A model selection method and system for complex scene tasks

Abstract

The present invention relates to the field of data processing technology, and specifically discloses a model selection method and system for complex scene tasks. The present invention first obtains the task characteristic attributes for complex scenes, obtains the task delay time according to the task characteristic attributes, then determines the structured data type and the unstructured data type for the task characteristic attributes, and obtains the first result response time of the structured data type and the second result response time of the unstructured data type, and then calculates the result response time ratio of the two, and judges whether the result response time ratio is greater than a preset threshold. If it is greater, it is determined that the task characteristic attribute is a task with a high calculation ratio, and arranges the calculation power model for processing, otherwise, arranges the image power model for processing, so that the model selection for complex scene tasks can be optimized, thereby selecting the corresponding power model through the attributes of complex scene tasks, and then avoiding the problem of slow model processing data.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a model selection method and system for complex scenario tasks. Background Art

[0002] With the in-depth development of the field of artificial intelligence, there are often many different models to choose from when facing complex scenario tasks. Among them, the model refers to the big data model, and the big data model usually refers to the mathematical or computational model used to process and analyze large-scale data sets. With the development of information technology, the amount of data has exploded, and traditional data processing methods can no longer meet the needs. Therefore, big data technology came into being. It includes a series of technologies and tools to effectively store, process and analyze large amounts of data. Different models correspond to different computing power, as well as different performance indicators, data requirements and computing resource consumption.

[0003] When processing tasks, a computing power model with higher similarity is matched to the task (usually tasks for complex scenarios are composed of image data processing and computing data processing). However, computing power models with higher similarity can only perform computing power calculations step by step, but cannot specifically select computing power models (image computing power models and computing computing power models) with corresponding attributes (for example, tasks with a large amount of image data but less computing data, or vice versa, tasks with less image data and more computing data). This will cause the model to process data slowly. Therefore, a model selection method and system for complex scenario tasks are needed to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to provide a model selection method and system for complex scene tasks to solve the technical problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A model selection method for complex scene tasks, including:

[0007] Acquire task characteristic attributes for complex scenarios, where the task characteristic attributes include task data structure characteristic information, task data volume, and task time limit characteristic information;

[0008] Acquire the task delay time according to the task time limit characteristic information;

[0009] Splitting the task data structure characteristic information according to preset time nodes to obtain a plurality of node task structures, and determining the task data structure type according to the plurality of node task structures, wherein the task data structure type includes a structured data type and an unstructured data type;

[0010] Acquire first computing power characteristic information of the structured data type, acquire a first computing power value and a first image computing power value according to the first computing power characteristic information, and calculate a first result response time according to the first computing power value, the first image computing power value, the task data volume, and the task delay time;

[0011] Acquire second computing power characteristic information of the unstructured data type, acquire a second computing power value and a second image computing power value according to the second computing power characteristic information, and calculate a second result response time according to the second computing power value, the second image computing power value, the task data volume, and the task delay time;

[0012] Calculate a result response time ratio value according to the first result response time and the second result response time;

[0013] Determine whether the result response time ratio is greater than a preset threshold;

[0014] If the result response time ratio is greater than a preset threshold, the task characteristic attribute is determined to be a task with a high calculation ratio, and a calculation power model for arranging tasks for complex scenarios is processed;

[0015] If the result response time ratio is not greater than a preset threshold, the task characteristic attribute is determined to be a task with a high image ratio, and an image computing model is arranged to process the task for complex scenes.

[0016] Preferably, the step of obtaining task feature attributes for complex scenarios includes:

[0017] Acquire task data for complex scenarios, acquire a corresponding task data format for complex scenarios according to the task data, and analyze the task data format based on Python to obtain task type information and data format information;

[0018] Identifying the data format information based on a preset image data format to obtain a plurality of image data formats;

[0019] Identify the data format information based on a preset computing power structure data format to obtain a plurality of computing power structure data formats, wherein the preset computing power structure data format is a preset fixed format;

[0020] The feature information set corresponding to the multiple image data formats and the multiple computing power structure data formats is used as the task data structure feature information.

