Financial data processing method and related apparatus

Abstract

A financial data processing method, which is applied to the processing of financial data in the financial field. In the method, a computational graph comprising nodes and edges is used to represent a financial data processing flow, and the nodes in the computational graph can indicate input financial data, and an algorithm model function and an operation rule function which are used for executing processing on the financial data. During the execution of the computational graph, by means of sequentially calling the data or functions indicated by the nodes, a series of processing procedures for the financial data can be implemented, such that the data, an algorithm model and an operation rule are visually combined together, thereby improving the interpretability of the financial data processing flow. Moreover, when the financial data processing flow is visually presented on the basis of the computational graph, the financial data processing flow can be changed by means of only adjusting the nodes in the computational graph, thereby facilitating the analysis and adjustment of the financial data processing flow.

Description

A financial data processing method and related device

[0001] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office on December 15, 2023, with application number 202311734349.3 and application name “A Financial Data Processing Method and Related Devices”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of financial computing technology, and in particular to a financial data processing method and related devices. Background Art

[0003] Financial forecasting is based on data generated by the company's business and financial activities over a period of time, external information such as the macro market and competing companies, combined with relevant rules in the company's operating value chain, and using systematic quantitative analysis techniques to predict the company's future financial status and operating level.

[0004] The purpose of financial forecasting is to enhance the proactive nature of financial management, anticipate risks and quantify their potential impact, reduce uncertainty in business management, align the projected goals of financial plans with the changing external environment and economic conditions, and promptly quantify the effectiveness of financial plan implementation. In short, financial forecasting is a crucial basis for business managers to implement lean management and make informed decisions.

[0005] Currently, financial forecasting is performed by financial professionals using domain knowledge (e.g., cross-references between financial indicators) to establish operational relationships between various types of financial data. This creates a financial forecasting model and then executes it to obtain the final forecast results. However, existing financial forecasting models typically run and output the final forecast results in code, making the operational logic between the financial data invisible and hindering analysis and adjustment of the financial forecasting model.

[0006] Summary of the Invention

[0007] This application provides a financial data processing method that can improve the visualization of the financial data processing process and facilitate the analysis and adjustment of the financial data processing process.

[0008] The first aspect of the present application provides a financial data processing method, which is applied to processing financial data in the financial field. The method comprises: first, obtaining a first calculation graph. The first calculation graph is a directed acyclic graph, which is used to indicate the processing flow of financial data. In addition, the first calculation graph includes multiple nodes and multiple directed edges, and the multiple nodes are connected by multiple directed edges, and the multiple directed edges are used to represent the data dependency relationship between the nodes. That is, the directed edges between the nodes are directional and represent the flow direction of the data. In addition, the multiple nodes include a first node, a second node and a third node, the first node is used to indicate the financial input data, the second node is used to indicate the pre-registered algorithm model function, and the third node is used to indicate the rule function constructed based on the operation rules of the financial data.

[0009] Then, based on the data dependency relationship between the nodes in the first computation graph, the execution order of the multiple nodes in the first computation graph can be determined, so that the multiple nodes in the first computation graph are executed in sequence to obtain the output result, which includes the output data corresponding to the multiple nodes.

[0010] Among them, the process of executing the first node includes obtaining financial input data and using the financial input data as the input data of the node connected to the first node, the process of executing the second node includes calling the algorithm model function to process the input data of the second node, and the process of executing the third node includes calling the rule function to perform rule operations on the input data of the third node.

[0011] In this solution, a computational graph consisting of nodes and edges is used to represent the processing flow of financial data. The nodes in the computational graph can indicate the input financial data, the algorithmic model functions that process the financial data, and the operation rule functions. When executing the computational graph, by sequentially calling the data or functions indicated by the nodes, a series of processing processes for the financial data can be implemented, thereby visually combining the data, algorithmic model, and operation rules, and improving the interpretability of the financial data processing flow. By defining the algorithmic model and the operation rules based on expert experience as different functions and integrating them into the same processing flow, the algorithmic model's capabilities in complex operations such as solving optimization problems and time series prediction can be effectively utilized, compensating for the disadvantage of the operation rules based on expert experience that are difficult to handle complex operations, and improving the efficiency and effectiveness of financial data analysis and processing. Moreover, when visually presenting the financial data processing flow based on a computational graph, the financial data processing flow can be modified simply by adjusting the nodes in the computational graph, making it easier to analyze and adjust the financial data processing flow.

[0012] In one possible implementation, the algorithm model function is obtained by registering the target algorithm model as an external function. The second node may specifically indicate a call address for the algorithm model function, so that when the second node is executed, the algorithm model can be called based on the call address indicated by the second node. The target algorithm model includes a statistical learning algorithm model and / or a machine learning algorithm model. For example, the target algorithm model includes an optimization problem solving model or an artificial intelligence model.

[0013] In this solution, by pre-registering the target algorithm model as an external function and indicating the calling method of the external function on the node of the calculation graph, the external target algorithm model can be organically integrated with other financial rule operations in the same calculation process, thereby combining the advantages of the algorithm model and conventional financial rule operations in the same calculation process, and flexibly responding to complex data processing needs in the financial field.

[0014] In one possible implementation, rule functions are derived from pre-built expert experience models, which indicate the multiple operations to be performed sequentially on input data. Simply put, for specific types of financial data, a corresponding expert experience model can be pre-built based on expert experience to indicate the process of performing operations on these specific types of financial data. By defining the expert experience model as a rule function and specifying the defined function at a node, the expert experience model can be invoked to complete the processing of specific types of financial data.

[0015] In this solution, by defining the pre-built expert experience model in the form of a rule function, it is convenient to indicate the entire expert experience model with one node in the calculation graph, ensuring that the expert experience model can be reused when constructing various calculation graphs, and facilitating the organic integration of the expert experience model with other operations in the calculation graph. There is no need to display the internal detailed structure of the expert experience model on the calculation graph, which is conducive to improving the visualization of the calculation graph.

[0016] In one possible implementation, the financial data processing method further includes obtaining a second computation graph, where the second computation graph is obtained by adjusting some nodes in the first computation graph. Optionally, the nodes in the first computation graph that are adjusted include, for example, any one or more of the following: nodes indicating financial input data, nodes indicating algorithmic model functions, or nodes indicating rule functions.

[0017] Then, based on the position of the node performing the adjustment in the second computation graph, a target node in the second computation graph where the output data will change relative to the first computation graph is determined. The target node includes the node performing the adjustment and nodes that can be reached by the node performing the adjustment via directed edges.

[0018] Secondly, based on the data dependency relationship between the nodes in the second computation graph, multiple nodes in the second computation graph are executed in sequence.

[0019] Finally, based on the execution results of the first computation graph and the execution results of the second computation graph, the change in the output data of the target node is displayed. Since the node whose output data is affected is the target node, after the second computation graph is executed, the change in the output data of the target node can be obtained and displayed by comparing the execution results of the second computation graph with the execution results of the first computation graph (i.e., the output data of each node). The change in the output data of the target node can refer to information such as the value before and after the output data change, the percentage change in the output data, and the amount of change in the output data.

[0020] In this solution, by analyzing the nodes affected before and after the calculation graph adjustment and presenting the specific changes in the affected nodes, users can quickly understand the impact of the calculation graph adjustment on the overall financial data processing flow.

[0021] In one possible implementation, based on the data dependency relationship between the nodes in the first computation graph, multiple nodes in the first computation graph are executed sequentially, specifically including: based on the data dependency relationship between the nodes in the first computation graph, arranging a first node queue and a second node queue, the first node queue and the second node queue both including multiple nodes sorted in sequence, and there is no data dependency relationship between the nodes included in the first node queue and the second node queue; then, executing the first node queue and the second node queue in parallel, wherein the execution order of the nodes in the first node queue and the second node queue is determined based on the sorting results of the nodes in the node queues.

[0022] In this solution, multiple node queues that can be executed in parallel are generated based on the data dependencies between nodes in the computation graph, and the node queues include multiple nodes sorted in sequence. This can achieve parallel execution of branches in the computation graph that do not have data dependencies, thereby improving the execution efficiency of the computation graph.

[0023] In one possible implementation, multiple nodes in the first computation graph are sequentially executed based on data dependencies between the nodes in the first computation graph, further comprising: arranging a third node queue based on the data dependencies between the nodes in the first computation graph, the third node queue comprising multiple sequentially ordered nodes. The third node queue has data dependencies with both the first node queue and the second node queue, i.e., the third node queue depends on the outputs of the first node queue and the second node queue. The output data of the first node queue and the output data of the second node queue are then used as input data for the third node queue, and the third node queue is executed.

[0024] In this solution, when generating node queues based on the data dependencies between nodes in the computation graph, in addition to generating node queues that can be executed in parallel, node queues that can be executed serially are also generated, thereby ensuring that the generated node queues can comply with the computational logic of the computation graph, ensuring that the computational logic of the entire computation graph can be executed by executing the node queues, and effectively improving the execution efficiency of the computation graph.

