for actual business scenarios. The feature service can provide services to the decision system through a unified HTTP interface, Dubbo interface, or asynchronous messages, and is suitable for various application scenarios such as real-time decision-making, batch decision-making, and modeling.

[0112] For example, such as Figure 3 As shown, it receives streaming data, mainly in JSON format, as well as batch data from big data storage. Next, it applies a hyperconverged infrastructure processing method to the streaming data to extract the intermediate data required for feature calculation. Finally, it aggregates the hyperconverged feature intermediate data to form the final feature result data, and provides support to external systems through the feature service, which helps in real-time or batch decision-making, model building and other application scenarios.

[0113] This process design not only improves data processing efficiency but also enhances the system's flexibility and scalability, making data analysis and decision support in different business scenarios more convenient and efficient.

[0114] This embodiment focuses on solving the feature computation problem in the field of financial risk control, and proposes an innovative hyperconverged computing model. This model is specifically designed to meet the feature computation needs of different application scenarios, and is particularly suitable for time-slice scenarios, feature detail computation, and feature dimension computation. For example, one of the commonly used indicators in financial risk control is the number of transactions for a particular card number in the last 7 days. The method in this embodiment optimizes the storage and processing of this type of event data through a hyperconverged computing model to achieve efficient feature computation.

[0115] This hyperconverged model is not limited to specific application scenarios but also possesses high flexibility and scalability, supporting risk control feature calculations across various scenarios. It allows users to customize feature calculation logic according to specific business needs, significantly improving system adaptability and responsiveness. Furthermore, in terms of non-functionality, the method in this embodiment exhibits high compatibility, ease of expansion, and superior cost-effectiveness, achieving true "hyperconvergence"—integrating multiple computing capabilities on a single platform—effectively reducing system complexity and operating costs. In summary, the method in this embodiment provides a comprehensive, efficient, and flexible solution to address the diverse feature calculation challenges in the financial risk control field.

[0116] The aforementioned hyper-converged feature computation method for big data risk control scenarios acquires streaming data, clarifies the feature dimension attributes corresponding to event data and whether time-slice grouping is necessary, assesses the possibility of feature-supported incremental computation, selects the optimal data processing structure and computational logic to process data, and achieves efficient storage and fast access. This method not only reduces resource consumption but also enhances the system's response speed and data processing capabilities, especially in terms of flexibility and scalability. Compared to existing technical solutions, this invention can reduce costs while improving the system's adaptability and response efficiency to different business needs, ensuring high-performance performance in application scenarios such as real-time decision-making and batch decision-making, effectively overcoming the limitations and shortcomings of traditional data processing solutions. This method makes data processing more intelligent and efficient, greatly improving the overall processing capabilities and response speed in big data risk control scenarios.

[0117] Figure 4 This is a schematic block diagram of a hyperconverged feature calculation system 300 based on a big data risk control scenario provided in an embodiment of the present invention. Figure 4 As shown, corresponding to the above-described hyperconverged feature calculation method based on big data risk control scenarios, this invention also provides a hyperconverged feature calculation system 300 based on big data risk control scenarios. This hyperconverged feature calculation system 300 includes a unit for executing the above-described hyperconverged feature calculation method based on big data risk control scenarios, and the system can be configured in a server. Specifically, please refer to... Figure 4 The hyperconverged feature computing system 300 based on big data risk control scenarios includes an acquisition unit 301, a hyperconverged unit 302, a storage unit 303, a computing unit 304, and an output unit 305.

[0118] The acquisition unit 301 is used to acquire streaming data from a big data risk control scenario to obtain event data; the hyperconverged unit 302 is used to determine the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluate whether the features support incremental calculation and select a data processing structure, and process the streaming data according to the selected structure and calculation logic to obtain hyperconverged feature data; the storage unit 303 is used to store the hyperconverged feature data; the calculation unit 304 is used to perform aggregation calculation on the hyperconverged feature data to obtain the final value of the hyperconverged feature; and the output unit 305 is used to output the final value of the hyperconverged feature.

[0119] In one embodiment, the hyperconverged unit 302 includes: The analysis subunit is used to classify the features corresponding to the event data, clarify the dimensional attributes, time slice characteristics and calculation logic, and evaluate whether it is suitable for incremental calculation to obtain analysis results; the calculation subunit is used to determine the data processing structure based on the analysis results and whether the features depend on the event details, and to process the streaming data according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0120] In one embodiment, the analysis subunit includes: The classification module is used to systematically classify the features corresponding to the event data, identify the dimensional attributes of each feature, whether it has time slice characteristics, and determine its calculation logic to obtain classification results; the determination module is used to determine the features suitable for incremental calculation based on the classification results; wherein, for features that support incremental calculation, the required intermediate process variables are further clarified; for features that do not support incremental calculation, they are processed based on event details to obtain analysis results.

