A data processing method, apparatus, storage medium, and electronic device
By using a unified timeline for processing multi-source heterogeneous data, the problems of temporal semantic alignment and fusion delay of multi-source heterogeneous time-series data are solved, enabling fast and accurate resource inventory prediction and reducing system computational complexity and computing power overhead.
Patent Information
- Application Number
- CN202611227695.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-13
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional resource data processing systems suffer from problems such as high overhead in time semantic alignment, high latency in fusion of offline and real-time data, and inability to quickly and accurately extrapolate the evolution trend of existing data when dealing with multi-source heterogeneous time-series data.
By employing a unified timeline for multi-source heterogeneous data atomization modeling, a view fusion mechanism combining static historical baselines and dynamic incremental differences, a scenario parameter-driven spatiotemporal transfer extrapolation mechanism, and a bucket aggregation and continuous recursion mechanism within the prediction window, standardized mapping and high-concurrency dynamic fusion of heterogeneous event data from different sources are achieved.
It reduces the system's computational complexity and computing power overhead, and can quickly, smoothly, and accurately derive a continuous sequence of resource inventory prediction values within any target prediction window, providing support for efficient processing of resource data transfer.
Smart Images

Figure CN123114650A_ABST
Abstract
Description
Technical Field
[0001] This specification relates to the field of computer technology, and in particular to a data processing method, apparatus, storage medium and electronic device. Background Technology
[0002] In modern online service platforms and resource transfer management scenarios (such as cloud computing power dispatch centers, power load allocation platforms, carbon resource data management systems, and financial resource service systems), service platforms need to process massive amounts of quantifiable and transferable data, such as computing power resources, power resources, carbon resources, and financial data resources, at high frequency, depending on the specific service scenario. These application service scenarios often involve multi-source heterogeneous data environments. Resource transfer event data generated by different source nodes typically exhibit complex data patterns and heterogeneous temporal characteristics. Furthermore, under the corresponding service prediction requirements, it is necessary to accurately monitor the dynamic resource transfer process across nodes and simulate and aggregate future resource flow trends and stock changes. Summary of the Invention
[0003] This specification provides a data processing method, apparatus, storage medium, and electronic device, the technical solutions of which are as follows: Firstly, this specification provides a data processing method, the method comprising: Extract resource transfer event data from multiple sources and heterogeneous types, and perform data atomization modeling on the resource transfer event data based on a unified time axis to obtain data transfer flow atomic units; Obtain the target differential change data corresponding to the dynamic resource transfer event, and merge the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data; In response to the target prediction task request, based on the resource scheduling prediction scenario parameters, data transfer inference processing is performed on the global resource transfer view data to obtain the prediction scenario resource flow data. The prediction scenario resource flow data is aggregated within the target prediction window of the unified time axis to obtain the net resource change amount, and a sequence of resource stock prediction values is obtained based on the net resource change amount.
[0004] Secondly, this specification provides a data processing apparatus, the apparatus comprising: The atomic modeling module is used to extract resource transfer event data of multiple heterogeneous types and perform data atomization modeling on the resource transfer event data based on a unified time axis to obtain data transfer flow atomic units. The global merging module is used to obtain the target differential change data corresponding to the dynamic resource transfer event, and merge the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data. The request-response module is used to respond to the target prediction task request, determine the resource scheduling prediction scenario parameters, perform data transfer inference processing on the global resource transfer view data to obtain the prediction scenario resource flow data, aggregate the prediction scenario resource flow data within the target prediction window of the unified time axis to obtain the net resource change amount, and synthesize the net resource change amount to obtain the resource stock prediction value sequence.
[0005] Thirdly, this specification provides a computer storage medium storing at least one instruction adapted for loading by a processor and executing method steps of one or more embodiments of this specification.
[0006] Fourthly, this specification provides a computer program product storing at least one instruction adapted to be loaded by a processor and to execute the method steps of one or more embodiments of this specification.
[0007] Fifthly, this specification provides an electronic device that may include: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the method steps of one or more embodiments of this specification.
[0008] The beneficial effects of the technical solutions provided in some embodiments of this specification include at least the following: In one or more embodiments of this specification, a multi-source heterogeneous data atomic modeling mechanism based on a unified time axis, a view fusion mechanism of static historical baseline and dynamic incremental difference, a scenario parameter-driven spatiotemporal transfer extrapolation mechanism, and a bucket aggregation and continuous inventory recursion mechanism within the prediction window are implemented. These mechanisms address the problems of traditional resource data processing systems when dealing with multi-source heterogeneous time-series data, such as high temporal semantic alignment overhead, high latency in offline and real-time data fusion, and the inability to quickly and accurately recursively extrapolate inventory evolution trends during scenario simulation and prediction. They achieve standardized mapping and high-concurrency dynamic fusion of heterogeneous event data from different sources, transforming complex prediction extrapolation and inventory state calculation into view fusion on a unified time axis. This reduces system computational complexity and power consumption, enabling the rapid, smooth, and accurate extrapolation of continuous resource inventory prediction value sequences within any target prediction window, thus supporting efficient resource data transfer processing. Attached Figure Description
[0009] To more clearly illustrate the technical solutions in this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0010] Figure 1 This is a schematic diagram of a data processing system provided in this manual; Figure 2 This is a flowchart illustrating a data processing method provided in this manual; Figure 3 This is a flowchart illustrating a data atomization modeling process provided in this manual; Figure 4 This is a schematic diagram of a data merging process provided in this manual; Figure 5 This is a schematic diagram of a data transfer and deduction process provided in this manual; Figure 6 This is a schematic diagram of an polymerization process provided in this specification; Figure 7 This is a schematic diagram of the structure of a data processing device provided in this specification; Figure 8 This is a schematic diagram of the structure of an electronic device provided in this specification. Detailed Implementation
[0011] The technical solutions in this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.
[0012] In the description of this specification, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this specification, it should be noted that, unless otherwise expressly specified and limited, "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. Those skilled in the art can understand the specific meaning of the above terms in this specification based on the specific circumstances. Furthermore, in the description of this specification, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.
[0013] The present specification will now be described in detail with reference to specific embodiments.
[0014] Please see Figure 1 This is a schematic diagram of a data processing system provided in this specification. Figure 1 As shown, the data processing system may include at least a client cluster and a service platform 100.
[0015] The client cluster may include at least one client, such as Figure 1 As shown, it specifically includes client 1 corresponding to user 1, client 2 corresponding to user 2, ..., client n corresponding to user n, where n is an integer greater than 0.
[0016] Each client in a client cluster can be an electronic device with communication capabilities, including but not limited to: wearable devices, handheld devices, personal computers, tablets, in-vehicle devices, smartphones, computing devices, or other processing devices connected to a wireless modem. Electronic devices may have different names in different networks, such as: user equipment, access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, electronic device, wireless communication device, user agent or user device, cellular phone, cordless phone, personal digital assistant (PDA), and electronic devices in 5G networks or future evolved networks.
[0017] The service platform 100 can be a standalone server device, such as a rack-mount, blade, tower, or cabinet-type server device, or a workstation, mainframe, or other hardware device with strong computing power; or it can be a server cluster composed of multiple servers. The servers in the service cluster can be composed in a symmetrical manner, wherein each server is functionally and hierarchically equivalent in the transaction chain, and each server can provide services independently. The independent provision of services can be understood as not requiring the assistance of other servers.