[0025] In one possible implementation, obtaining a first computation graph includes obtaining multiple node creation instructions and multiple node connection instructions, wherein the multiple node creation instructions are each used to instruct the creation of a node in the first computation graph, and the multiple node connection instructions are each used to instruct the connection of already created nodes. Any of the multiple node creation instructions may be used to instruct the creation of a data node or a computation node, such as creating a node indicating financial input data, or creating a node indicating a rule function or an algorithmic model function. Thus, by creating multiple nodes based on the multiple node creation instructions and creating multiple directed edges based on the multiple node connection instructions, the first computation graph can be obtained.

[0026] That is, the execution device creates nodes based on the node creation instruction, and connects the created nodes based on the node connection instruction, and finally obtains a first computational graph including multiple nodes and multiple directed edges.

[0027] In one possible implementation, the first node is specifically used to indicate the type of financial input data (e.g., inventory data, historical price data, historical shipment volume, forecasted price, etc.). There is a mapping relationship between the type of financial input data and the target data structure. That is, the first node is mapped to the target data structure by indicating the type of financial data.

[0028] Thus, when the first node is executed, the data indicated by the target data structure may be used as the financial input data based on the type and mapping relationship of the financial input data indicated by the first node. The target data structure may be, for example, a table, a queue, or an array, for storing the financial input data.

[0029] In one possible implementation, the first computation graph further includes a fourth node, which indicates a financial indicator cross-reference model. Executing the fourth node includes calling the financial indicator cross-reference model to perform a cross-reference operation on the multiple input data of the fourth node. Specifically, the financial indicator cross-reference model is a pre-built model used to perform operations on financial indicators with cross-reference relationships. For example, a simple financial indicator cross-reference model can be a model for calculating net profit. The operational logic of the financial indicator cross-reference model is: net profit = total revenue - total cost.

[0030] The second aspect of the present application provides a financial data processing device, including: an acquisition module, used to obtain a first calculation graph, the first calculation graph includes multiple nodes and multiple directed edges, the multiple nodes are connected by multiple directed edges, the multiple directed edges are used to represent data dependencies between nodes, the multiple nodes include a first node, a second node and a third node, the first node is used to indicate financial input data, the second node is used to indicate a pre-registered algorithm model function, and the third node is used to indicate a rule function constructed based on the operation rules of financial data; a processing module, used to execute multiple nodes in the first calculation graph in sequence based on the data dependencies between the nodes in the first calculation graph to obtain output results, and the output results include output data corresponding to multiple nodes; wherein the process of executing the first node includes acquiring financial input data and using the financial input data as input data of the node connected to the first node, the process of executing the second node includes calling the algorithm model function to process the input data of the second node, and the process of executing the third node includes calling the rule function to perform rule operations on the input data of the third node.

[0031] In one possible implementation, the algorithm model function is obtained by registering a target algorithm model as an external function, where the target algorithm model includes a statistical learning algorithm model and / or a machine learning algorithm model.

[0032] In a possible implementation, the rule function is obtained based on a pre-built expert experience model, where the expert experience model is used to indicate a plurality of operations to be sequentially performed on input data.

[0033] In one possible implementation, the acquisition module is also used to obtain a second computational graph, which is obtained by adjusting some nodes in the first computational graph; the processing module is also used to determine, based on the position of the adjusted node in the second computational graph, a target node in the second computational graph whose output data will change relative to the first computational graph; the processing module is also used to execute multiple nodes in the second computational graph in sequence based on the data dependency relationship between the nodes in the second computational graph; the processing module is also used to display the change in the output data of the target node based on the execution results of the first computational graph and the execution results of the second computational graph.

[0034] In a possible implementation, some of the nodes include any one or more of the following nodes: a node for indicating financial input data, a node for indicating an algorithm model function, or a node for indicating a rule function.

[0035] In one possible implementation, the processing module is further used to: arrange a first node queue and a second node queue based on the data dependency relationship between the nodes in the first computational graph, wherein the first node queue and the second node queue both include a plurality of nodes sorted in sequence, and there is no data dependency relationship between the nodes included in the first node queue and the second node queue; execute the first node queue and the second node queue in parallel, wherein the execution order of the nodes in the first node queue and the second node queue is determined based on the sorting results of the nodes in the node queues.

[0036] In one possible implementation, the processing module is further used to: based on the data dependency relationship between the nodes in the first computation graph, arrange a third node queue, where the third node queue includes a plurality of nodes sorted in sequence; use the output data of the first node queue and the output data of the second node queue as input data of the third node queue, and execute the third node queue.

[0037] In one possible implementation, the acquisition module is further used to obtain multiple node creation instructions and multiple node connection instructions, where the multiple node creation instructions are all used to instruct the creation of nodes in the first computational graph, and the multiple node connection instructions are all used to instruct the connection of already created nodes; the processing module is further used to create multiple nodes based on the multiple node creation instructions, and to create multiple directed edges based on the multiple node connection instructions to obtain the first computational graph.

[0038] In one possible implementation, the first node is specifically used to indicate the type of financial input data, and there is a mapping relationship between the type of financial input data and the target data structure; the acquisition module is further used to call the data indicated by the target data structure as the financial input data based on the type of financial input data indicated by the first node and the mapping relationship.

[0039] In one possible implementation, the first calculation graph also includes a fourth node, which is used to indicate a financial indicator cross-reference model; wherein the process of executing the fourth node includes calling the financial indicator cross-reference model to perform a cross-reference operation on multiple input data of the fourth node.

[0040] A third aspect of the present application provides a financial data processing device, which may include a processor coupled to a memory, the memory storing program instructions. When the program instructions stored in the memory are executed by the processor, the method of the first aspect or any implementation of the first aspect is implemented. For details of the steps in each possible implementation of the first aspect executed by the processor, please refer to the first aspect and will not be repeated here.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the method of any implementation of the first aspect.

[0042] A fifth aspect of the present application provides a circuit system, the circuit system including a processing circuit, and the processing circuit is configured to execute a method of any implementation manner of the above-mentioned first aspect.

[0043] In a sixth aspect, the present application provides a computer program product, which includes program code. When the computer program product is run on a computer, it enables the computer to execute the method of any implementation manner of the first aspect.

[0044] The seventh aspect of the present application provides a chip system, which includes a processor for supporting a server or a feature screening device to implement the functions involved in any implementation of the first aspect, for example, processing the data and / or information involved in the above method. In one possible design, the chip system also includes a memory for storing program instructions and data necessary for the server or feature screening device. The chip system can be composed of a chip, or it can include a chip and other discrete devices.

[0045] The beneficial effects of the second to seventh aspects mentioned above can be referred to the introduction of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] FIG1 is a schematic diagram of a system architecture provided in an embodiment of the present application;

[0047] FIG2 is a schematic structural diagram of an execution device 101 provided in an embodiment of the present application;

[0048] FIG3 is a flow chart of a method for processing financial data according to an embodiment of the present application;

[0049] FIG4 is a schematic diagram of a calculation graph provided in an embodiment of the present application;

[0050] FIG5 is a schematic diagram of a decision tree model provided in an embodiment of the present application;

[0051] FIG6 is a schematic diagram of an execution of a decision tree model provided in an embodiment of the present application;

[0052] FIG7 is a schematic diagram of adjusting a decision tree model provided in an embodiment of the present application;

[0053] FIG8 is a schematic diagram of creating a node for indicating a time series prediction algorithm model provided by an embodiment of the present application;

[0054] FIG9 is a schematic diagram of generating a node queue based on a computation graph according to an embodiment of the present application;

[0055] FIG10 is another flow chart of a method for processing financial data provided by an embodiment of the present application;

[0056] FIG11 is a schematic diagram showing an adjusted calculation graph provided in an embodiment of the present application;

[0057] FIG12 is a schematic diagram of the structure of a financial data processing device provided in an embodiment of the present application;

[0058] FIG13 is a schematic diagram of the structure of an execution device provided in an embodiment of the present application;

[0059] FIG14 is a schematic diagram of the structure of a chip provided in an embodiment of the present application;

[0060] FIG15 is a schematic diagram of the structure of a computer-readable storage medium provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of this application more clear, the embodiments of this application are described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only embodiments of a part of this application, rather than all embodiments. It is known to those skilled in the art that with the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0062] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the descriptions used in this way can be interchangeable where appropriate so that the embodiments can be implemented in a sequence other than that illustrated or described in this application. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The named or numbered process steps can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved. The division of units in this application is a logical division. In actual application, there may be other division methods. For example, multiple units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, and the indirect coupling or communication connection between units can be electrical or other similar forms, which are not limited in this application. Moreover, the units or sub-units described as separate components may or may not be physically separated, may or may not be physical units, or may be distributed into multiple circuit units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this application.

[0063] To facilitate understanding, some technical terms involved in the embodiments of this application are first introduced below.