[0021] Preferably, the step of obtaining the task delay time according to the task time limit feature information comprises:

[0022] Get the preset processing time corresponding to the task data volume;

[0023] Classifying the task data in a format based on a preset time node to obtain a plurality of classified task formats;

[0024] Acquire multiple first processing times corresponding to multiple classification task formats, and accumulate the multiple first processing times to obtain a first processing total time;

[0025] The difference between the preset processing time and the first processing total time is calculated to obtain the difference time, and the difference time is used as the task delay time.

[0026] Preferably, the step of determining the type of the task data structure according to the plurality of node task structures comprises:

[0027] Acquire structure information of each task data structure type according to the plurality of node task structures, wherein the structure information includes at least one of text structure information, image structure information and video structure information;

[0028] Determining in sequence whether the plurality of structure information are consistent with a preset unstructured data type;

[0029] If the structure information is consistent with the preset unstructured data type, taking the task data structure type corresponding to the plurality of structure information as the unstructured data type, and obtaining a first quantity of the plurality of structure information corresponding to the unstructured data type;

[0030] If the structure information is inconsistent with the preset unstructured data type, taking the task data structure type corresponding to the plurality of structure information as the structure data type, and obtaining a second quantity of the plurality of structure information corresponding to the structure data type;

[0031] Acquire a quantity ratio value according to the first quantity and the second quantity, and determine whether the quantity ratio value is greater than a preset quantity threshold;

[0032] If the quantity ratio is greater than a preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as an unstructured data type;

[0033] If the quantity ratio is not greater than a preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as a structured data type.

[0034] Preferably, the step of calculating the first result response time according to the first computing power value, the first image computing power value, the task data volume and the task delay time includes:

[0035] The first result response time is calculated according to the first computing power value, the first image computing power value, the task data volume and the task delay time, wherein the calculation formula is:

[0036]

[0037] in, Indicates the first result response time, Indicates the first computing power value, Indicates the computing power value of the first image, Indicates the amount of task data, Indicates the task delay time.

[0038] Preferably, the step of calculating the second result response time according to the second computing power value, the second image computing power value, the task data volume and the task delay time includes:

[0039] The second result response time is calculated according to the second computing power value, the second image computing power value, the task data volume and the task delay time, wherein the calculation formula is:

[0040]

[0041] in, Indicates the second result response time, Indicates the second computing power value, Indicates the computing power value of the second image, Indicates the amount of task data, Indicates the task delay time.

[0042] This application also provides a model selection system for complex scene tasks, including:

[0043] The first acquisition module is used to acquire task characteristic attributes for complex scenarios, wherein the task characteristic attributes include task data structure characteristic information, task data volume and task time limit characteristic information;

[0044] A second acquisition module is used to acquire the task delay time according to the task time limit feature information;

[0045] A first splitting module is used to split the task data structure characteristic information according to preset time nodes to obtain multiple node task structures, and determine the task data structure type according to the multiple node task structures, wherein the task data structure type includes structured data type and unstructured data type;

[0046] a third acquisition module, configured to acquire first computing power characteristic information of the structured data type, acquire a first computing power value and a first image computing power value according to the first computing power characteristic information, and calculate a first result response time according to the first computing power value, the first image computing power value, the task data volume, and the task delay time;

[0047] a fourth acquisition module, configured to acquire second computing power characteristic information of an unstructured data type, acquire a second computing power value and a second image computing power value according to the second computing power characteristic information, and calculate a second result response time according to the second computing power value, the second image computing power value, the task data volume, and the task delay time;

[0048] A first calculation module, used to calculate a result response time ratio value according to the first result response time and the second result response time;

[0049] A first judgment module is used to judge whether the result response time ratio is greater than a preset threshold;

[0050] If the result response time ratio is greater than a preset threshold, the task characteristic attribute is determined to be a task with a high calculation ratio, and a calculation power model for arranging tasks for complex scenarios is processed;

[0051] If the result response time ratio is not greater than a preset threshold, the task characteristic attribute is determined to be a task with a high image ratio, and an image computing model is arranged to process the task for complex scenes.

[0052] Preferably, the first acquisition module includes:

[0053] A first acquisition unit is used to acquire task data for complex scenarios, acquire a corresponding task data format for complex scenarios according to the task data, and analyze the task data format based on Python to obtain task type information and data format information;

[0054] A first recognition unit, configured to recognize data format information based on a preset image data format to obtain a plurality of image data formats;

[0055] A second identification unit is used to identify the data format information based on a preset computing power structure data format to obtain a plurality of computing power structure data formats, wherein the preset computing power structure data format is a preset fixed format;

[0056] The feature information set corresponding to the multiple image data formats and the multiple computing power structure data formats is used as the task data structure feature information.