[0121] In one embodiment, the computing subunit includes: The confirmation module is used to confirm whether the feature depends on the event details for incremental calculation based on the analysis results; if the feature can be incrementally calculated, the time slice structure is determined as the data processing structure; if the feature cannot be incrementally calculated but depends on the event details, the event details structure is determined as the data processing structure. The processing module is used to process the streaming data according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0122] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the hyperconverged feature computing system 300 based on big data risk control scenarios and each unit can be referred to the corresponding description in the aforementioned method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0123] The aforementioned hyperconverged feature computing system 300 based on big data risk control scenarios can be implemented as a computer program, which can be used in scenarios such as... Figure 5 It runs on the computer device shown.

[0124] Please see Figure 5 , Figure 5 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.

[0125] See Figure 5The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0126] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a hyper-converged feature calculation method based on a big data risk control scenario.

[0127] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0128] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a hyper-converged feature calculation method based on big data risk control scenarios.

[0129] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0130] The processor 502 is used to run a computer program 5032 stored in the memory to perform the following steps: Acquire streaming data from big data risk control scenarios to obtain event data; determine the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices; evaluate whether the features support incremental calculation and select a data processing structure; process the streaming data according to the selected structure and calculation logic to obtain hyperconverged feature data; store the hyperconverged feature data; perform aggregation calculation on the hyperconverged feature data to obtain the final value of the hyperconverged feature; output the final value of the hyperconverged feature.

[0131] The event data also includes batch data formed by persisting streaming data to a big data warehouse or RDMS.

[0132] In one embodiment, when the processor 502 implements the steps of determining the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluating whether the features support incremental computation and selecting a data processing structure, and processing the streaming data according to the selected structure and computational logic to obtain hyper-converged feature data, the specific implementation is as follows: The features corresponding to the event data are classified, the dimensional attributes, time slice characteristics and calculation logic are clarified, and the suitability for incremental calculation is evaluated to obtain the analysis results. Based on the analysis results, the data processing structure is determined according to whether the features depend on the event details, and the streaming data is processed according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0133] In one embodiment, when the processor 502 performs the steps of classifying the features corresponding to the event data, clarifying the dimensional attributes, time slice characteristics and calculation logic, and evaluating whether incremental calculation is suitable to obtain the analysis results, the specific implementation is as follows: The features corresponding to the event data are systematically classified, and the dimensional attributes, time slice characteristics, and calculation logic of each feature are identified to obtain the classification results. Based on the classification results, features suitable for incremental calculation are determined. For features that support incremental calculation, the required intermediate process variables are further clarified. For features that do not support incremental calculation, they are processed based on event details to obtain the analysis results.

[0134] In one embodiment, when the processor 502 implements the step of determining the data processing structure based on whether the features depend on event details according to the analysis results, and processing the streaming data according to the selected structure and computational logic to obtain hyper-converged feature data, the specific implementation steps are as follows: Based on the analysis results, it is confirmed whether the feature depends on the event details for incremental calculation; if the feature can be incrementally calculated, the time slice structure is determined as the data processing structure; if the feature cannot be incrementally calculated but depends on the event details, the event details structure is determined as the data processing structure; the streaming data is processed according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0135] The time-slice structure divides continuous event data into different time-slice units according to time and stores statistical indicators in the form of a dimension table to achieve incremental calculation of features.

[0136] The event detail structure records the necessary original attributes of each transaction to support feature calculations that depend on details and are time-sensitive. It enables flexible processing and extended calculation of features through custom aggregation logic and a unified data access system.

[0137] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0138] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0139] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform the following steps: Acquire streaming data from big data risk control scenarios to obtain event data; determine the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices; evaluate whether the features support incremental calculation and select a data processing structure; process the streaming data according to the selected structure and calculation logic to obtain hyperconverged feature data; store the hyperconverged feature data; perform aggregation calculation on the hyperconverged feature data to obtain the final value of the hyperconverged feature; output the final value of the hyperconverged feature.

[0140] The event data also includes batch data formed by persisting streaming data to a big data warehouse or RDMS.

[0141] In one embodiment, when the processor executes the computer program to determine the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluate whether the features support incremental computation and select a data processing structure, and process the streaming data according to the selected structure and computational logic to obtain hyperconverged feature data, the processor specifically implements the following steps: The features corresponding to the event data are classified, the dimensional attributes, time slice characteristics and calculation logic are clarified, and the suitability for incremental calculation is evaluated to obtain the analysis results. Based on the analysis results, the data processing structure is determined according to whether the features depend on the event details, and the streaming data is processed according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0142] In one embodiment, when the processor executes the computer program to classify the features corresponding to the event data, determine the dimensional attributes, time slice characteristics and calculation logic, and evaluate whether incremental calculation is suitable to obtain the analysis results, the processor specifically implements the following steps: The features corresponding to the event data are systematically classified, and the dimensional attributes, time slice characteristics, and calculation logic of each feature are identified to obtain the classification results. Based on the classification results, features suitable for incremental calculation are determined. For features that support incremental calculation, the required intermediate process variables are further clarified. For features that do not support incremental calculation, they are processed based on event details to obtain the analysis results.