[0064] (1) Cross-reference relationship

[0065] A cross-checking relationship refers to the necessary, mutually reconcilable relationship between relevant indicators and figures in accounting books and financial statements. For example, the ending balance of each general ledger account and the sum of the ending balances of its subordinate secondary or detailed ledger accounts are mutually consistent and reconcilable. Similarly, the totals of sales revenue, sales tax, sales plant costs, sales expenses, technology transfer fees, and sales profits on a product sales statement are reconcilable with the amounts of the same items on the income statement.

[0066] Generally speaking, the cross-reference relationship between financial indicators can usually be expressed through arithmetic operations.

[0067] (2) Four arithmetic operations

[0068] The four arithmetic operations refer to addition, subtraction, multiplication and division.

[0069] (3) Aggregation operations

[0070] Aggregation operations are operations that calculate a single value from a collection of values. For example, you can calculate an average, a maximum, or a cumulative value from a set of values.

[0071] (4) Directed acyclic graph (DAG)

[0072] In mathematics, especially graph theory and computer science, a directed acyclic graph is a directed graph without loops. Specifically, if a directed graph cannot return to a vertex through a set of edges, then the graph is a directed acyclic graph.

[0073] (5) Statistical Learning

[0074] Statistical learning, also known as statistical machine learning, is a discipline in which computers construct probabilistic statistical models based on data and use these models for prediction and analysis. Data is the subject of statistical learning. The fundamental assumption of statistical learning about data is that similar data exhibit certain statistical regularities. This is the premise of statistical learning. These data share certain common properties, and because of this statistical regularity, they can be processed using statistical learning methods.

[0075] In general, statistical learning methods can be summarized as follows: starting from a given, finite, set of training data used for learning, it is assumed that the data are generated independently and identically distributed; and it is assumed that the model to be learned belongs to a set of functions, called the hypothesis space; applied to a certain evaluation criterion, an optimal model is selected from the hypothesis space so that it has the best prediction for known training data and unknown test data under the given evaluation criterion; the selection of the optimal model is implemented by the algorithm.

[0076] (6) Machine Learning

[0077] Machine learning is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance.

[0078] (7) Application Programming Interface (API)

[0079] An API is a set of predefined functions that allows applications and developers to access a set of routines based on a piece of software or hardware without having to access the source code or understand the details of the internal workings.

[0080] Currently, financial forecasting is performed by financial professionals using domain knowledge (e.g., cross-references between financial indicators) to establish operational relationships between various types of financial data. This creates a financial forecasting model and then executes it to obtain the final forecast results. However, existing financial forecasting models typically run and output the final forecast results in code, making the operational logic between the financial data invisible and hindering analysis and adjustment of the financial forecasting model.

[0081] Furthermore, when building financial forecasting models based on financial domain knowledge, experts often struggle to manually exhaust all possible combinations of hypothetical factors and arrive at the optimal combination strategy, resulting in the constructed financial forecasting models failing to achieve optimal results. Furthermore, the forecasting logic or cross-references for some financial indicators lack clear rules, requiring experts to extract patterns from the data, which is difficult and inefficient.

[0082] The present application provides a financial data processing method, which uses a calculation graph including nodes and edges to represent the processing flow of financial data, and the nodes in the calculation graph can indicate the input financial data, the algorithm model function that performs processing on the financial data, and the operation rule function. When executing the calculation graph, by sequentially calling the data or functions indicated by the nodes, a series of processing processes of the financial data can be realized, thereby visually combining the data, the algorithm model, and the operation rules, and improving the interpretability of the financial data processing flow. By defining the algorithm model and the operation rules based on expert experience as different functions and integrating them into the same processing flow, the algorithm model's ability in complex operations such as solving optimization problems and time series prediction can be effectively utilized, making up for the disadvantage that the operation rules based on expert experience are difficult to handle complex operations, and improving the efficiency and effect of financial data analysis and processing. Moreover, when visually presenting the financial data processing flow based on the calculation graph, the financial data processing flow can be changed by simply adjusting the nodes in the calculation graph, which facilitates the analysis and adjustment of the financial data processing flow.

[0083] Please refer to Figure 1, which is a schematic diagram of a system architecture provided in an embodiment of the present application. As shown in Figure 1, in the system architecture, the execution device 101 can be, for example, a personal computer, a laptop computer, or a server. In addition, the execution device 101 is communicatively connected to the data storage system 102 to obtain data stored in the data storage system 102. The data storage system 102 can be implemented by a storage device deployed on the execution device 101, for example, the execution device 101 is a personal computer, and the data storage system 102 is a hard disk deployed on the personal computer. The data storage system 102 can also be implemented by a storage device independent of the execution device, for example, the execution device 101 is a computing server, and the data storage system 102 is a data server dedicated to storing data.

[0084] During operation, the execution device 101 can obtain a computation graph representing the financial data processing flow. This computation graph is a directed acyclic graph, where nodes represent financial data and the computational methods used to process the financial data (e.g., algorithmic model functions or rule functions used to process the financial data). Based on the connections between the nodes in the computation graph, the execution device 101 can sequentially execute each node and, when executing a node, call the corresponding financial data or function from the data storage system, thereby obtaining the output results (i.e., the processing results of the financial data) after executing the computation graph.

[0085] Please refer to Figure 2, which is a schematic diagram of the structure of an execution device 101 provided in an embodiment of the present application. As shown in Figure 2, the execution device 101 used in the financial data processing method provided in an embodiment of the present application includes a processor 103, which is coupled to a system bus 105. The processor 103 can be one or more processors, each of which can include one or more processor cores. A display adapter (video adapter) 107 can drive a display 109, which is coupled to the system bus 105. The system bus 105 is coupled to an input / output (I / O) bus via a bus bridge 111. An I / O interface 115 is coupled to the I / O bus. The I / O interface 115 communicates with various I / O devices, such as an input device 117 (e.g., a touch screen), an external memory 121 (e.g., a hard drive, floppy disk, optical disk, or USB flash drive), a multimedia interface, etc., a transceiver 123 (capable of sending and / or receiving radio communication signals), a camera 155 (capable of capturing still and dynamic digital video images), and an external USB port 125. Optionally, the interface connected to the I / O interface 115 may be a USB interface.

[0086] The processor 103 may be any conventional processor, including a reduced instruction set computing (RISC) processor, a complex instruction set computing (CISC) processor, or a combination thereof. Alternatively, the processor may be a dedicated device such as an ASIC.

[0087] Execution device 101 can communicate with software deployment server 149 via network interface 129. Exemplarily, network interface 129 is a hardware network interface, such as a network card. Network 127 can be an external network, such as the Internet, or an internal network, such as Ethernet or a virtual private network (VPN). Alternatively, network 127 can be a wireless network, such as a WiFi network or a cellular network.

[0088] The hard drive interface 131 is coupled to the system bus 105. The hard drive interface is connected to the hard drive 133. The internal memory 135 is coupled to the system bus 105. The data running in the internal memory 135 may include an operating system (OS) 137, an application program 143, and a scheduler for executing the device 101.

[0089] The operating system consists of a shell 139 and a kernel 141. Shell 139 is an interface between the user and the operating system's kernel. The shell is the outermost layer of the operating system. The shell manages the interaction between the user and the operating system: it waits for user input, interprets user input to the operating system, and processes various operating system output.

[0090] The kernel 141 consists of the parts of the operating system that manage memory, files, peripherals, and system resources. The kernel 141 directly interacts with the hardware. The operating system kernel typically runs processes and provides inter-process communication, CPU time slice management, interrupts, memory management, and I / O management.

[0091] The above introduces the system architecture and execution equipment used by the method provided in the embodiment of the present application. The following will introduce in detail the execution process of the financial data processing method provided in the embodiment of the present application.

[0092] Please refer to Figure 3, which is a flow chart of a financial data processing method provided by an embodiment of the present application. As shown in Figure 3, the financial data processing method includes the following steps 301-302.

[0093] Step 301: Obtain a first computation graph, where the first computation graph includes multiple nodes and multiple directed edges. The multiple nodes are connected by multiple directed edges, and the multiple directed edges are used to represent data dependencies between the nodes. The multiple nodes include a first node, a second node, and a third node. The first node is used to indicate financial input data, the second node is used to indicate a pre-registered algorithm model function, and the third node is used to indicate a rule function constructed based on operation rules of financial data.

[0094] In this embodiment, the first computation graph is a DAG, which is used to indicate the processing flow of financial data. With respect to the multiple nodes and multiple directed edges included in the first computation graph, each node in the first computation graph is connected to at least one directed edge, and each directed edge is connected to two nodes. In addition, the directed edges between nodes are directional, representing the flow of data, and thus can represent the data dependency relationship between nodes. For example, assuming that node A and node B are connected by a directed edge, and the direction of the directed edge is from node A to node B, then it means that the output data of node A will flow to node B, that is, the output data of node A will serve as the input data of node B, and node B will depend on the data output by node A.