[0057] The present application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0058] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0059] The beneficial effects of the present application are as follows: the present invention first obtains task data structure characteristic information, task data volume and task time limit characteristic information for complex scenarios, then obtains the task delay time according to the task time limit characteristic information, then splits the task data structure characteristic information according to preset time nodes to obtain multiple node task structures, determines the task data structure type according to the multiple node task structures, wherein the task data structure type includes structured data types and unstructured data types, then obtains the first computing power characteristic information of the structured data type, and obtains the corresponding first result response time according to the first computing power characteristic information, then obtains the second computing power characteristic information of the unstructured data type, and obtains the corresponding first result response time according to the second computing power characteristic information. Take the corresponding second result response time, calculate the ratio of the first result response time and the second result response time, and determine whether the result response time ratio is greater than the preset threshold; if the result response time ratio is greater than the preset threshold, then the task characteristic attribute is determined to be a task with a high calculation ratio, and the computing power model is arranged to process the task for complex scenes; if the result response time ratio is not greater than the preset threshold, then the task characteristic attribute is determined to be a task with a high image ratio, and the image computing power model is arranged to process the task for complex scenes. In this way, the model selection for the task for complex scenes can be optimized, thereby selecting the corresponding computing power model according to the attributes of the task for complex scenes, thereby avoiding the problem of slow data processing by the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The figure is a schematic diagram of a method flow of an embodiment of the present application.

[0061] Figure 2 A schematic diagram of the system structure of an embodiment of the present application.

[0062] Figure 3 A schematic diagram of the internal structure of a computer device according to an embodiment of the present application.

[0063] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0064] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0065] like Figure 1-Figure 3 As shown, the present application provides a model selection method for complex scene tasks, including:

[0066] S1. Acquire task characteristic attributes for complex scenarios, wherein the task characteristic attributes include task data structure characteristic information, task data volume, and task time limit characteristic information;

[0067] S2. Obtaining the task delay time according to the task time limit characteristic information;

[0068] S3, splitting the task data structure characteristic information according to preset time nodes to obtain multiple node task structures, and determining the task data structure type according to the multiple node task structures, wherein the task data structure type includes structured data type and unstructured data type;

[0069] S4. Obtain first computing power characteristic information of the structured data type, obtain a first computing power value and a first image computing power value according to the first computing power characteristic information, and calculate a first result response time according to the first computing power value, the first image computing power value, the task data volume, and the task delay time;

[0070] S5. Obtain second computing power characteristic information of the unstructured data type, obtain a second computing power value and a second image computing power value according to the second computing power characteristic information, and calculate a second result response time according to the second computing power value, the second image computing power value, the task data volume, and the task delay time;

[0071] S6. Calculate a result response time ratio according to the first result response time and the second result response time;

[0072] S7, determining whether the result response time ratio is greater than a preset threshold;

[0073] If the result response time ratio is greater than a preset threshold, the task characteristic attribute is determined to be a task with a high calculation ratio, and a calculation power model for arranging tasks for complex scenarios is processed;

[0074] If the result response time ratio is not greater than a preset threshold, the task characteristic attribute is determined to be a task with a high image ratio, and an image computing model is arranged to process the task for complex scenes.