[0018] In one or more embodiments of this specification, the service platform 100 can establish a communication connection with at least one client in the client cluster, and complete data interaction during data processing based on this communication connection. The client devices can be categorized into various types of terminals, server nodes, data acquisition terminals, resource data scheduling nodes, resource data monitoring nodes, and user service interaction terminals in the corresponding resource service processing scenarios.
[0019] During operation, the client primarily handles the following core processing steps: Resource transfer event generation and reporting: Resource transfer events are generated by various client devices in corresponding scenarios, such as computing resource scheduling events in computing resource scenarios, power load transfer events in power resource scenarios, carbon emission transfer events in carbon resource scenarios, and financial resource transfer events in financial service scenarios. Data generated by client devices at different nodes typically exhibits multi-source heterogeneity, and the formats of their internal time dimension tags (such as creation time, effective time, completion time, and settlement time) differ from time semantic standards. Client devices report these multi-source heterogeneous resource transfer event data to the service platform 100 in real-time or asynchronously via the communication network.
[0020] Prediction Task Request Initiation: Users or upper-level control systems can submit a target prediction task request to the service platform 100 through any client device. The target prediction task request carries resource scheduling prediction scenario parameters, which can be structured numerical configurations or scenario description text containing unstructured natural language (e.g., "assuming that computing power demand will increase by 20% next quarter and be accompanied by peak computing power consumption").
[0021] Prediction Results and Warning Reception: The client device receives the response message returned by the service platform 100. The response message encapsulates the predicted scenario resource flow data and the sequence of predicted resource inventory values on a unified time axis.
[0022] The service platform 100 internally comprises the following core processing procedures, either logically or in hardware: Connect to the communication network, responsible for receiving multi-source heterogeneous resource transfer event data from the client cluster, loading a preset time semantic transformation matrix, normalizing heterogeneous time tags into universally coordinated time (UTC) absolute timestamps on a unified time axis, and then encapsulating them into standardized data tuples (i.e., data transfer stream atomic units).
[0023] Obtain the target differential change data corresponding to the dynamic resource transfer event, and merge the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data; In response to a prediction task request initiated by a client device, based on resource scheduling prediction scenario parameters, data transfer inference processing is performed on the global resource transfer view data to obtain prediction scenario resource flow data. The prediction scenario resource flow data is aggregated within the target prediction window of the unified time axis to obtain net resource change amount, and a sequence of predicted resource inventory values is obtained based on the net resource change amount.
[0024] Furthermore, a response message is sent to the 100 client of the service platform.
[0025] It should be noted that the service platform 100 establishes a communication connection with at least one client in the client cluster for interactive communication via a network. This network can be a wireless network or a wired network. Wireless networks include, but are not limited to, cellular networks, wireless LANs, infrared networks, or Bluetooth networks. Wired networks include, but are not limited to, Ethernet, Universal Serial Bus (USB), or Controller Area Network (CAN). In one or more embodiments of the specification, technologies and / or formats including HyperText Markup Language (HTML), Extensible Markup Language (XML), etc., are used to represent data exchanged over the network (such as target compressed packets). Furthermore, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), and Internet Protocol Security (IPsec) can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can be used to replace or supplement the aforementioned data communication technologies.
[0026] The data processing system embodiments provided in this specification and the data processing methods described in one or more embodiments belong to the same concept. The execution entity corresponding to the data processing method involved in one or more embodiments of this specification can be the aforementioned service platform 100; the execution entity corresponding to the data processing method involved in one or more embodiments of this specification can also be the electronic device corresponding to the client, specifically determined based on the actual application environment. The implementation process of the data processing system embodiments can be detailed in the following method embodiments, and will not be repeated here.
[0027] based on Figure 1 The following is a detailed description of the data processing methods provided by one or more embodiments of this specification, as illustrated in the schematic diagram of the scenario.
[0028] Please see Figure 2 This document provides a flowchart illustrating a data processing method according to one or more embodiments. This method can be implemented using a computer program and can run on a data processing device based on the von Neumann architecture. The computer program can be integrated into an application or run as a standalone utility application. The data processing device can be a service platform.
[0029] Specifically, the data processing method includes: S102: Extract resource transfer event data of multi-source heterogeneous types, perform data atomization modeling on the resource transfer event data based on a unified time axis, and obtain data transfer flow atomic units; Multi-source heterogeneous resource transfer event data refers to raw event records generated by nodes distributed from different sources (such as various heterogeneous business servers, data acquisition terminals, distributed computing nodes, etc.), which have significant differences in data format, field definition, structural pattern, and time reference system, and are used to characterize the changes in the state, ownership, or location of a certain quantifiable resource (such as computing power resources, power resources, network bandwidth resources, financial service resources, etc.) between different nodes.
[0030] A unified time axis refers to a global time coordinate system established within the service platform (target computing node) with a unified reference benchmark and standard time scale system. It is used to eliminate time and data semantic confusion caused by heterogeneous source data due to clock deviation, time zone differences, or different sampling frequencies.
[0031] Data atomization modeling refers to the process of normalizing the temporal semantics, decoupling fields, and refining attributes of extracted raw heterogeneous event data, and reconstructing it into a data structure with definite spatiotemporal semantics and conforming to a standard format.
[0032] Data transfer flow atomic units refer to the core metadata generated after data atomization modeling. For example, in some embodiments, core metadata such as standardized time coordinates, resource transfer direction, and resource change values may be included, and these core metadata units serve as the smallest standardized data units within the computing node.
[0033] To illustrate, the service platform subscribes to and receives multiple resource transfer event data streams of various heterogeneous source node types. It extracts the original time identifiers and resource transfer parameters of each event data in the data stream, and maps the heterogeneous time identifiers to discrete time coordinate points corresponding to a unified time axis through time coordinate transformation. On this basis, the event data is decomposed into fields and reconstructed into attributes. The extracted core event features are encapsulated into standardized data tuples with globally unique identifiers and discrete time coordinates to construct the atomic units of the data transfer stream.
[0034] S104: Obtain the target differential change data corresponding to the dynamic resource transfer event, and merge the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data; Dynamic resource transfer events refer to events that occur in real time or frequently during the operation of resource data services, representing the latest dynamic changes in the status of resources, such as real-time resource data transfer-in / transfer-out events. Target differential change data refers to discrete or continuous differential data extracted or calculated based on the dynamic resource transfer event, used to reflect the incremental changes in resource data within a specific time node or time period.
[0035] Baseline data refers to a static or phased full resource transfer dataset pre-built in the target computing nodes maintained by the service platform, consisting of atomic units of historically processed data transfer flows, which serves as the initial reference benchmark for subsequent resource state evolution.
[0036] Global resource transfer view data refers to a data view that reflects the overall picture of global system resource transfer after aligning static historical baseline data and dynamic target differential change data on a unified time axis, comparing data versions, and overlaying and fusing numerical values.
[0037] In a schematic manner, the latest dynamic resource transfer events occurring within the resource data service are captured periodically or on demand through an event listening interface. The incremental resource change values and timestamps in the events are extracted to generate corresponding target differential change data. Baseline data composed of atomic units of data transfer streams from past historical events is read and mapped onto a unified time axis together with the target differential change data. Based on this, using the time scale on the unified time axis as a reference, the baseline data and target differential change data are matched for time coordinates and verified for versions. Dynamic incremental values at the same time coordinate point or the same time interval are superimposed onto the corresponding resource values of the baseline data, thereby completing the merging of the offline baseline and the real-time differential, and generating global resource transfer view data reflecting the dynamic flow of global resources.