[0095] In the first computation graph, the multiple nodes of the first computation graph actually include two types of nodes: data nodes and computation nodes. Data nodes are nodes used to indicate data, such as nodes indicating input data to the first computation graph or nodes indicating output data from computation nodes. Computation nodes are nodes used to indicate the execution of computations on data. Specifically, computation nodes may include, for example, nodes indicating algorithmic model functions, nodes indicating rule functions, nodes indicating financial cross-checking models, nodes performing decision-making and judgment, and nodes performing optimization solutions.

[0096] Specifically, in this embodiment, the multiple nodes of the first computation graph include a first node, a second node, and a third node. The first node is used to indicate financial input data, i.e., the first node belongs to the aforementioned data node; the second node is used to indicate a pre-registered algorithm model function; and the third node is used to indicate a rule function constructed based on the operation rules of the financial data, i.e., the second and third nodes belong to the aforementioned operation nodes, and the types of operations indicated by the second and third nodes are different.

[0097] Optionally, for the algorithm model function indicated by the second node, the algorithm model function can be obtained by registering the target algorithm model as an external function. Then, in the case where the target algorithm model is pre-registered as an external function, the registered algorithm model function can be obtained, and the second node can specifically indicate the calling address of the algorithm model function, so that when the second node is executed, the target algorithm model execution data can be processed by calling the calling address indicated by the second node through the application programming interface (API).

[0098] Among them, the target algorithm model includes a statistical learning algorithm model and / or a machine learning algorithm model. For example, the target algorithm model includes an optimization problem solving model (referred to as the optimization solution model), which is used to obtain the optimal solution under certain constraints (for example, based on warehouse inventory, the distance between multiple customers and the warehouse, and the goods removal situation of multiple customers and other constraint information to obtain the optimal distribution strategy). For another example, the target algorithm model can be an artificial intelligence (AI) model, and specifically a time series prediction algorithm model, which is used to predict future data based on past historical data (for example, predict future price data based on price data at historical time).

[0099] In this solution, by pre-registering the target algorithm model as an external function and indicating the calling method of the external function on the node of the calculation graph, the external target algorithm model can be organically integrated with other financial rule operations in the same calculation process, thereby combining the advantages of the algorithm model and conventional financial rule operations in the same calculation process, and flexibly responding to complex data processing needs in the financial field.

[0100] Optionally, for the rule function indicated by the third node, the rule function may be obtained based on a pre-built expert experience model, where the expert experience model is used to indicate a plurality of operations to be sequentially performed on the input data of the third node.

[0101] It is understandable that in the financial field, for some specific types of financial data, these financial data can be processed or obtained based on the type of financial data and the field to which it belongs, combined with expert experience. For example, for the distribution strategy of a certain product (such as a mobile phone or a router), a series of judgment processes can be performed based on the inventory of the product and the customer share, so as to determine the distribution quantity for each customer, and then obtain the distribution strategy of the product. Therefore, the distribution strategy of the product actually needs to be obtained by processing the corresponding product inventory data and customer share data in combination with expert experience. Moreover, for the same type of financial data, in most cases, the way of processing such financial data based on expert experience often does not change frequently. Therefore, in this embodiment, for specific types of financial data, corresponding expert experience models can be pre-built based on expert experience to indicate the process of performing calculations on these specific types of financial data.

[0102] Moreover, by defining the pre-built expert experience model as a rule function, it is convenient to indicate the entire expert experience model with one node in the calculation graph, ensuring that the expert experience model can be reused when constructing various calculation graphs, and facilitating the organic integration of the expert experience model with other operations in the calculation graph. There is no need to display the internal detailed structure of the expert experience model on the calculation graph, which is conducive to improving the visualization of the calculation graph.

[0103] Alternatively, the rule function indicated by the third node may be obtained based on any one or more operations selected from the four arithmetic operations, aggregation operations, or conditional judgments. Conditional judgments may be performed on input data, and corresponding actions are performed based on whether the input data satisfies the conditions. For example, when defining a rule function, one or more operations selected from the four arithmetic operations, aggregation operations, or conditional judgments may be combined to obtain a defined rule function.

[0104] In general, the rule function can be a function predefined based on financial field knowledge, satisfying the operation rules of the financial field and capable of performing operations on financial data to obtain corresponding operation results. This embodiment does not limit the specific implementation method of the rule function.

[0105] Optionally, the first computation graph in this embodiment may be pre-installed on the execution device, or obtained by the execution device from a network or other device. The first computation graph may also be constructed by the execution device in response to a user's instruction.

[0106] Exemplarily, in the process of constructing the first computational graph, the execution device may obtain multiple node creation instructions and multiple node connection instructions, wherein the multiple node creation instructions are all used to instruct the creation of nodes in the first computational graph, and the multiple node connection instructions are all used to instruct the connection of the created nodes. Among them, any one of the multiple node creation instructions may be used to instruct the creation of a data node or an operation node, such as creating a node indicating financial input data, or creating a node indicating a rule function or an algorithm model function. This embodiment does not specifically limit this. Specifically, the user can generate a node creation instruction by executing one or more operations (such as specifying the type of node to be created, specifying the operation indicated by the node or the function called, etc.) on the computational graph construction interface displayed by the execution device. In addition, the node connection instruction needs to indicate the connection direction of the node, that is, from which node to which node, so as to ensure that a directed edge can be generated based on the indicated node connection direction.

[0107] Then, the execution device creates multiple nodes based on the multiple node creation instructions and creates multiple directed edges based on the multiple node connection instructions, thereby obtaining a first computation graph. In other words, the execution device creates nodes based on the node creation instructions and connects the created nodes based on the node connection instructions, ultimately obtaining a first computation graph comprising multiple nodes and multiple directed edges.

[0108] This embodiment does not limit the order in which the execution device obtains multiple node creation instructions and multiple node connection instructions. Generally speaking, once the execution device has obtained at least two node creation instructions and has created at least two nodes, the execution device can obtain node connection instructions to connect the created nodes. Generally speaking, the execution device obtains node creation instructions and node connection instructions alternately. That is, after the execution device obtains some node creation instructions and creates the corresponding nodes, it can obtain node connection instructions for these nodes, thereby connecting these nodes.

[0109] Step 302: Based on the data dependency relationship between the nodes in the first computation graph, execute multiple nodes in the first computation graph in sequence to obtain output results, where the output results include output data corresponding to the multiple nodes.

[0110] Since each node is connected to at least one directed edge, and the directed edges between nodes represent the data dependency between nodes (i.e., the data flow between nodes), the data dependency between nodes in the first computation graph can determine the execution order of multiple nodes in the first computation graph. In this way, based on the execution order of multiple nodes in the first computation graph, multiple nodes are executed in sequence to obtain the output result of the first computation graph. Among them, the output result of the first computation graph may include the output data of each of the multiple nodes. In this way, after obtaining the output result of the first computation graph, by visually presenting the output data of each node behind the node of the first computation graph, it is easy for users to clearly understand the calculation results of the financial data after each calculation step. It should be noted that for the data nodes of the first computation graph, the output data of the data node can be the financial data indicated by the data node itself.

[0111] In this embodiment, the process of executing the first node in the first computation graph includes obtaining financial input data indicated by the first node and using the financial input data as input data of the node connected to the first node.

[0112] Exemplarily, the first node can be specifically used to indicate the type of financial input data (such as inventory data, historical price data, historical shipment volume, forecast price, etc.), and there is a mapping relationship between the type of financial input data and the target data structure. That is, the first node is mapped to the target data structure by indicating the type of financial data. In this way, when the first node is executed, the data indicated by the target data structure can be called as the financial input data based on the type of financial input data indicated by the first node and the mapping relationship. Among them, the target data structure can be, for example, a data structure such as a table, a queue or an array, which is used to store the above-mentioned financial input data. For example, historical price data can be stored in the form of a table, and the table used to store historical price data is the target data structure; by indicating that the type of financial input data is historical price data on the first node, it is possible to call the table storing historical price data as the financial input data indicated by the first node based on the mapping relationship between the historical price data type and the target data structure.

[0113] Since the second node indicates a pre-registered algorithm model function, the process of executing the second node includes calling the algorithm model function to process the input data of the second node. For example, the second node can specifically indicate the name of the algorithm model function and the calling address of the algorithm model function. In this way, based on the calling address indicated by the second node, the algorithm model function can be called in the form of an API call, and the input data of the second node is passed to the called algorithm model function, and finally the operation result returned by the algorithm model function is obtained, and the operation result is used as the output data of the second node.

[0114] Similarly, because the third node indicates a rule function, the process of executing the third node includes calling the rule function to perform a rule operation on the input data of the third node. For example, the third node may specifically indicate the name of the rule function and the call address of the rule function. Based on the call address indicated by the third node, the rule function can be called and the input data of the third node can be passed to the rule function, ultimately obtaining the operation result returned by the rule function.