[0075] As described in the above steps S1-S7, during task processing, a computing power model with a higher similarity is matched to the task (usually a task for complex scenes is composed of image data processing and computing data processing), and the computing power model with a higher similarity can only perform computing power calculations step by step, and cannot specifically select computing power models (image computing power models and computing computing power models) with corresponding attributes (for example: tasks with a large amount of image data but less computing data, or vice versa, tasks with less image data and more computing data), which will cause the model to process data slowly. Therefore, the present invention first obtains the task feature attributes for complex scenes, wherein the task feature attributes include task data structure feature information, task data volume and task time limit feature information, so that the task feature attributes can be used to A comprehensive understanding of the characteristics of the calculated tasks can also provide accuracy and reliability for the selection of models. Since the purpose of task calculation is to calculate quickly, but the processing time of different tasks is different, the waiting delay time is also different. On this basis, it is necessary to obtain the task delay time according to the task time limit feature information, so that the computing resources can be reasonably allocated through the task delay time to ensure that the task is completed on time. Since the proportion of the task data structure type determines the selection of the task model, it is necessary to split the task data structure feature information according to the preset time nodes to obtain multiple node task structures, and determine the task data structure type according to the multiple node task structures, where the task data structure type includes structured data types and non-structured data types. The first computing power characteristic information of the structured data type is then obtained, and a first computing power value and a first image computing power value are obtained according to the first computing power characteristic information, and a first result response time is calculated according to the first computing power value, the first image computing power value, the task data volume, and the task delay time, so as to provide an accurate prediction of the result response time, help optimize task scheduling, and provide data support for the processing time ratio of the structured data type and the unstructured data type, and then the second computing power characteristic information of the unstructured data type is obtained, and a second computing power value and a second image computing power value are obtained according to the second computing power characteristic information, and a first computing power value and a second image computing power value are calculated according to the second computing power characteristic information. The second result response time is calculated based on the first image computing power value, the second image computing power value, the task data volume and the task delay time. In this way, the second result response time can provide data support for the processing time ratio of structured data types and unstructured data types. Then, the result response time ratio value is calculated according to the first result response time and the second result response time. In this way, the result response time ratio value can be used to analyze the ratio type of the task structure and select the optimal computing power model for complex scene tasks. On this basis, it is determined whether the result response time ratio value is greater than a preset threshold value. If the result response time ratio value is greater than the preset threshold value, it is determined that the task characteristic attribute is a task with a high computing ratio, and the computing power model is arranged for processing the complex scene tasks.If the result response time ratio is not greater than the preset threshold, the task characteristic attribute is determined to be a task with a high image ratio, and the image computing model is arranged to process the complex scene-oriented task. In this way, by comparing the result response time ratio with the preset threshold, the ratio of a large amount of image data and computing data in the complex scene-oriented task can be reasonably judged, and then the model selection for the complex scene-oriented task can be optimized, so that the corresponding computing model can be selected according to the attributes of the complex scene-oriented task, thereby avoiding the problem of slow data processing by the model.

[0076] In one embodiment, the step S1 of acquiring task feature attributes for complex scenarios includes:

[0077] S101, obtaining task data for complex scenarios, obtaining a corresponding task data format for complex scenarios according to the task data, and analyzing the task data format based on Python to obtain task type information and data format information;

[0078] S102, identifying data format information based on a preset image data format to obtain multiple image data formats;

[0079] S103, identifying the data format information based on a preset computing power structure data format to obtain a plurality of computing power structure data formats, wherein the preset computing power structure data format is a preset fixed format;

[0080] S104: Use a set of characteristic information corresponding to the plurality of image data formats and the plurality of computing power structure data formats as task data structure characteristic information.

[0081] As described in the above steps S101-S104, the present invention first obtains task data for complex scenarios, obtains the corresponding task data format for complex scenarios according to the task data, and analyzes the task data format based on Python to obtain task type information and data format information, which can provide necessary basic information for the proportion of data structure types in subsequent steps, ensuring the accuracy and completeness of task feature attribute acquisition. Since the tasks to be processed are composed of structured data types and unstructured data types, the structured data types are data that have been organized into a fixed format, usually stored in a relational database, and can be represented in a table. Each row represents a record, and each column represents an attribute or field, such as function formula calculations. Unstructured data types are data without a fixed format or structure, usually including multimedia data such as text, images, audio, and video. These data cannot be directly stored in traditional relational databases and need to be identified, judged, and processed separately through models. On this basis, the data format information is first identified based on the preset image data format to obtain multiple image data formats, and then the data format information is identified based on the preset computing power structure data format to obtain multiple computing power structure data formats, wherein the preset computing power structure data format is a pre-set fixed format. Finally, the feature information set corresponding to the multiple image data formats and the multiple computing power structure data formats is used as the task data structure feature information, which can provide a comprehensive and accurate basis for obtaining task feature attributes.