[0038] S106: In response to the target prediction task request, based on the resource scheduling prediction scenario parameters, perform data transfer inference processing on the global resource transfer view data to obtain prediction scenario resource flow data, aggregate the prediction scenario resource flow data within the target prediction window of the unified time axis to obtain net resource change amount, and synthesize the net resource change amount to obtain a resource inventory prediction value sequence.
[0039] A target prediction task request refers to an instruction initiated by a user, terminal device, or upper-level control system to trigger simulation and prediction calculations of resource stock evolution. Resource scheduling prediction scenario parameters refer to the parameters or descriptive information carried in the request that express specific future scheduling assumptions, service change trends, control strategy adjustments, and external environmental disturbances.
[0040] Data transfer simulation processing refers to the operation process of applying data transfer simulation processing to existing global resource transfer view data based on resource scheduling prediction scenario parameters; Predictive scenario resource flow data refers to simulated resource transfer flows generated through extrapolation processing, used to characterize the resource flow status at various future time points under a specific predicted scenario.
[0041] The target forecast window refers to a pre-defined future time span or time series interval used for forecasting resource stock. Net resource change refers to the net increase or decrease in the forward and reverse resource transfer scalars after algebraic summation and cancellation at a specific time step or time bucket node within the target prediction window.
[0042] The resource stock prediction value sequence refers to a set of deterministic values that reflect the future evolution trend of the stock, derived by continuously accumulating and recursively extrapolating the net resource change at each time step according to the known resource stock benchmark value of the target resource node corresponding to the service platform at the initial moment, following the time sequence of a unified time axis.
[0043] Indicatively, in response to a target prediction task request sent by the client, resource scheduling prediction scenario parameters (such as control parameters used to express assumptions about future specific computing resource demand amplification, reduction, or advance / delay of computing resource tasks) contained in the request are extracted and parsed. By analyzing the resource scheduling prediction scenario parameters, scenario transition rules (mathematical control operators derived from the prediction scenario parameters to guide the scaling, addition, reduction, or replacement of computing power data scalars) and spatiotemporal offset coefficients (vector parameters used to determine the translation duration and direction of computing power data on a unified time axis) are determined. The scenario transition rules are applied to the computing power change values in the global resource transfer view data for recalculation, generating scenario adjustment vectors (vector data containing the direction and magnitude change increments generated by recalculating the original computing power scalars through the scenario transition rules). The scenario adjustment vectors are then mapped and superimposed onto the target time point on the unified time axis according to the spatiotemporal offset coefficients to generate prediction scenario resource flow data. Then, within the preset target prediction window (a pre-defined continuous future time span used to perform resource data inventory extrapolation and status prediction, such as the next 2 hours), it is divided into a time bucket sequence arranged in chronological order (a set of discrete time intervals divided within the target prediction window according to a preset fixed time granularity, such as 15 minutes per time bucket; within each time bucket, the predicted scenario resource flow data falling within that bucket is algebraically summed; the algebraic difference between the forward resource data request scalar and the reverse resource data release scalar is calculated to determine the net resource change amount corresponding to each time bucket (the pure increase or decrease value after algebraic summation and cancellation of the forward and reverse resource data scalars within a specific time bucket). The resource data processing service obtains the known available resource data inventory baseline value at the initial moment. According to the time sequence of the time buckets, the net resource change amount of each time bucket is sequentially added to the initial resource data inventory baseline value. The resource data inventory prediction value at the end of each time bucket is recursively calculated. Finally, a continuous resource inventory prediction value sequence is constructed (a set of continuous prediction state values reflecting the future evolution trend of resource data inventory is formed by continuously adding and recursively adding the net resource change amount of each time bucket according to the time sequence based on the initial inventory) and output.
[0044] For example, in the application scenario of intelligent computing center computing power resource scheduling, the target computing node maintained by the service platform is the data center computing power scheduling server; when the computing power scheduling server receives the target prediction task request sent by the control terminal, it first extracts and parses the resource scheduling prediction scenario parameters contained in the request (e.g., "assuming that due to a sudden artificial intelligence large model training task in the next 2 hours, the new computing power application demand will be scaled down by 1.5 times, and the release of existing computing power tasks will be delayed by 15 minutes"). The server performs data transfer simulation processing, applying the parsed scenario transformation rules (a scaling operator that multiplies the forward computing power request scalar by 1.5) to the computing power change values in the previously merged global computing power resource transfer view data. It then maps and overlays the recalculated change values onto the corresponding time nodes on the unified time axis according to the spatiotemporal offset coefficient (a vector parameter that shifts the timestamp of the reverse computing power release scalar by 15 minutes on a unified time axis), thus generating the predicted scenario computing power resource flow data. Within the target prediction window of the next 2 hours, the server divides the data into 8 time buckets (time bucket 1 to time bucket 8) arranged at 15-minute intervals. Within each time bucket, the server accumulates the forward computing power request scalar and the reverse computing power release scalar from the predicted scenario computing power resource flow data and calculates the algebraic difference to determine the net computing power change for each time bucket (e.g., the net computing power change for time bucket 1 is +600 cores, for time bucket 2 it is +300 cores, and for time bucket 3 it is -100 cores, etc.). Finally, the server reads the known available computing power baseline value of the data center at the initial moment (e.g., 10,000 cores) and performs discrete state recursion according to the chronological order of the time bucket sequence. The calculation involves adding 600 cores to the net computing power change of time bucket 1 and then adding it to the initial baseline value to obtain the predicted computing power inventory of 10,600 cores at the end of time bucket 1. Then, adding 300 cores to the net computing power change of time bucket 2 and then adding it to obtain the predicted computing power inventory of 10,900 cores at the end of time bucket 2. Finally, subtracting 100 cores from the net computing power change of time bucket 3 and then adding it to obtain the predicted computing power inventory of 10,800 cores at the end of time bucket 3, and so on, to derive the predicted computing power inventory at the end of all 8 time buckets. Finally, these predictions are combined to construct a sequence of predicted computing power inventory for each node every 15 minutes in the next 2 hours. This sequence is then encapsulated into a response message and output to the automated computing power scheduling engine, thus completing the extrapolation and inventory prediction processing.
[0045] This specification presents an atomic modeling mechanism for multi-source heterogeneous data based on a unified timeline, a view fusion mechanism combining static historical baselines and dynamic incremental differences, a scenario parameter-driven spatiotemporal transfer extrapolation mechanism, and a bucket aggregation and continuous inventory recursion mechanism within the prediction window. These mechanisms address the problems of traditional resource data processing systems when dealing with multi-source heterogeneous time-series data, such as high temporal semantic alignment overhead, high latency in offline and real-time data fusion, and the inability to quickly and accurately recursively extrapolate inventory evolution trends during scenario simulation and prediction. The system achieves standardized mapping and high-concurrency dynamic fusion of heterogeneous event data from different sources, transforming complex prediction extrapolation and inventory state calculations into view fusion on a unified timeline. This reduces system computational complexity and power consumption, enabling the rapid, smooth, and accurate extrapolation of continuous resource inventory prediction sequences within any target prediction window, thus supporting efficient resource data transfer processing.
[0046] Please see Figure 3 , Figure 3 This is a schematic diagram of a data atomization modeling process proposed in one or more embodiments of this specification. Specifically, the execution of data atomization modeling based on a unified time axis on the resource transfer event data to obtain data transfer flow atomic units can be referenced as follows: S202: Perform temporal semantic alignment on the resource transfer event data on a unified timeline to obtain aligned event data; A unified time axis refers to a global time coordinate system with a globally unified reference benchmark and standard time scale system (such as Coordinated Universal Time, UTC), which is used as a unified scale to eliminate clock deviations and sampling accuracy differences in multi-source heterogeneous data.