[0115] Optionally, the first calculation graph may further include a fourth node, which is used to indicate a financial indicator cross-reference model. The financial indicator cross-reference model is a pre-built model, which is a model used to calculate financial indicators with cross-reference relationships. For example, a simple financial indicator cross-reference model may be a model for calculating net profit, and the calculation logic of the financial indicator cross-reference model is: net profit = total revenue - total cost. In actual applications, various types of financial indicator cross-reference models can be pre-built based on specific business scenarios, such as a quantity-cost-price model, a pipeline model, a carry-over quantity allocation model, and a country risk model. Financial indicator cross-reference models, this embodiment does not limit the specific implementation method of the financial indicator cross-reference model.

[0116] The process of executing the fourth node includes calling the financial indicator cross-reference model to perform a cross-reference operation on the plurality of input data of the fourth node. For example, when the financial indicator cross-reference model indicated by the fourth node is a model for calculating net profit, and the plurality of input data of the fourth node include total revenue and total cost, then the total profit (i.e., the output data of the fourth node) can be obtained by subtracting the total cost from the total revenue based on the financial indicator cross-reference model.

[0117] For example, please refer to Figure 4, which is a schematic diagram of a calculation graph provided in an embodiment of the present application. As shown in Figure 4, the calculation graph includes multiple nodes and multiple directed edges, and each node is connected to at least one directed edge. In addition, in the calculation graph, the types of nodes are divided into operation nodes and data nodes. Among them, the operation nodes include nodes indicating decision tree models, nodes indicating optimization solution models, nodes indicating time series prediction algorithm models, nodes for solving revenue, nodes for solving costs, and nodes for solving gross rate.

[0118] Specifically, the node indicating the decision tree model is, for example, the third node mentioned above, that is, the decision tree model indicated by the node is an expert experience model constructed based on expert experience, and the decision tree model is defined in the calculation graph in the form of a function. For example, please refer to Figure 5, which is a schematic diagram of a decision tree model provided in an embodiment of the present application. As shown in Figure 5, the decision tree model is actually a model pre-built by a user (such as a financial expert) to indicate how to determine the distribution strategy based on the number of days of supply (DOS) and the proportion of customers.

[0119] In addition, please refer to Figure 6, which is a schematic diagram of the execution of a decision tree model provided in an embodiment of the present application. As shown in Figure 6, the structure displayed by the decision tree model at the front end can be a top-down judgment process according to the user's usage habits, which is more in line with the user's usage habits and is convenient for the user to check or adjust the decision tree model. When executing the node corresponding to the decision tree model, the back end reverses the decision tree model structure through the middle layer conversion to adapt it to the structure of the knowledge representation meta-path. Among them, the knowledge representation meta-path is a minimum unit for expressing complete semantic logic (computational logic) under the knowledge modeling in the financial field. It carries the minimum logic of input + judgment condition + action + output, and is also the minimum unit of analytical reasoning through the graph model.

[0120] Please refer to Figure 7, which is a schematic diagram of the adjustment of a decision tree model provided in an embodiment of the present application. As shown in Figure 7, in the process of constructing a decision tree model, the user can pop up modification options for each node by clicking on each node in the decision tree model on the display interface. For example, when the user clicks on the "goods distribution strategy" node, the display interface can pop up the options of "add judgment node" and "add output node" to facilitate the user to continue to add judgment nodes or output nodes to the decision tree model. For another example, when the user clicks on the judgment node "Product stage = ramp-up period", the display interface can pop up the options of "edit node information", "add judgment node" and "add output node"; and, when the user clicks on "edit node information", the display interface further pops up the operation method that can be edited for the node. For another example, when the user clicks on the output node "no goods distribution", the display interface can pop up "edit node information" to facilitate the user to modify the operation method indicated by the node.

[0121] In addition, in Figure 4, the optimization solution model is an algorithm model used to solve optimization problems in statistical learning, and the time series prediction algorithm model is a model used to predict prices in machine learning. Therefore, the node indicating the optimization solution model and the node indicating the time series prediction algorithm model are, for example, the second node mentioned above, that is, the node used to indicate the algorithm model function. Among them, the optimization solution model and the time series prediction algorithm model are both registered as external functions in advance, so the nodes in the calculation graph can actually be the call addresses used to indicate the optimization solution model and the time series prediction algorithm model after they are registered as functions.

[0122] For example, please refer to Figure 8, which is a schematic diagram of a node for indicating a time series prediction algorithm model provided by an embodiment of the present application. As shown in Figure 8, when creating a node, a node indicating an external function can be created by entering the interface for referencing an external function. Then, by selecting the external function to be referenced (such as the time series prediction algorithm model in Figure 8) on the interface for referencing the external function, and determining the input data of the external function, the node creation can be completed.

[0123] In Figure 4, the solutions for revenue, cost, and gross margin are all based on the financial indicator cross-reference model. Therefore, the nodes for revenue, cost, and gross margin are equivalent to the fourth node, indicating the financial indicator cross-reference model. Each financial indicator cross-reference model predefines the calculation method between financial indicators, for example, revenue = business volume * predicted unit price, and cost = business volume * predicted unit volume.

[0124] The above describes the calculation graph provided by this embodiment in detail with reference to examples. The following describes the execution process of the calculation graph in detail.

[0125] It's understandable that in practical applications, a computation graph may contain a large number of nodes, with different nodes located on different branches of the computation graph, and nodes on different branches have no data dependencies. To improve computation graph execution efficiency, nodes on different branches without data dependencies can actually be executed in parallel, thereby speeding up the execution of the entire computation graph.

[0126] Exemplarily, in the process of sequentially executing multiple nodes in the first calculation graph as described in step 302 above, the following process may be specifically included: First, based on the data dependency relationship between the nodes in the first calculation graph, a first node queue and a second node queue are arranged. The first node queue and the second node queue both include multiple nodes that are sorted in sequence, and there is no data dependency relationship between the nodes included in the first node queue and the second node queue. That is to say, based on the nodes with data dependency on a branch in the first calculation graph, a first node queue can be generated, and the sorting result between the multiple nodes in the first node queue is determined based on the data dependency relationship of the nodes on the branch. Based on the nodes with data dependency on another branch in the first calculation graph, a second node queue can be generated, and the sorting result between the multiple nodes in the second node queue is determined based on the data dependency relationship of the nodes on the branch. In this way, corresponding node queues can be generated for different branches that do not have data dependency relationships in the first calculation graph.

[0127] Then, the first node queue and the second node queue are executed in parallel, where the order in which the nodes in the first node queue and the second node queue are executed is determined based on the ordering of the nodes within the node queues. Because there is no data dependency between the first node queue and the second node queue, the first node queue and the second node queue can be executed in parallel. Furthermore, when executing either of the first node queue and the second node queue, nodes can be sequentially retrieved for execution based on the ordering of the nodes within the node queue, thereby ensuring that the order in which the nodes are executed matches the data dependencies between the nodes.

[0128] It should be noted that the above example uses the generation and parallel execution of a first node queue and a second node queue to illustrate how to execute nodes on different branches of a computation graph in parallel. In practical applications, two or more node queues capable of parallel execution may be generated for a single computation graph, and this is not a specific limitation.

[0129] In this solution, multiple node queues that can be executed in parallel are generated based on the data dependencies between nodes in the computation graph, and the node queues include multiple nodes sorted in sequence. This can achieve parallel execution of branches in the computation graph that do not have data dependencies, thereby improving the execution efficiency of the computation graph.

[0130] Optionally, in some embodiments, in addition to being able to orchestrate the first node queue and the second node queue described above, a third node queue may also be orchestrated based on the data dependency relationship between nodes in the first computation graph. The third node queue includes a plurality of sequentially ordered nodes. The third node queue has a data dependency relationship with both the first node queue and the second node queue, i.e., the third node queue depends on the outputs of the first node queue and the second node queue.

[0131] In this way, after the first node queue and the second node queue are executed, the output data of the first node queue and the output data of the second node queue can be used as input data of the third node queue to execute the third node queue.

[0132] It is understandable that multiple branches in the computational graph that do not have data dependencies may converge to the same node, that is, the output data of multiple branches are all used as input data of the same node. Then this node and other nodes after this node can constitute a convergence branch, and this convergence branch is dependent on the previous multiple branches.

[0133] It should be noted that the above describes executing the first and second node queues in parallel, followed by the third node queue in serial execution. However, in some possible embodiments, a node may be divided into multiple branches after a node. In this case, one node queue is executed first, followed by multiple node queues in parallel. In other words, the execution order between node queues is determined by the specific structure of the computation graph. For branches without data dependencies, node queues can be generated for parallel execution.

[0134] In general, when generating node queues based on the data dependencies between nodes in the computation graph, this solution generates node queues for serial execution in addition to those that can be executed in parallel, thereby ensuring that the generated node queues conform to the computational logic of the computation graph, ensuring that the computational logic of the entire computation graph can be executed by executing the node queues, and effectively improving the execution efficiency of the computation graph.