[0082] In one embodiment, the step S2 of acquiring the task delay time according to the task time limit characteristic information includes:

[0083] Get the preset processing time corresponding to the task data volume;

[0084] S201, classifying the task data in a format based on a preset time node to obtain a plurality of classified task formats;

[0085] S202, obtaining a plurality of first processing times corresponding to a plurality of classification task formats, and accumulating the plurality of first processing times to obtain a first total processing time;

[0086] S203: Calculate the difference between the preset processing time and the first processing total time to obtain the difference time, and use the difference time as the task delay time.

[0087] As described in the above steps S201-S203, the present invention first performs format classification on the task data based on the preset time node to obtain multiple classified task formats. In this way, standardized format classification of task data can provide an important basis for extracting the task format. Then, multiple first processing times corresponding to the multiple classified task formats are obtained, and the multiple first processing times are accumulated to obtain the first total processing time. In this way, the processing time of each classified task format can be counted, and a basis for optimizing the task processing time can be provided. Finally, the difference between the preset processing time and the first total processing time is calculated to obtain the difference time, and the difference time is used as the task delay time. By calculating the task delay time, it can provide an accurate basis for standardizing the judgment of the additional error time generated by the task data during the processing process. At the same time, it can also reduce the error in the calculation of the result response time and provide calculation support for the calculation of the result response time.

[0088] In one embodiment, the step S3 of determining the task data structure type according to the plurality of node task structures comprises:

[0089] S301, acquiring structure information of each task data structure type according to the plurality of node task structures, wherein the structure information includes at least one of text structure information, image structure information and video structure information;

[0090] S302, determining in sequence whether the plurality of structure information are consistent with a preset unstructured data type;

[0091] If the structure information is consistent with the preset unstructured data type, taking the task data structure type corresponding to the plurality of structure information as the unstructured data type, and obtaining a first quantity of the plurality of structure information corresponding to the unstructured data type;

[0092] If the structure information is inconsistent with the preset unstructured data type, taking the task data structure type corresponding to the plurality of structure information as the structure data type, and obtaining a second quantity of the plurality of structure information corresponding to the structure data type;

[0093] S303: Obtain a quantity proportion value according to the first quantity and the second quantity, and determine whether the quantity proportion value is greater than a preset quantity threshold;

[0094] If the quantity ratio is greater than a preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as an unstructured data type;

[0095] If the quantity ratio is not greater than a preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as a structured data type.

[0096] As described in the above steps S301-S303, since the unstructured data type is mainly reflected in image data, etc., and image data mainly appears in the text structure information, image structure information and video structure information, etc., therefore, the present invention first obtains the structure information of each task data structure type according to the multiple node task structures, wherein the structure information includes at least one of the text structure information, image structure information and video structure information, so as to ensure that each task data structure type is correctly identified, providing a basis for subsequent processing, and then judges in turn whether the multiple structure information is consistent with the preset unstructured data type. If the structure information is consistent with the preset unstructured data type, the task data structure type corresponding to the multiple structure information is taken as the unstructured data type, and the first number of the multiple structure information corresponding to the unstructured data type is obtained. If the structure information is consistent with the preset unstructured data type, If the types are inconsistent, the task data structure type corresponding to the multiple structural information is taken as the structured data type, and the second quantity of the multiple structural information corresponding to the structured data type is obtained, so that the first quantity and the second quantity can provide a basis for the selection of task data computing power, and then the quantity proportion value is obtained according to the first quantity and the second quantity, and it is determined whether the quantity proportion value is greater than the preset quantity threshold. If the quantity proportion value is greater than the preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as an unstructured data type. If the quantity proportion value is not greater than the preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as a structured data type. After determining the task data structure type, it can provide a basis for the subsequent optimal model selection. At the same time, through type determination, the task scheduling strategy is optimized to improve the overall efficiency of task processing.

[0097] In one embodiment, the step S5 of calculating the first result response time according to the first computing power value, the first image computing power value, the task data volume and the task delay time includes:

[0098] S501. Calculate a first result response time according to the first computing power value, the first image computing power value, the task data volume, and the task delay time, wherein the calculation formula is:

[0099]

[0100] in, Indicates the first result response time, Indicates the first computing power value, Indicates the computing power value of the first image, Indicates the amount of task data, Indicates the task delay time.