[0143] In one embodiment, when the processor executes the computer program to implement the step of determining the data processing structure based on the analysis results according to whether the features depend on event details, and processing the streaming data according to the selected structure and computational logic to obtain hyper-converged feature data, the processor specifically implements the following steps: Based on the analysis results, it is confirmed whether the feature depends on the event details for incremental calculation; if the feature can be incrementally calculated, the time slice structure is determined as the data processing structure; if the feature cannot be incrementally calculated but depends on the event details, the event details structure is determined as the data processing structure; the streaming data is processed according to the selected structure and calculation logic to obtain hyper-converged feature data.

[0144] The time-slice structure divides continuous event data into different time-slice units according to time and stores statistical indicators in the form of a dimension table to achieve incremental calculation of features.

[0145] The event detail structure records the necessary original attributes of each transaction to support feature calculations that depend on details and are time-sensitive. It enables flexible processing and extended calculation of features through custom aggregation logic and a unified data access system.

[0146] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0147] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0148] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0149] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the system of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0150] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0151] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A hyper-converged feature calculation method based on big data risk control scenarios, characterized in that, include: Acquire streaming data from big data risk control scenarios to obtain event data; Define the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluate whether the features support incremental calculation and select a data processing structure, and process the streaming data according to the selected structure and calculation logic to obtain hyper-converged feature data; Store the hyper-converged feature data; The hyperconverged feature data is aggregated and calculated to obtain the final value of the hyperconverged feature. Output the final value of the hyperfusion feature.

2. The hyper-converged feature calculation method based on big data risk control scenarios according to claim 1, characterized in that, The event data also includes batch data formed by persisting streaming data to a big data warehouse or RDMS.

3. The hyper-converged feature calculation method based on big data risk control scenarios according to claim 2, characterized in that, The process involves clarifying the feature dimension attributes corresponding to the event data and whether they are grouped based on time slices, evaluating whether the features support incremental computation and selecting a data processing structure, and processing the streaming data according to the selected structure and computational logic to obtain hyperconverged feature data, including: The features corresponding to the event data are classified, the dimensional attributes, time slice characteristics and calculation logic are clarified, and it is evaluated whether incremental calculation is suitable to obtain the analysis results. Based on the analysis results, the data processing structure is determined according to whether the features depend on event details. The streaming data is then processed according to the selected structure and computational logic to obtain hyper-converged feature data.

4. The hyper-converged feature calculation method based on big data risk control scenarios according to claim 3, characterized in that, The process of classifying the features corresponding to the event data, clarifying dimensional attributes, time slice characteristics and calculation logic, and evaluating whether incremental calculation is suitable to obtain analysis results includes: The features corresponding to the event data are systematically classified, and the dimensional attributes of each feature, whether it has time slice characteristics, and its calculation logic are identified to obtain the classification results. Based on the classification results, features suitable for incremental calculation are determined; for features that support incremental calculation, the required intermediate process variables are further identified; for features that do not support incremental calculation, they are processed based on event details to obtain the analysis results.

5. The hyper-converged feature calculation method based on big data risk control scenarios according to claim 3, characterized in that, The process involves determining the data processing structure based on the analysis results, considering whether the features depend on event details, and then processing the streaming data according to the selected structure and computational logic to obtain hyper-converged feature data, including: Based on the analysis results, it is confirmed whether the feature depends on the event details for incremental calculation; if the feature can be incrementally calculated, the time slice structure is determined to be the data processing structure; if the feature cannot be incrementally calculated but depends on the event details, the event details structure is determined to be the data processing structure. The streaming data is processed according to the selected structure and computational logic to obtain hyperconverged feature data.

6. The hyper-converged feature calculation method based on big data risk control scenarios according to claim 5, characterized in that, The time-slice structure divides continuous event data into different time-slice units according to time and stores statistical indicators in the form of a dimension table to achieve incremental calculation of features.

7. The hyper-converged feature calculation method based on big data risk control scenarios according to claim 5, characterized in that, The event detail structure records the necessary original attributes of each transaction to support feature calculations that depend on details and are time-sensitive. It enables flexible processing and extended calculation of features through custom aggregation logic and a unified data access system.

8. A hyperconverged feature computing system based on big data risk control scenarios, characterized in that: include: The acquisition unit is used to acquire streaming data from big data risk control scenarios to obtain event data; The hyperconverged unit is used to determine the feature dimension attributes corresponding to the event data and whether it is grouped based on time slices, evaluate whether the features support incremental calculation and select a data processing structure, and process the streaming data according to the selected structure and calculation logic to obtain hyperconverged feature data. Storage unit for storing the hyperconverged feature data; A computing unit is used to perform aggregation calculations on the hyperconverged feature data to obtain the final value of the hyperconverged feature; The output unit is used to output the final value of the hyper-fusion feature.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images