[0135] For example, please refer to Figure 9, which is a schematic diagram of a node queue generated based on a calculation graph provided in an embodiment of the present application. As shown in Figure 9, the calculation graph includes a total of 9 nodes, wherein node 1 and node are connected to node 3, node 3 is connected to node 4, node 5 is connected to node 6, node 6 is connected to node 7, node 4 and node 7 are connected to node 8, and node 8 is connected to node 9. Based on the connection relationship between the nodes in the calculation graph (i.e., data dependency), a first node queue, a second node queue, and a third node queue can be generated. The first node queue includes nodes 1, node 2, node 3, and node 4 sorted in sequence; the second node queue includes nodes 5, node 6, and node 7 sorted in sequence; and the third node queue includes nodes 8 and node 9 sorted in sequence. In addition, the first node queue and the second node queue are executed in parallel, while the third node queue is executed after the first node queue and the second node queue are executed.

[0136] The above describes the process of building and executing a computation graph to process financial data. In some scenarios, users may need to adjust the calculation methods or financial data in the computation graph and perform inference calculations based on the adjusted computation graph to determine the impact of the adjusted calculation methods or financial data on the financial data processing process, thereby enabling the deduction of various business development scenarios under different business scenarios.

[0137] Based on this, this embodiment also provides a correlation analysis method before and after the calculation graph adjustment, which can analyze the nodes affected before and after the calculation graph adjustment and present the specific changes of the affected nodes, so that users can quickly understand the impact of the calculation graph adjustment on the overall financial data processing process.

[0138] For example, please refer to Figure 10, which is another flowchart of a financial data processing method provided in an embodiment of the present application. As shown in Figure 10, based on the embodiment shown in Figure 3, the following steps 303-306 may also be included.

[0139] Step 303: Obtain a second computation graph, where the second computation graph is obtained by adjusting some nodes in the first computation graph.

[0140] In this embodiment, the second computation graph is obtained by adjusting some of the nodes in the first computation graph based on the first computation graph. Optionally, the nodes to be adjusted include any one or more of the following: nodes indicating financial input data, nodes indicating algorithmic model functions, or nodes indicating rule functions. In other words, the second computation graph can be obtained by adjusting the computational methods such as the financial input data or the algorithmic model functions and rule functions in the first computation graph.

[0141] For example, financial input data may change over time, so the nodes indicating financial input data may need to be adjusted. For example, taking the calculation graph shown in FIG4 as an example, if the grossing rate needs to be determined weekly based on the calculation graph, since the price history and predicted sales order numbers may change each week, the nodes indicating the price history and predicted sales order numbers may need to be adjusted to obtain a new calculation graph.

[0142] For another example, for the same type of financial data, when constructing a calculation graph, the user may want to compare the impact of different calculation methods on the final processing flow, so the nodes indicating the calculation methods such as the algorithm model function and the rule function can be adjusted. For example, taking the calculation graph shown in Figure 4 as an example, when the user registers multiple different time series prediction algorithm models as external functions, the user may want to compare the impact of different time series prediction algorithm models on the entire data processing flow, so the nodes indicating the time series prediction algorithm models can be adjusted to obtain a new calculation graph.

[0143] Step 304 : Based on the position of the node to be adjusted in the second computation graph, determine the target node in the second computation graph whose output data will change relative to the first computation graph.

[0144] Because each node in the second computation graph is connected by a directed edge, the target node in the second computation graph whose output data will change relative to the first computation graph can be determined based on the position of the node performing the adjustment in the second computation graph and the connection relationships between the nodes in the second computation graph. The target node includes the node performing the adjustment and nodes that can be reached by the node performing the adjustment via directed edges.

[0145] Step 305: Based on the data dependency relationship between the nodes in the second computation graph, execute multiple nodes in the second computation graph in sequence.

[0146] The process of executing the second computation graph is similar to the process of executing the first computation graph. For details, please refer to step 302 above, which will not be described in detail here.

[0147] Step 306: Based on the execution results of the first computation graph and the execution results of the second computation graph, the change of the output data of the target node is displayed.

[0148] Since the node whose output data is affected is the target node, after the second computation graph is executed, the change in the output data of the target node can be obtained and displayed by comparing the execution results of the second computation graph with the execution results of the first computation graph (i.e., the output data of each node). The change in the output data of the target node can refer to information such as the value before and after the change in the output data, the percentage change in the output data, the amount of change in the output data, etc., which is not specifically limited here.

[0149] Exemplarily, please refer to Figure 11, which is a schematic diagram showing an adjusted calculation graph provided in an embodiment of the present application. As shown in Figure 11, the calculation graph shown in Figure 11 is obtained after adjusting the node indicating the predicted number of sales orders on the basis of the calculation graph shown in Figure 4. After the calculation graph is adjusted, based on the data dependency relationship between the nodes in the calculation graph, it can be determined that the nodes affected by the output data include nodes indicating the optimization solution model, nodes indicating the shipment volume, nodes indicating the revenue solution method, nodes indicating revenue, nodes indicating the grossing rate, and nodes indicating the grossing rate. That is, the above-mentioned target nodes may, for example, include the adjusted nodes and affected nodes shown in Figure 11.

[0150] In addition, for the node indicating the predicted number of sales orders, the predicted number of sales orders indicated by the node is reduced. In this way, after executing the calculation graph shown in Figure 4 and the calculation graph shown in Figure 11, the changes in the output data of the affected nodes can be displayed in the calculation graph shown in Figure 11 based on the execution results of the two calculation graphs. As shown in Figure 11, the output data (i.e., the shipment volume) of the node indicating the shipment volume in the calculation graph is reduced, the output data (i.e., the revenue) of the node indicating the revenue is reduced, and the output data (i.e., the grossing rate) of the node indicating the grossing rate is reduced. Moreover, for the final output data of the calculation graph (i.e., the grossing rate), the change ratio after the calculation graph is adjusted compared to before the adjustment can be displayed, that is, the grossing rate decreases by 30%, thereby intuitively showing the impact of the adjustment of the predicted number of sales orders on the entire data processing flow.

[0151] The above describes in detail the method provided by the embodiment of the present application. Next, the device provided by the embodiment of the present application for executing the above method will be introduced.

[0152] Please refer to Figure 12, which is a schematic diagram of the structure of a financial data processing device provided by an embodiment of the present application. As shown in Figure 11, the financial data processing device provided by an embodiment of the present application includes: an acquisition module 1201, which is used to acquire a first calculation graph, the first calculation graph including multiple nodes and multiple directed edges, the multiple nodes are connected by multiple directed edges, the multiple directed edges are used to represent data dependencies between the nodes, the multiple nodes include a first node, a second node and a third node, the first node is used to indicate financial input data, the second node is used to indicate a pre-registered algorithm model function, and the third node is used to indicate a rule function constructed based on the operation rules of the financial data; a processing module 1202, which is used to execute multiple nodes in the first calculation graph in sequence based on the data dependencies between the nodes in the first calculation graph to obtain output results, the output results including output data corresponding to the multiple nodes; wherein the process of executing the first node includes acquiring financial input data and using the financial input data as input data of the node connected to the first node, the process of executing the second node includes calling the algorithm model function to process the input data of the second node, and the process of executing the third node includes calling the rule function to perform rule operation on the input data of the third node.

[0153] In one possible implementation, the algorithm model function is obtained by registering a target algorithm model as an external function, where the target algorithm model includes a statistical learning algorithm model and / or a machine learning algorithm model.

[0154] In a possible implementation, the rule function is obtained based on a pre-built expert experience model, where the expert experience model is used to indicate a plurality of operations to be sequentially performed on input data.

[0155] In one possible implementation, the acquisition module 1201 is also used to obtain a second computational graph, which is obtained by adjusting some nodes in the first computational graph; the processing module 1202 is also used to determine, based on the position of the node in the second computational graph on which the adjustment is performed, a target node in the second computational graph whose output data will change relative to the first computational graph; the processing module 1202 is also used to execute multiple nodes in the second computational graph in sequence based on the data dependency relationship between the nodes in the second computational graph; the processing module 1202 is also used to display the change in the output data of the target node based on the execution results of the first computational graph and the execution results of the second computational graph.

[0156] In a possible implementation, some of the nodes include any one or more of the following nodes: a node for indicating financial input data, a node for indicating an algorithm model function, or a node for indicating a rule function.

[0157] In one possible implementation, the processing module 1202 is further used to: arrange a first node queue and a second node queue based on the data dependency relationship between the nodes in the first computational graph, wherein the first node queue and the second node queue both include a plurality of sequentially sorted nodes, and there is no data dependency relationship between the nodes included in the first node queue and the second node queue; execute the first node queue and the second node queue in parallel, wherein the execution order of the nodes in the first node queue and the second node queue is determined based on the sorting results of the nodes in the node queues.