[0101] As described in the above step S501, the present invention calculates the first result response time based on the first computing power value, the first image computing power value, the task data volume and the task delay time, so that the structured data type can be quantified to evaluate the task processing time, and can also provide computing data support for the image and computing proportions in the subsequent task feature attributes.

[0102] In one embodiment, the step S6 of calculating the second result response time according to the second computing power value, the second image computing power value, the task data volume and the task delay time includes:

[0103] S601, calculating a second result response time according to the second computing power value, the second image computing power value, the task data volume and the task delay time, wherein the calculation formula is:

[0104]

[0105] in, Indicates the second result response time, Indicates the second computing power value, Indicates the computing power value of the second image, Indicates the amount of task data, Indicates the task delay time.

[0106] As described in the above step S601, the present invention calculates the second result response time based on the second computing power value, the second image computing power value, the task data volume and the task delay time, so that the evaluation task processing time of the unstructured data type can be quantified, and computing data support can be provided for the image and computing proportions in the subsequent task feature attributes.

[0107] This application also provides a model selection system for complex scene tasks, including:

[0108] The first acquisition module 1 is used to acquire task characteristic attributes for complex scenarios, wherein the task characteristic attributes include task data structure characteristic information, task data volume and task time limit characteristic information;

[0109] A second acquisition module 2 is used to acquire the task delay time according to the task time limit feature information;

[0110] The first splitting module 3 is used to split the task data structure characteristic information according to preset time nodes to obtain multiple node task structures, and determine the task data structure type according to the multiple node task structures, wherein the task data structure type includes structured data type and unstructured data type;

[0111] A third acquisition module 4 is used to acquire first computing power characteristic information of the structured data type, acquire a first computing power value and a first image computing power value according to the first computing power characteristic information, and calculate a first result response time according to the first computing power value, the first image computing power value, the task data volume, and the task delay time;

[0112] a fourth acquisition module 5, configured to acquire second computing power characteristic information of an unstructured data type, acquire a second computing power value and a second image computing power value according to the second computing power characteristic information, and calculate a second result response time according to the second computing power value, the second image computing power value, the task data volume, and the task delay time;

[0113] A first calculation module 6, used to calculate a result response time ratio value according to the first result response time and the second result response time;

[0114] The first judgment module 7 is used to judge whether the result response time ratio is greater than a preset threshold;

[0115] If the result response time ratio is greater than a preset threshold, the task characteristic attribute is determined to be a task with a high calculation ratio, and a calculation power model for arranging tasks for complex scenarios is processed;

[0116] If the result response time ratio is not greater than a preset threshold, the task characteristic attribute is determined to be a task with a high image ratio, and an image computing model is arranged to process the task for complex scenes.

[0117] In one embodiment, the first acquisition module includes:

[0118] A first acquisition unit is used to acquire task data for complex scenarios, acquire a corresponding task data format for complex scenarios according to the task data, and analyze the task data format based on Python to obtain task type information and data format information;

[0119] A first recognition unit, configured to recognize data format information based on a preset image data format to obtain a plurality of image data formats;

[0120] A second identification unit is used to identify the data format information based on a preset computing power structure data format to obtain a plurality of computing power structure data formats, wherein the preset computing power structure data format is a preset fixed format;

[0121] The feature information set corresponding to the multiple image data formats and the multiple computing power structure data formats is used as the task data structure feature information.

[0122] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned model selection method for complex scene tasks are implemented.

[0123] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0124] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0125] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A model selection method for complex scene tasks, characterized in that: include: Acquire task characteristic attributes for complex scenarios, where the task characteristic attributes include task data structure characteristic information, task data volume, and task time limit characteristic information; Acquire the task delay time according to the task time limit characteristic information; The task data structure characteristic information is split according to preset time nodes to obtain a plurality of node task structures, and the task data structure type is determined according to the plurality of node task structures, wherein the task data structure type includes a structured data type and an unstructured data type; Obtaining first computing power characteristic information of a structured data type, obtaining a first computing power value and a first image computing power value according to the first computing power characteristic information, and calculating a first result response time according to the first computing power value, the first image computing power value, the task data volume, and the task delay time; Acquire second computing power characteristic information of the unstructured data type, acquire a second computing power value and a second image computing power value according to the second computing power characteristic information, and calculate a second result response time according to the second computing power value, the second image computing power value, the task data volume, and the task delay time; Calculate a result response time ratio value according to the first result response time and the second result response time; Determine whether the result response time ratio is greater than a preset threshold; If the result response time ratio is greater than a preset threshold, the task characteristic attribute is determined to be a task with a high calculation ratio, and a calculation power model for arranging tasks for complex scenarios is processed; If the result response time ratio is not greater than a preset threshold, the task characteristic attribute is determined to be a task with a high image ratio, and an image computing model is arranged to process the task for complex scenes.