[0047] Temporal semantic alignment refers to the process of eliminating the time scale disorder caused by the acquisition clock deviation, time zone definition difference or sampling accuracy difference (such as second level and millisecond level) of multi-source heterogeneous data, and uniformly mapping and normalizing them to a global reference system. Aligned event data refers to event data that has been aligned with time semantics and has definite and standard time coordinates.
[0048] To illustrate, the initial time identifier and the corresponding source system node identifier carried in the resource transfer event data are first extracted. The time semantic transformation mapping dictionary in local persistent storage is read, and the corresponding time zone offset, time format parsing benchmark, and sampling precision rules are matched according to the source system node identifier. Using the standard time scale on the unified time axis as a reference, the initial time identifier is converted into a standard time scale point on the unified time axis through format parsing, time zone difference compensation, and precision scaling calculation. The converted standard time scale point is then filled back into the resource transfer event data to replace the original initial time identifier, thereby completing the time semantic alignment on the unified time axis and generating aligned event data with a unified time scale.
[0049] Optionally, to specifically perform time semantic alignment on the resource transfer event data on a unified timeline to obtain aligned event data, the following methods can be used: Step C2: Load a preset time semantic transformation matrix based on the source attribute of the resource transfer event data; Source attributes refer to the identifying information that characterizes the physical or logical features of the source of resource transfer event data, such as the time zone, operating system environment, underlying data acquisition protocol, and system version identifier of the source node.
[0050] The time semantic transformation matrix refers to a pre-built and configured dictionary that records the mapping relationship between different source attributes and the global unified standard. It may contain time zone compensation difference, time format parsing mask, and unit conversion coefficient of resource scalar.
[0051] Step C4: Using the time semantic transformation matrix, convert the different time dimension labels in the resource transfer event data into a universal coordinated timestamp, and convert the resource transfer parameters in the resource transfer event data into standard resource transfer data; Different time dimension tags refer to the time stamps attached to events by heterogeneous source systems. For example, they may include local time strings with time differences, counters with different precision (such as microseconds, milliseconds or seconds), and time with different business dimensions (such as the time of request occurrence and the time recorded by the system).
[0052] The Coordinated Universal Time Stamp (UTC) is an absolute numerical time expression that is unaffected by geographical time zones and is calculated from a unified reference time system.
[0053] Step C6: Based on the preset global time granularity, perform time sampling on the general coordination timestamp to obtain the event time point on the unified time axis, and align the standard resource transfer data to the event time point to obtain aligned event data.
[0054] Resource transfer parameters refer to the quantitative indicators of resources recorded in the original data; Indicatively, steps C2-C6 are executed. After obtaining the original resource transfer event data, the message header metadata is extracted and parsed. Based on the source attributes of the resource transfer event data (referring to the identification information characterizing the source of the data, including the source's geographical time zone, system architecture type, and interface protocol version number), a preset time semantic transformation matrix (a two-dimensional mapping table or mathematical matrix recording time zone difference compensation, format parsing rules, and dimensional conversion coefficients between heterogeneous formats of different source systems and the global standard system) is loaded. The time semantic transformation matrix is used to perform clock offset operations and numerical type casting to accurately convert the different time dimension labels carried in the resource transfer event data into a Universal Coordinated Time (UTC) timestamp. The system uses an absolute numerical timestamp calculated from a zero-point reference. Simultaneously, through the unit conversion factor recorded in the matrix, the resource transfer parameters (the non-standard format resource change direction and original quantity fields embedded in the original event) in the resource transfer event data are transformed into standard resource transfer data with unified dimensions and defined signs. Based on the preset global time granularity, the general coordination timestamp is sampled (using mathematical truncation algorithms such as rounding down and modulo to remove extra time precision tails that exceed the global time granularity, and discretizing consecutive timestamps to the nearest valid step size node). This yields event time points that fully conform to the time scale specification and fall on a unified time axis. The standard resource transfer data is precisely mounted and aligned to the event time points, thus combining them to obtain aligned event data that eliminates time sequence disorder and scalar ambiguity.
[0055] S204: Based on the preset metadata field configuration, the alignment event data is encapsulated into standardized data tuples to obtain data transfer stream atomic units.
[0056] Predefined metadata field configurations refer to standardized templates or constraint rules that are predefined or dynamically loaded in the target compute node to regulate the data storage structure and field mapping relationships. This configuration specifies the set of fields in the data structure, the order of the fields, the data types, and the validation rules. Standardized data tuples refer to fixed-format, multi-dimensional structured data formed by filling and encapsulating the decoupled event backbone features according to the field configuration.
[0057] In a schematic manner, the system reads the preset metadata field configuration, which defines the constraints of each field in the standardized data model in the form of a structured dictionary. The data extraction engine is then activated to traverse and scan the alignment event data, stripping away and discarding redundant log headers or system-level debugging information unrelated to resource transfer. The system accurately extracts the core business fields, such as resource flow direction, core resource values, and aligned standard time coordinates. A fixed-length array or structure is instantiated, and the extracted core business fields are mapped and filled into their corresponding slots according to the index order and type requirements specified in the metadata field configuration, generating standardized data tuples with a uniform format. Finally, a globally unique identifier generated by a hash algorithm is appended to this standardized data tuple, formally registering and encapsulating it as an atomic unit of the data transfer flow.
[0058] This specification addresses the challenge of unifying and merging multi-source heterogeneous data due to clock skew and format inconsistencies by performing temporal semantic alignment on a unified timeline and encapsulating it based on metadata field configuration. This eliminates temporal semantic ambiguity across data sources, ensuring absolute accuracy of the global timeline. Simultaneously, it reconstructs loosely structured raw events into standardized data transfer stream atomic units, significantly reducing data parsing overhead at downstream nodes and improving the efficiency and stability of high-concurrency access to massive amounts of heterogeneous data.
[0059] Optional, please refer to Figure 4 , Figure 4 This is a flowchart illustrating a data merging process. Specifically, it involves acquiring the target differential change data corresponding to the dynamic resource transfer event, and merging the baseline data composed of the atomic units of the data transfer stream with the target differential change data on the unified time axis to obtain global resource transfer view data. The following method can be used as a reference: S302: Read the baseline data composed of the data transfer stream atomic units in the historical time period; The historical time period refers to a preset time interval before the current view construction calculation is triggered. The data state within this interval is relatively stable and is used to delineate the time boundaries between the static historical state and the dynamic incremental state of the system.
[0060] S304: Capture dynamic resource transfer events from multiple heterogeneous data processing source nodes, and generate target differential change data based on the dynamic resource transfer events; S306: Global resource transfer view data is obtained by merging the baseline data and the target differential variation data on the unified time axis.
[0061] Indicatively, steps S302-S306 are executed. First, view construction is initiated. Baseline data (a static full resource transfer dataset accumulated and solidified on disk by the system during the historical time period) composed of atomic units of the data transfer flow is read through the distributed database interface. Then, the distributed log capture engine is started, and dynamic resource transfer events (system event data newly generated in the current running cycle representing the immediate change in the allocation or release status of computing power resources) are captured in real time from multiple heterogeneous data processing source nodes using change data capture technology. The core resource scalar increase / decrease values are extracted, and target differential change data is generated based on the dynamic resource transfer events. Using a preset global universal time coordinate system as a reference, the baseline data and target differential change data are merged on a unified time axis (by comparing consistent time coordinate nodes, the scalar values of the differential change data are algebraically superimposed onto the corresponding positions of the baseline data in chronological order, and the version number is used to eliminate duplicate merging). This accurately reconstructs and generates global resource transfer view data that can completely reflect the current point in time and the full resource flow status of the system in the future.