[0158] In one possible implementation, the processing module 1202 is further used to: based on the data dependency relationship between the nodes in the first computation graph, arrange a third node queue, where the third node queue includes a plurality of nodes sorted in sequence; use the output data of the first node queue and the output data of the second node queue as input data of the third node queue, and execute the third node queue.

[0159] In one possible implementation, the acquisition module 1201 is also used to obtain multiple node creation instructions and multiple node connection instructions, where the multiple node creation instructions are all used to indicate the creation of nodes in the first computational graph, and the multiple node connection instructions are all used to indicate the connection of already created nodes; the processing module 1202 is also used to create multiple nodes based on the multiple node creation instructions, and to create multiple directed edges based on the multiple node connection instructions to obtain the first computational graph.

[0160] In one possible implementation, the first node is specifically used to indicate the type of financial input data, and there is a mapping relationship between the type of financial input data and the target data structure; the acquisition module 1201 is further used to call the data indicated by the target data structure as the financial input data based on the type of financial input data indicated by the first node and the mapping relationship.

[0161] In one possible implementation, the first calculation graph also includes a fourth node, which is used to indicate a financial indicator cross-reference model; wherein the process of executing the fourth node includes calling the financial indicator cross-reference model to perform a cross-reference operation on multiple input data of the fourth node.

[0162] Please refer to Figure 13, which is a schematic diagram of the structure of an execution device provided in an embodiment of the present application. The execution device 1300 can be specifically manifested as a server, a personal computer, a laptop computer, etc., which is not limited here. Specifically, the execution device 1300 includes: a receiver 1301, a transmitter 1302, a processor 1303 and a memory 1304 (wherein the number of processors 1303 in the execution device 1300 can be one or more, and Figure 13 uses one processor as an example). In some embodiments of the present application, the receiver 1301, the transmitter 1302, the processor 1303 and the memory 1304 can be connected via a bus or other means.

[0163] Memory 1304 may include read-only memory and random access memory, and provides instructions and data to processor 1303. A portion of memory 1304 may also include non-volatile random access memory (NVRAM). Memory 1304 stores processor and operation instructions, executable modules, or data structures, or subsets or extended sets thereof. The operation instructions may include various operation instructions for implementing various operations.

[0164] Processor 1303 controls the operation of the execution device. In specific applications, the various components of the execution device are coupled together via a bus system. In addition to a data bus, the bus system may also include a power bus, a control bus, and a status signal bus. However, for clarity, all bus systems are referred to as a bus system in the figure.

[0165] The method disclosed in the above embodiment of the present application can be applied to the processor 1303, or implemented by the processor 1303. The processor 1303 can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 1303 or instructions in the form of software. The above-mentioned processor 1303 can be a general-purpose processor, a digital signal processor (digital signal processing, DSP), a microprocessor or a microcontroller, and can further include an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0166] The processor 1303 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. A general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 1304, and the processor 1303 reads the information in the memory 1304 and completes the steps of the above method in combination with its hardware.

[0167] Receiver 1301 can be used to receive input digital or character information and generate signal input related to executing device-related settings and function control. Transmitter 1302 can be used to output digital or character information through the first interface. Transmitter 1302 can also be used to send instructions to the disk group through the first interface to modify data in the disk group. Transmitter 1302 can also include a display device such as a display screen.

[0168] The execution device provided in the embodiments of the present application may specifically be a chip, which includes: a processing unit and a communication unit. The processing unit may be, for example, a processor, and the communication unit may be, for example, an input / output interface, a pin, or a circuit. The processing unit may execute computer-executable instructions stored in the storage unit so that the chip in the execution device executes the method described in the above embodiment. Optionally, the storage unit is a storage unit in the chip, such as a register, a cache, etc. The storage unit may also be a storage unit located outside the chip in the wireless access device, such as a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM), etc.

[0169] Specifically, see Figure 14 , which is a schematic diagram of the structure of a chip provided in an embodiment of the present application. The chip can be a neural network processor (NPU) 1400. NPU 1400 is mounted on the host CPU (host CPU) as a coprocessor, and the host CPU assigns tasks. The core of the NPU is arithmetic circuit 1403, which is controlled by controller 1404 to extract matrix data from memory and perform multiplication operations.

[0170] In some implementations, the arithmetic circuit 1403 includes multiple processing units (PEs). In some implementations, the arithmetic circuit 1403 is a two-dimensional systolic array. The arithmetic circuit 1403 can also be a one-dimensional systolic array or other electronic circuitry capable of performing mathematical operations such as multiplication and addition. In some implementations, the arithmetic circuit 1403 is a general-purpose matrix processor.

[0171] For example, assume there are input matrix A, weight matrix B, and output matrix C. The arithmetic circuit retrieves the corresponding data of matrix B from weight memory 1402 and caches it on each PE in the arithmetic circuit. The arithmetic circuit retrieves the data of matrix A from input memory 1401 and performs a matrix operation on matrix B. The partial or final matrix result is stored in accumulator 1408.

[0172] Unified memory 1406 is used to store input and output data. Weight data is directly transferred to weight memory 1402 through the Direct Memory Access Controller (DMAC) 1405. Input data is also transferred to unified memory 1406 through the DMAC.

[0173] BIU stands for Bus Interface Unit, i.e., bus interface unit 1410 , which is used for interaction between the AXI bus, DMAC, and instruction fetch buffer (IFB) 1409 .

[0174] The bus interface unit 1410 (BIU) is used for the instruction fetch memory 1409 to obtain instructions from the external memory, and is also used for the storage unit access controller 1405 to obtain the original data of the input matrix A or the weight matrix B from the external memory.

[0175] DMAC is mainly used to transfer input data in the external memory DDR to the unified memory 1406 or transfer weight data to the weight memory 1402 or transfer input data to the input memory 1401.

[0176] The vector calculation unit 1407 includes multiple operation processing units. When necessary, it further processes the output of the operation circuit 1403, such as vector multiplication, vector addition, exponential operation, logarithmic operation, size comparison, etc. It is mainly used for non-convolutional / fully connected layer network calculations in neural networks, such as batch normalization, pixel-level summation, and upsampling of feature planes.

[0177] In some implementations, the vector calculation unit 1407 can store the processed output vector to the unified memory 1406. For example, the vector calculation unit 1407 can apply a linear function or a nonlinear function to the output of the operation circuit 1403, such as linear interpolation of the feature plane extracted by the convolution layer, or accumulate a vector of values ​​to generate an activation value. In some implementations, the vector calculation unit 1407 generates a normalized value, a pixel-level summed value, or both. In some implementations, the processed output vector can be used as an activation input to the operation circuit 1403, for example, for use in subsequent layers in a neural network.

[0178] An instruction fetch buffer 1409 connected to the controller 1404 is used to store instructions used by the controller 1404;

[0179] Unified memory 1406, input memory 1401, weight memory 1402, and instruction fetch memory 1409 are all on-chip memories. External memories are private to the NPU hardware architecture.

[0180] The processor mentioned in any of the above places can be a general-purpose central processing unit, a microprocessor, an ASIC, or one or more integrated circuits for controlling the execution of the above program.

[0181] Please refer to Figure 15, which is a schematic diagram of the structure of a computer-readable storage medium provided in an embodiment of the present application. This application also provides a computer-readable storage medium. In some embodiments, the methods disclosed in the above embodiments can be implemented as computer program instructions encoded in a machine-readable format on a computer-readable storage medium or on other non-transitory media or products.

[0182] 15 schematically illustrates a conceptual partial view of an example computer-readable storage medium including a computer program for executing a computer process on a computing device, arranged in accordance with at least some embodiments presented herein.

[0183] In one embodiment, the computer readable storage medium 1500 is provided using a signal bearing medium 1501. The signal bearing medium 1501 may include one or more program instructions 1502, which when executed by one or more processors may provide the functions or part of the functions described in the above embodiments.

[0184] In some examples, the signal bearing medium 1501 may include a computer readable medium 1503 such as, but not limited to, a hard drive, a compact disk (CD), a digital video disk (DVD), a digital tape, a memory, a ROM or RAM, and the like.

[0185] In some embodiments, the signal-bearing medium 1501 may include a computer-recordable medium 1504, such as, but not limited to, a memory, a read / write (R / W) CD, a R / W DVD, or the like. In some embodiments, the signal-bearing medium 1501 may include a communication medium 1505, such as, but not limited to, a digital and / or analog communication medium (e.g., a fiber optic cable, a waveguide, a wired communication link, a wireless communication link, or the like). Thus, for example, the signal-bearing medium 1501 may be communicated via a wireless form of the communication medium 1505 (e.g., a wireless communication medium conforming to the IEEE 802.X standard or other transmission protocol).

[0186] The one or more program instructions 1502 may be, for example, computer-executable instructions or logic-implemented instructions. In some examples, the computing device may be configured to provide various operations, functions, or actions in response to the program instructions 1502 communicated to the computing device via one or more of computer-readable media 1503, computer-recordable media 1504, and / or communication media 1505.

[0187] It should also be noted that the device embodiments described above are merely illustrative, in which the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines.

[0188] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memories, special components, etc. In general, all functions performed by computer programs can be easily implemented with corresponding hardware, and the specific hardware structures used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, training equipment, or network equipment, etc.) to execute the methods of each embodiment of the present application.

[0189] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0190] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, computer, training equipment or data center to another website, computer, training equipment or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training equipment, data center, etc. that includes one or more available media integrations. Available media can be magnetic media, (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state drive (SSD)), etc.

Claims

1. A method for processing financial data, characterized in that: include: Acquire a first computation graph, the first computation graph comprising a plurality of nodes and a plurality of directed edges, the plurality of nodes being connected by the plurality of directed edges, the plurality of directed edges being used to represent data dependency relationships between the nodes, the plurality of nodes comprising a first node, a second node, and a third node, the first node being used to indicate financial input data, the second node being used to indicate a pre-registered algorithm model function, and the third node being used to indicate a rule function constructed based on a calculation rule of financial data; Based on the data dependency relationship between the nodes in the first computation graph, execute multiple nodes in the first computation graph in sequence to obtain an output result, where the output result includes output data corresponding to the multiple nodes; Among them, the process of executing the first node includes obtaining the financial input data and using the financial input data as the input data of the node connected to the first node, the process of executing the second node includes calling the algorithm model function to process the input data of the second node, and the process of executing the third node includes calling the rule function to perform rule operations on the input data of the third node.

2. The method according to claim 1, characterized in that: The algorithm model function is obtained by registering a target algorithm model as an external function, and the target algorithm model includes a statistical learning algorithm model and / or a machine learning algorithm model.

3. The method according to claim 1 or 2, characterized in that: The rule function is obtained based on a pre-built expert experience model, and the expert experience model is used to indicate a plurality of operations to be sequentially performed on input data.

4. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: Obtaining a second computation graph, where the second computation graph is obtained by adjusting some nodes in the first computation graph; Determine, based on the position of the node in the second computation graph that performs the adjustment, a target node in the second computation graph where output data will change relative to the first computation graph; Based on the data dependency relationship between the nodes in the second computation graph, sequentially execute multiple nodes in the second computation graph; Based on the execution results of the first computation graph and the execution results of the second computation graph, the changes in the output data of the target node are displayed.

5. The method according to claim 4, characterized in that The partial nodes include any one or more of the following nodes: a node for indicating financial input data, a node for indicating an algorithm model function, or a node for indicating a rule function.

6. The method according to any one of claims 1 to 5, characterized in that: The executing the plurality of nodes in the first computation graph in sequence based on the data dependency relationship between the nodes in the first computation graph comprises: Based on the data dependency relationship between the nodes in the first computing graph, a first node queue and a second node queue are arranged, wherein the first node queue and the second node queue both include a plurality of nodes that are sequentially arranged, and there is no data dependency relationship between the nodes included in the first node queue and the second node queue; The first node queue and the second node queue are executed in parallel, wherein the execution order of the nodes in the first node queue and the second node queue is determined based on the sorting result of the nodes in the node queues.

7. The method according to claim 6, characterized in that The executing the plurality of nodes in the first computation graph in sequence based on the data dependency relationship between the nodes in the first computation graph further includes: Based on the data dependency relationship between the nodes in the first computing graph, a third node queue is arranged, where the third node queue includes a plurality of nodes arranged in sequence; The output data of the first node queue and the output data of the second node queue are used as input data of the third node queue, and the third node queue is executed.

8. The method according to any one of claims 1 to 7, characterized in that: The obtaining of the first computation graph includes: Obtaining a plurality of node creation instructions and a plurality of node connection instructions, wherein the plurality of node creation instructions are all used to instruct the creation of nodes in the first computation graph, and the plurality of node connection instructions are all used to instruct the connection of already created nodes; The multiple nodes are created based on the multiple node creation instructions, and the multiple directed edges are created based on the multiple node connection instructions to obtain the first computational graph.

9. The method according to any one of claims 1 to 8, characterized in that: The first node is specifically used to indicate the type of the financial input data, and there is a mapping relationship between the type of the financial input data and the target data structure; The obtaining of the financial input data comprises: Based on the type of the financial input data indicated by the first node and the mapping relationship, the data indicated by the target data structure is called as the financial input data.

10. The method according to any one of claims 1 to 9, characterized in that: The first calculation graph further includes a fourth node, and the fourth node is used to indicate a financial indicator cross-reference model; The process of executing the fourth node includes calling the financial indicator cross-reference model to perform a cross-reference operation on multiple input data of the fourth node.

11. A financial data processing device, characterized in that: include: An acquisition module is used to acquire a first computation graph, wherein the first computation graph includes a plurality of nodes and a plurality of directed edges, wherein the plurality of nodes are connected by the plurality of directed edges, wherein the plurality of directed edges are used to represent data dependency relationships between the nodes, wherein the plurality of nodes include a first node, a second node, and a third node, wherein the first node is used to indicate financial input data, the second node is used to indicate a pre-registered algorithm model function, and the third node is used to indicate a rule function constructed based on a calculation rule of financial data; A processing module, configured to sequentially execute multiple nodes in the first computation graph based on the data dependency relationship between the nodes in the first computation graph to obtain an output result, wherein the output result includes output data corresponding to the multiple nodes; Among them, the process of executing the first node includes obtaining the financial input data and using the financial input data as the input data of the node connected to the first node, the process of executing the second node includes calling the algorithm model function to process the input data of the second node, and the process of executing the third node includes calling the rule function to perform rule operations on the input data of the third node.

12. The device according to claim 11, characterized in that The algorithm model function is obtained by registering a target algorithm model as an external function, and the target algorithm model includes a statistical learning algorithm model and / or a machine learning algorithm model.

13. The device according to claim 11 or 12, characterized in that The rule function is obtained based on a pre-built expert experience model, and the expert experience model is used to indicate a plurality of operations to be sequentially performed on input data.

14. The device according to any one of claims 11 to 13, characterized in that: The acquisition module is further used to acquire a second computation graph, where the second computation graph is obtained by adjusting some nodes in the first computation graph; The processing module is further used to determine, based on the position of the node in the second computation graph that performs the adjustment, a target node in the second computation graph where output data will change relative to the first computation graph; The processing module is further used to sequentially execute multiple nodes in the second computation graph based on the data dependency relationship between the nodes in the second computation graph; The processing module is also used to display the change of the output data of the target node based on the execution result of the first calculation graph and the execution result of the second calculation graph.

15. The device according to claim 14, characterized in that The partial nodes include any one or more of the following nodes: a node for indicating financial input data, a node for indicating an algorithm model function, or a node for indicating a rule function.

16. The device according to any one of claims 11 to 15, characterized in that: The processing module is further used for: Based on the data dependency relationship between the nodes in the first computation graph, a first node queue and a second node queue are arranged, wherein the first node queue and the second node queue both include a plurality of nodes arranged in sequence, and There are no data dependencies between the included nodes; The first node queue and the second node queue are executed in parallel, wherein the execution order of the nodes in the first node queue and the second node queue is determined based on the sorting result of the nodes in the node queues.

17. The device according to claim 16, characterized in that The processing module is further used for: Based on the data dependency relationship between the nodes in the first computing graph, a third node queue is arranged, where the third node queue includes a plurality of nodes arranged in sequence; The output data of the first node queue and the output data of the second node queue are used as input data of the third node queue, and the third node queue is executed.

18. The device according to any one of claims 11 to 17, characterized in that: The acquisition module is further used to acquire multiple node creation instructions and multiple node connection instructions, wherein the multiple node creation instructions are all used to instruct the creation of nodes in the first calculation graph, and the multiple node connection instructions are all used to instruct the connection of the created nodes; The processing module is further used to create the multiple nodes based on the multiple node creation instructions, and to create the multiple directed edges based on the multiple node connection instructions, so as to obtain the first computational graph.

19. The device according to any one of claims 11 to 18, characterized in that: The first node is specifically used to indicate the type of the financial input data, and there is a mapping relationship between the type of the financial input data and the target data structure; The acquisition module is further used to call the data indicated by the target data structure as the financial input data based on the type of the financial input data indicated by the first node and the mapping relationship.

20. The device according to any one of claims 11 to 19, characterized in that: The first calculation graph further includes a fourth node, and the fourth node is used to indicate a financial indicator cross-reference model; The process of executing the fourth node includes calling the financial indicator cross-reference model to perform a cross-reference operation on multiple input data of the fourth node.

21. A financial data processing device, characterized in that: The device comprises a memory and a processor; the memory stores codes, the processor is configured to execute the codes, and when the codes are executed, the device executes the method according to any one of claims 1 to 10.

22. A computer storage medium, characterized in that The computer storage medium stores instructions, which, when executed by a computer, cause the computer to implement the method of any one of claims 1 to 10.

Images