[0062] In this specification, steps S302-S306 are executed. By reading the baseline data fixed within a historical time period and capturing dynamic resource transfer events from multiple heterogeneous source nodes in real time to generate target differential change data, the static baseline and dynamic differential change data are accurately merged and reconstructed on a unified time axis. This achieves seamless, low-latency dynamic fusion of historical offline base and real-time incremental status. While avoiding the reconstruction of the entire underlying data and frequent refresh and reconstruction of physical indexes, it ensures the absolute real-time nature and temporal consistency of resource flow status. It also provides a highly reliable, panoramic unified view data base to support subsequent accurate resource scenario extrapolation and stock prediction.
[0063] Optionally, the following methods can be used to implement S306: Step C200: Establish a circular buffer version bitmap for the time bucket sequence on the unified time axis; A time bucket sequence on a unified time axis refers to a set of equidistant time intervals continuously divided along a unified general time coordinate axis in the memory of the target computing node according to a fixed discrete time step, in order to slice the continuous resource state into time slices.
[0064] A circular buffer version bitmap refers to a contiguous block of memory that is connected end to end. Using a binary bit array format, it records the current version status, update flag, or validity identifier of data within each time bucket, bit by bit. Its circular structure ensures that the memory space can be reused cyclically as time progresses without frequent reallocation.
[0065] Step C204: Perform bitmasking filtering on the target differential variation data based on the version bitmap of the circular buffer to obtain the filtered target differential variation data; Bitwise masking filtering refers to the technical process of using a preset binary mask and the version / status feature code carried in the target differential change data to perform bitwise logical operations (such as AND, OR, XOR) to quickly remove expired, out-of-order, or conflicting redundant differential records.
[0066] Step C206: Generate a time bucket index offset vector based on the filtered target differential change data, and superimpose the time bucket index offset vector onto the index slot of the baseline data in the corresponding time bucket in an unlocked state to obtain the global resource transfer view data.
[0067] The time bucket index offset vector refers to the vector data entity that combines the scalar form of differential change data (pure increment and decrement values) with the timestamp coordinates of its occurrence, and transforms it into a location pointer pointing to the storage address space of a specific time bucket in memory and a value increment.
[0068] Lock-free state refers to the use of hardware-supported atomic operations (such as compare and swap) to replace traditional mutex or spinlock mechanisms in a multi-threaded, high-concurrency environment. An index slot refers to the precise memory storage address unit in memory corresponding to a specific time bucket for baseline data.
[0069] Indicatively, steps C202-C206 are executed. First, based on the global time span of the unified time axis and the preset fixed discrete time step, a fixed-size array space is continuously allocated to establish a circular buffer version bitmap for the time bucket sequence on the unified time axis. This bitmap uses multi-bit binary tuples to accurately map the version number and write status of the merged data in the corresponding time bucket. When the target computing node receives the target differential change data to be merged, it extracts the event version stamp of the change data and performs a bitwise logical AND operation with the corresponding time bucket bit array in the circular buffer version bitmap. By performing bitwise mask filtering on the target differential change data based on the circular buffer version bitmap, the computing node intercepts the data in real time at the hardware instruction level. Expired version data or duplicate pushed dirty data caused by network latency are truncated and discarded, thus obtaining the filtered target differential change data with valid version and correct time sequence. The pure increase / decrease scalar of resources and its corresponding timestamp coordinates are extracted from the filtered target differential change data, and a time bucket index offset vector directly pointing to the memory baseband address is generated using a hash address mapping algorithm. The underlying atomic update instruction is called to accurately accumulate and superimpose the resource increase / decrease scalar carried by the time bucket index offset vector into the index slot of the baseline data on the corresponding time bucket in a lock-free state without blocking any parallel computing threads. In this way, while ensuring extremely high concurrency throughput, the absolutely consistent and time-sequential global resource transfer view data is reconstructed and generated.
[0070] In this specification, based on steps C2-C6, a ring-shaped version bitmap is constructed for bit-mask filtering, and differential data is superimposed using index offset vectors in a lock-free state, thus resolving the lock contention and out-of-order conflicts during massive data merging. This mechanism achieves instantaneous interception of dirty data and lock-free, ultra-fast fusion of incremental data, significantly improving concurrent throughput and ensuring low latency and temporal consistency in view reconstruction.
[0071] Optional, please refer to Figure 5 , Figure 5 This is a flowchart illustrating a data transfer and extrapolation process. Specifically, based on resource scheduling prediction scenario parameters, data transfer and extrapolation processing is performed on the global resource transfer view data to obtain predicted scenario resource flow data, including: S402: Analyze the resource scheduling prediction scenario parameters of the target prediction task request, and determine the scenario transition rules and corresponding spatiotemporal offset coefficients within the target time interval; The target time interval refers to the continuous time span (e.g., between 2 and 4 hours in the future) defined in the target prediction task request, which requires the system to perform simulation calculations and extrapolation analysis for a specific future or historical replay.
[0072] Scenario transition rules refer to the mathematical calculation rules and operators extracted based on the business change logic assumed by the prediction task (such as load surge, policy rate limiting, etc.) to guide the original resource transfer scalar to perform numerical scaling, proportional increase or decrease, or function replacement.
[0073] The spatiotemporal offset coefficient refers to a quantitative parameter that characterizes the timing of resource transfer events occurring on a unified time axis, including the time of advance or delay, as well as the logical location migration of resource scheduling source / destination nodes.
[0074] Indicatively, step S402 is executed to parse the resource scheduling prediction scenario parameters (structured or semi-structured configuration information input by the user or upper-level control system to characterize future sudden business loads, strategy adjustments, or external environmental disturbances) of the target prediction task request, thereby accurately determining the scenario transformation rules (mathematical transformation operators used to guide the original resource transfer scalars to perform proportional scaling, fixed value increase / decrease, and function substitution) and the corresponding spatiotemporal offset coefficients (characteristic scalar parameters used to indicate the advance or delay of resource transfer events on a unified time axis and the location of the transfer).
[0075] Optionally, the specific execution of parsing the resource scheduling prediction scenario parameters to determine the scenario transition rules and corresponding spatiotemporal offset coefficients within the target time interval can refer to the following methods: Step D2: Obtain the scenario description text from the target prediction task request; Context description text refers to the character sequence contained in the target prediction task request, written in natural language, pseudocode, semi-structured script, etc., to describe the intent of future business assumptions, sudden traffic surges, changes in the external environment, or adjustments to scheduling strategies.
[0076] Step D4: Based on the scenario description text, use the scenario parsing large model to extract scenario semantic vectors, and map the scenario semantic vectors to the spatiotemporal resource state space for vector solution to obtain scenario transformation rules and spatiotemporal offset coefficients that satisfy resource transfer space constraints.
[0077] Contextual parsing large models refer to large models (LLMs) that are deployed on computing nodes or accessed through the cloud, based on machine learning network architectures (such as Transformer), and pre-trained and fine-tuned using massive resource scheduling domain corpora.
[0078] Context semantic vectors refer to the high-dimensional continuous floating-point feature arrays output by the context parsing big model after feature encoding of the context description text, which are used to represent abstract scheduling intentions.
[0079] The spatiotemporal resource state space refers to a multi-dimensional modeling space constructed in the memory of the target computing node, which includes the time axis dimension, the spatial topology node dimension, and the resource capacity constraint dimension. It is used to digitally map the boundaries and topological relationships of real physical resource networks.
[0080] Resource transfer space constraints refer to objectively existing, insurmountable boundary restrictions in real physical networks, such as the maximum capacity of a single physical server, the maximum concurrent bandwidth threshold of a network link, and the rigidly defined latest response latency baseline.
[0081] Schematic, step D2 involves parsing the data packet payload of the received target prediction task request. Through data cleaning and deserialization, the context description text within the request is extracted. Step D4 then invokes the deployed context parsing model based on the context description text. The underlying natural language understanding network (a deep neural network text encoder built on a multi-layer self-attention architecture) of this model performs lexical decomposition and multi-dimensional feature encoding on the text sequence, autonomously extracting the context semantic vector. The context parsing model then calls its integrated state space mapping layer to directly... The scenario analysis model is then projected into the spatiotemporal resource state space constructed within the model. Within this model's internal space, the scenario analysis model autonomously performs vector computation using its built-in neural decoder network (e.g., using tensor multiplication and nonlinear activation functions to perform dimensionality reduction projection optimization on high-dimensional feature tensors, and filtering illegal solutions through an end-to-end deep inference process). Finally, its output layer directly outputs scenario transition rules and spatiotemporal offset coefficients that have been verified by the model's internal logic and absolutely satisfy resource transfer space constraints (objective boundary constraints in the physical resource network, such as the physical upper limit of single-node resource capacity).
[0082] For example, let's take a look at the scenario of intelligent inference of cross-node computing power resources in cloud data centers. Suppose the scenario description text is "It is predicted that there will be an extremely high concurrency of user requests at 2 pm tomorrow. Under the premise of ensuring zero latency of computing power response of core service links, we should reduce the occupation of memory object cache pool by non-critical data analysis tasks and avoid computing power hotspot congestion as much as possible." Step D2 involves directly packaging this highly abstract natural language text into the context analysis model. Step D4: The context analysis model first uses its internal encoding layer to extract the natural language into a high-dimensional context semantic vector (this vector implicitly and continuously contains complex scheduling intentions such as extreme concurrency, core zero latency, and pressure reduction of non-critical caches). The hidden layer inside the model directly maps this semantic vector to its predefined "spatiotemporal resource state space." This space integrates the physical topology state of the current computing cluster and comes with strict resource transfer space constraints. The decoder inside the model automatically performs vector computation. Faced with non-quantized input, the model performs gradient optimization within the constraint space, automatically transforming the macroscopic intentions into underlying topology reconstruction and resource scheduling operators. Finally, after black-box feature space optimization calculations within the model, the model directly outputs the following without exceeding the physical boundaries: The calculated scenario transition rules output a cost function model for dynamic resource reduction, thereby reconstructing the operation weights of the memory object cache pool and reducing the concurrent request volume of its computing resources from the underlying mathematical logic. The calculated spatiotemporal offset coefficient: For secondary computing tasks that generate congestion conflicts, a non-linear delay offset vector is calculated and output based on the longest path boundary of the topology sorting of the underlying tasks. This controls the relevant computing tasks to strictly perform gradient shifts according to the topology hierarchy on a unified time axis, thereby perfectly smoothing the computing power peaks without destroying the original task dependencies.
[0083] Optionally, the following illustrates the model training process for a large-scale scenario analysis model: Model creation: Obtain the basic large language model and create an initial large scenario parsing model for the scenario parsing scenario. The basic large language model (MLLM) includes, but is not limited to, the DeepSeek large model, the GPT series large models, etc.
[0084] 1) Construct a historical scenario scheduling training sample set: Extract a large amount of historical scheduling task data from historical resource scheduling logs and the work order system. For each historical scheduling task, extract the corresponding natural language scheduling requirement text input from the front end and use it as the training input sample; at the same time, extract the scheduling execution policy records of the system that were actually successfully issued to the system for the requirement text in a historical moment without triggering any node resource data capacity alarms, and parse out the corresponding real-scenario transformation rules (e.g., the actual computing power reduction ratio scalar, denoted as R). true And the actual temporary offset coefficient (e.g., the actual task delay time, denoted as T). true This is used as the true supervisory label for the input sample. Through this method, the target computing node constructs a historical scene scheduling training sample set containing a massive number of "text-rule and offset" key-value pairs.
[0085] 2) Forward Propagation and Prediction Parameter Generation: After obtaining the historical scene scheduling training sample set, the input samples are batch-wise input into the initial scene parsing large model, which is in a non-converged state. The front-end encoder of the initial large model performs lexicalization on the input natural language scheduling request text and extracts the hidden layer semantic feature matrix through a multi-head self-attention mechanism. This feature matrix is input into the regression decoder at the end of the large model. Based on the network weights and bias parameters of the current round, the decoder performs fully connected mapping and numerical decoding, and outputs a set of prediction results, specifically including the predicted scene transformation rules and predicted spatiotemporal offset coefficients predicted by the current model.
[0086] 3) Loss Calculation: To quantify the deviation between the predicted parameters output by the large model and the true supervision labels, a pre-defined loss calculation formula (such as cross-entropy loss, Euclidean distance loss, and hinge loss) is used for error calculation. For example, the computation node calculates the mean squared error (MSE) for regression tasks with two different dimensions: rule scaling ratio and time offset duration, and then linearly combines them using pre-defined weighting coefficients. The loss function formula is as follows:
[0087] Where Loss is the total global loss value for the current batch of training, and N represents the total number of samples in the current training batch. and Let represent the predicted scenario transition rule parameters and the actual rule parameters for the i-th sample, respectively. and denoted as the predicted spatiotemporal offset coefficient and the actual offset coefficient, respectively, with a and b being pre-set empirical weight hyperparameters.
[0088] 4) Backpropagation and iterative update of network weights: Adjust the model parameters of the initial scenario parsing model based on the global total loss value until the model training termination condition is met to obtain the scenario parsing model.
[0089] Optionally, the model's training termination conditions may include, for example, the loss function value being less than or equal to a preset loss function threshold, or the number of iterations reaching a preset threshold. Specific training termination conditions can be determined based on actual circumstances and are not specifically limited here.
[0090] In this specification, based on steps D2-D4, macroscopic scenario semantics are extracted from a large model, and high-dimensional vector optimization is performed in the spatiotemporal state space. This solves the problem that traditional rule matching cannot parse non-quantitative abstract intentions and generate high-order low-level scheduling strategies. This mechanism realizes intelligent translation from natural language to complex cost functions and nonlinear offset operators. While ensuring that the inference parameters strictly adhere to physical boundary constraints, it significantly improves the decision-making accuracy and intelligence level of resource scheduling in complex and sudden scenarios.
[0091] S404: Based on the scenario transition rule, the resource change amount of the data transfer flow atomic unit corresponding to the time node in the global resource transfer view data is recalculated to obtain the scenario adjustment vector. The scenario adjustment vector is mapped and superimposed onto the target time point of the unified time axis according to the spatiotemporal offset coefficient to obtain the predicted scenario resource flow data.
[0092] The scenario adjustment vector refers to the intermediate vector entity generated after recalculating the scalar values in the original data transfer stream atomic units by applying scenario transformation rules. This vector entity contains a completely new level of resource change and spatiotemporal transformation characteristics.
[0093] Predictive scenario resource flow data refers to a complete global resource transfer data view that is reconstructed by mapping the scenario adjustment vector to a specific time node on a unified time axis according to the spatiotemporal offset coefficient. This view reflects the evolution trend of a specific hypothetical predicted environment.
[0094] Indicatively, the simulation calculation pipeline is initiated to execute step S404. Using scenario transformation rules, the resource change amounts of the data transfer flow atomic units falling into the corresponding time nodes in the global resource transfer view data are recalculated. This involves performing algebraic operations on the initial resource increase / decrease scalars carried by the atomic units and the aforementioned mathematical transformation operators to recalculate the scenario adjustment vector (an intermediate vector data containing new resource increment values and their spatiotemporal mapping attributes, generated after numerical reconstruction). The scenario adjustment vector is then subjected to time coordinate addition and subtraction operations according to the aforementioned spatiotemporal offset coefficient, precisely mapping and superimposing it onto the target time point of the unified time axis (the landing point scale redefined in the global time coordinate system after time translation). This reconstructs the predicted scenario resource flow data that fully conforms to the evolution trend of a specific hypothetical environment, for subsequent aggregation calculations.
[0095] In this specification, based on S402-S404, scenario transition rules and spatiotemporal offset coefficients are extracted by parsing the prediction task requests. Based on these, resource change recalculation and time coordinate mapping translation are directly performed on the transfer flow atomic units in the global view data. The complex scenario evolution and scheduling strategy adjustment are reduced in dimensionality and transformed into scalar value reset and discrete time coordinate point displacement on a unified time axis. This reduces the underlying computational complexity and memory consumption of spatiotemporal simulation, and realizes high-concurrency, low-latency simulation for sudden loads or limit control strategies.
[0096] Optional, please refer to Figure 6 , Figure 6 This is a flowchart illustrating an aggregation process. Specifically, it involves aggregating the resource stream data of the predicted scenario within the target prediction window on the unified time axis to obtain the net resource change amount, and then synthesizing the net resource change amount to obtain a sequence of predicted resource inventory values, including: S502: Within the target prediction window of the unified time axis, determine the time bucket sequence arranged in chronological order; The target prediction window refers to the future continuous time span (e.g., the next 2 hours or the next 24 hours) that the system sets based on the scheduling prediction task request and requires simulation and prediction of the evolution of resource stock.
[0097] A time bucket sequence refers to a set of discrete time intervals arranged in chronological order, formed by dividing a continuous time axis into equal parts according to a preset discrete time step (such as 15 minutes or 1 hour).
[0098] S504: Within each time bucket of the time bucket sequence, perform a comprehensive calculation on the resource inflow change and resource outflow change in the predicted scenario resource flow data falling within the time bucket to obtain the net resource change corresponding to each time bucket; Resource inflow changes refer to the sum of the scalar increases in resource data caused by positive operations such as new applications, allocations, or transfers within a specific time bucket; Resource outflow change refers to the sum of scalar reductions in resource data caused by reverse operations such as release, recycling, or destruction.
[0099] Net resource change refers to the pure increase or decrease in resource data obtained after offsetting (algebraic summation) the changes in resource inflow and outflow within a specific time bucket.
[0100] In illustratively, step S504 involves performing a full scan of the predicted scenario resource flow data falling within the boundary of each time bucket in the time bucket sequence. This scan aggregates and statistically analyzes the resource inflow change (a set of resource values with a positive increase indicator) and the resource outflow change (a set of resource values with a negative decrease indicator). A comprehensive calculation is then performed on both to accurately obtain the net resource change for each time bucket (a unique net scalar value representing the absolute increase or decrease trend of resources within that time period, obtained after comprehensive calculation). S506: Obtain the current initial resource inventory of the target resource node, and sequentially add the net resource change amount of each time bucket to the current initial resource inventory according to the time sequence of the time buckets, so as to obtain the resource inventory prediction value at the end of each time bucket, and obtain the resource inventory prediction value sequence based on all resource inventory prediction values.
[0101] The current initial resource stock refers to the physical baseline value of the available resources actually owned or remaining by the target resource node at the beginning zero point of the target prediction window.
[0102] The resource stock prediction value sequence refers to a set of continuous values that reflect the remaining state of resources at each discrete time node in the future, generated by superimposing the net change of each time bucket in the direction of time progression based on the initial resource stock.
[0103] Indicatively, step S506 involves querying the underlying resource status database to obtain the current initial resource inventory of the target resource node. A status accumulator is then initialized, and the net resource change amount of each time bucket is sequentially added to the current initial resource inventory according to the time sequence of the time buckets. This process recursively yields the predicted resource inventory value at the end of each time bucket (after each accumulation operation, the corresponding time node's resource inventory snapshot value is overwritten in memory). Finally, these snapshot values are combined and encapsulated according to their time-series index. Based on all the predicted resource inventory values, a sequence of predicted resource inventory values is obtained to guide future macro-level resource scheduling, completing the closed loop from micro-event aggregation to macro-level inventory prediction.
[0104] The following will combine Figure 7 This manual provides a detailed description of the data processing device provided. It should be noted that... Figure 7 The data processing apparatus shown is used to execute this specification. Figures 1-6 The methods of the embodiments shown are illustrated only in connection with this specification for ease of explanation. For specific technical details not disclosed, please refer to this specification. Figures 1-6 The example shown.
[0105] Please see Figure 7This diagram illustrates the structure of the data processing device described in this specification. The data processing device 1 can be implemented as all or part of a user's electronic device through software, hardware, or a combination of both. According to some embodiments, the data processing device 1 includes an atomic modeling module 11, a global merging module 12, and a request-response module 13, specifically used for: Atomic modeling module 11 is used to extract resource transfer event data of multi-source heterogeneous types, perform data atomization modeling on the resource transfer event data based on a unified time axis, and obtain data transfer flow atomic units; The global merging module 12 is used to obtain the target differential change data corresponding to the dynamic resource transfer event, and merge the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data. The request response module 13 is used to respond to the target prediction task request, determine the resource scheduling prediction scenario parameters, perform data transfer inference processing on the global resource transfer view data to obtain the prediction scenario resource flow data, aggregate the prediction scenario resource flow data within the target prediction window of the unified time axis to obtain the net resource change amount, and perform comprehensive analysis based on the net resource change amount to obtain the resource stock prediction value sequence.
[0106] It should be noted that the data processing apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when executing the data processing method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the data processing apparatus and the data processing method embodiments provided in the above embodiments belong to the same concept, and the implementation process can be found in the method embodiments, which will not be repeated here.
[0107] The serial numbers in this specification are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0108] This specification also provides a computer storage medium capable of storing multiple instructions adapted to be loaded and executed by a processor as described above. Figures 1-6 The data processing method described in the illustrated embodiment can be found in the following document for a detailed execution process. Figures 1-6 The specific details of the illustrated embodiments will not be elaborated here.
[0109] This specification also provides a computer program product that stores at least one instruction, said at least one instruction being loaded and executed by the processor as described above. Figures 1-6 The data processing method described in the illustrated embodiment can be found in the following document for a detailed execution process. Figures 1-6The specific details of the illustrated embodiments will not be elaborated here.
[0110] Please refer to Figure 8 This is a structural block diagram of an electronic device provided in an embodiment of this specification. The electronic device in this specification may include one or more of the following components: a processor 1010, a memory 1020, an input device 1030, an output device 1040, and a bus 1050. The processor 1010, memory 1020, input device 1030, and output device 1040 may be connected to each other via the bus 1050.
[0111] Processor 1010 may include one or more processing cores. Processor 1010 connects to various parts of the electronic device using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 1020, and by calling data stored in memory 1020. Optionally, processor 1010 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). Processor 1010 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem handles wireless communication. It is understood that the modem may also not be integrated into processor 1010 and may be implemented separately using a communication chip.
[0112] The memory 1020 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 1020 may include a non-transitory computer-readable medium. The memory 1020 may be used to store instructions, programs, code, code sets, or instruction sets.
[0113] The input device 1030 is used to receive input instructions or data, and includes, but is not limited to, a keyboard, mouse, camera, microphone, or touch device. The output device 1040 is used to output instructions or data, and includes, but is not limited to, a display device and a speaker. In this embodiment, the input device 1030 can be a temperature sensor for acquiring the operating temperature of the electronic device. The output device 1040 can be a speaker for outputting audio signals.
[0114] In addition, those skilled in the art will understand that the structure of the electronic device shown in the above figures does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, the electronic device may also include radio frequency circuits, input units, sensors, audio circuits, Wireless Fidelity (WIFI) modules, power supplies, Bluetooth modules, etc., which will not be described in detail here.
[0115] In the embodiments of this specification, the executing entity for each step can be the electronic device described above. Optionally, the executing entity for each step can be the operating system of the electronic device. The operating system can be Android, iOS, or other operating systems; this specification does not limit this.
[0116] exist Figure 8 In the electronic device, the processor 1010 can be used to call a program stored in the memory 1020 and execute it to implement the data processing method as described in the various method embodiments of this specification.
[0117] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.
[0118] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in the embodiments of this specification are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the features, data, and information involved in this specification were all obtained under full authorization.
[0119] The above-disclosed embodiments are merely preferred embodiments of this specification and should not be construed as limiting the scope of this specification. Therefore, any equivalent variations made in accordance with the claims of this specification shall still fall within the scope of this specification.
Claims
1. A data processing method, characterized by, The method includes: Extract resource transfer event data from multiple sources and heterogeneous types, and perform data atomization modeling on the resource transfer event data based on a unified time axis to obtain data transfer flow atomic units; Obtain the target differential change data corresponding to the dynamic resource transfer event, and merge the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data; In response to the target prediction task request, based on the resource scheduling prediction scenario parameters, data transfer inference processing is performed on the global resource transfer view data to obtain the prediction scenario resource flow data. The prediction scenario resource flow data is aggregated within the target prediction window of the unified time axis to obtain the net resource change amount, and a sequence of resource stock prediction values is obtained based on the net resource change amount.
2. The method of claim 1, wherein, The process of performing data atomization modeling on the resource transfer event data based on a unified time axis to obtain data transfer flow atomic units includes: The resource transfer event data is aligned temporally and semantically on a unified timeline to obtain aligned event data. Based on the preset metadata field configuration, the alignment event data is encapsulated into standardized data tuples to obtain data transfer stream atomic units.
3. The method of claim 2, wherein, The step of performing temporal semantic alignment on the resource transfer event data on a unified time axis to obtain aligned event data includes: A preset time semantic transformation matrix is loaded based on the source attribute of the resource transfer event data; Using the time semantic transformation matrix, the different time dimension labels in the resource transfer event data are converted into a universal coordinated timestamp, and the resource transfer parameters in the resource transfer event data are converted into standard resource transfer data; The general coordination timestamp is sampled according to the preset global time granularity to obtain the event time point on the unified time axis. The standard resource transfer data is then aligned to the event time point to obtain aligned event data.
4. The method of claim 1, wherein, The process of acquiring the target differential change data corresponding to the dynamic resource transfer event, and merging the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data, includes: Read baseline data composed of the data transfer stream atomic units from the historical time period; Dynamic resource transfer events are captured from multiple heterogeneous data processing source nodes, and target differential change data is generated based on the dynamic resource transfer events. Global resource transfer view data is obtained by merging the baseline data and the target differential variation data on the unified time axis.
5. The method of claim 4, wherein, The process of merging the baseline data and the target differential variation data on the unified time axis to obtain global resource transfer view data includes: Establish a circular buffer version bitmap for the time bucket sequence on the unified time axis; The target differential variation data is filtered by bitmasking based on the version bitmap of the circular buffer to obtain the filtered target differential variation data. Based on the filtered target differential change data, a time bucket index offset vector is generated, and in an unlocked state, the time bucket index offset vector is superimposed onto the index slot of the baseline data in the corresponding time bucket to obtain the global resource transfer view data.
6. The method of claim 1, wherein, The process of performing data transfer inference processing on the global resource transfer view data based on resource scheduling prediction scenario parameters to obtain prediction scenario resource flow data includes: Analyze the resource scheduling prediction scenario parameters of the target prediction task request to determine the scenario transition rules and corresponding spatiotemporal offset coefficients within the target time interval; Based on the scenario transition rules, the resource change amount of the data transfer flow atomic units at the corresponding time nodes in the global resource transfer view data is recalculated to obtain the scenario adjustment vector. The scenario adjustment vector is then mapped and superimposed onto the target time point of the unified time axis according to the spatiotemporal offset coefficient to obtain the predicted scenario resource flow data.
7. The method of claim 6, wherein, The step of parsing the resource scheduling prediction scenario parameters to determine the scenario transition rules and corresponding spatiotemporal offset coefficients within the target time interval includes: Obtain the scenario description text from the target prediction task request; Based on the scenario description text, a scenario parsing big model is used to extract scenario semantic vectors, and the scenario semantic vectors are mapped to the spatiotemporal resource state space for vector calculation to obtain scenario transformation rules and spatiotemporal offset coefficients that satisfy resource transfer space constraints.
8. The method of claim 1, wherein, The process of aggregating the resource flow data of the predicted scenario within the target prediction window of the unified time axis to obtain the net resource change amount, and then synthesizing the resource stock prediction value sequence based on the net resource change amount, includes: Within the target prediction window of the unified time axis, determine the time bucket sequence arranged in chronological order; Within each time bucket of the time bucket sequence, a comprehensive calculation is performed on the resource inflow change and resource outflow change in the predicted scenario resource flow data falling within the time bucket to obtain the net resource change corresponding to each time bucket. Obtain the current initial resource inventory of the target resource node. According to the time sequence of the time buckets, sequentially add the net resource change amount of each time bucket to the current initial resource inventory to obtain the resource inventory prediction value at the end of each time bucket. Based on all the resource inventory prediction values, obtain the resource inventory prediction value sequence.
9. A data processing apparatus, characterized by, The device includes: The atomic modeling module is used to extract resource transfer event data of multiple heterogeneous types and perform data atomization modeling on the resource transfer event data based on a unified time axis to obtain data transfer flow atomic units. The global merging module is used to obtain the target differential change data corresponding to the dynamic resource transfer event, and merge the baseline data composed of the data transfer stream atomic units with the target differential change data on the unified time axis to obtain global resource transfer view data. The request-response module is used to respond to the target prediction task request, determine the resource scheduling prediction scenario parameters, perform data transfer inference processing on the global resource transfer view data to obtain the prediction scenario resource flow data, aggregate the prediction scenario resource flow data within the target prediction window of the unified time axis to obtain the net resource change amount, and synthesize the net resource change amount to obtain the resource stock prediction value sequence.
10. A computer storage medium, characterized in that, The computer storage medium stores a plurality of instructions adapted for loading by a processor and executing the steps of the method as described in any one of claims 1 to 8.
11. A computer program product, characterised in that, The computer program product stores at least one instruction, which is loaded by a processor and executed according to the steps of the method as described in any one of claims 1 to 8.
12. An electronic device, comprising: include: A processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the steps of the method as described in any one of claims 1 to 8.