2. The model selection method for complex scene tasks according to claim 1, characterized in that: The step of obtaining task feature attributes for complex scenarios includes: Acquire task data for complex scenarios, acquire a corresponding task data format for complex scenarios according to the task data, and analyze the task data format based on Python to obtain task type information and data format information; Identifying the data format information based on a preset image data format to obtain a plurality of image data formats; Identify the data format information based on a preset computing power structure data format to obtain a plurality of computing power structure data formats, wherein the preset computing power structure data format is a preset fixed format; The feature information set corresponding to the multiple image data formats and the multiple computing power structure data formats is used as the task data structure feature information.

3. The model selection method for complex scene tasks according to claim 1, characterized in that: The step of obtaining the task delay time according to the task time limit characteristic information includes: Get the preset processing time corresponding to the task data volume; Classifying the task data in a format based on a preset time node to obtain a plurality of classified task formats; Acquire multiple first processing times corresponding to multiple classification task formats, and accumulate the multiple first processing times to obtain a first processing total time; The difference between the preset processing time and the first processing total time is calculated to obtain the difference time, and the difference time is used as the task delay time.

4. The model selection method for complex scene tasks according to claim 1, characterized in that: The step of determining the task data structure type according to the plurality of node task structures comprises: Acquire structure information of each task data structure type according to the plurality of node task structures, wherein the structure information includes at least one of text structure information, image structure information and video structure information; Determining in sequence whether the plurality of structure information are consistent with a preset unstructured data type; If the structure information is consistent with the preset unstructured data type, the task data structure type corresponding to the plurality of structure information is used as the unstructured data type, and a first quantity of the plurality of structure information corresponding to the unstructured data type is obtained; If the structure information is inconsistent with the preset unstructured data type, taking the task data structure type corresponding to the plurality of structure information as the structure data type, and obtaining a second quantity of the plurality of structure information corresponding to the structure data type; Acquire a quantity ratio value according to the first quantity and the second quantity, and determine whether the quantity ratio value is greater than a preset quantity threshold; If the quantity ratio is greater than a preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as an unstructured data type; If the quantity ratio is not greater than a preset quantity threshold, the task data structure type corresponding to the task data structure characteristic information is defined as a structured data type.

5. The model selection method for complex scene tasks according to claim 1, characterized in that: The step of calculating the first result response time according to the first computing power value, the first image computing power value, the task data volume and the task delay time includes: The first result response time is calculated according to the first computing power value, the first image computing power value, the task data volume and the task delay time, wherein the calculation formula is: in, Indicates the first result response time, Indicates the first computing power value, Indicates the computing power value of the first image, Indicates the amount of task data, Indicates the task delay time.

6. The model selection method for complex scene tasks according to claim 1, characterized in that: The step of calculating the second result response time according to the second computing power value, the second image computing power value, the task data volume and the task delay time includes: The second result response time is calculated according to the second computing power value, the second image computing power value, the task data volume and the task delay time, wherein the calculation formula is: in, Indicates the second result response time, Indicates the second computing power value, Indicates the computing power value of the second image, Indicates the amount of task data, Indicates the task delay time.

7. A model selection system for complex scene tasks, characterized in that: include: The first acquisition module is used to acquire task characteristic attributes for complex scenarios, wherein the task characteristic attributes include task data structure characteristic information, task data volume and task time limit characteristic information; A second acquisition module is used to acquire the task delay time according to the task time limit feature information; A first splitting module is used to split the task data structure characteristic information according to preset time nodes to obtain multiple node task structures, and determine the task data structure type according to the multiple node task structures, wherein the task data structure type includes structured data type and unstructured data type; a third acquisition module, configured to acquire first computing power characteristic information of the structured data type, acquire a first computing power value and a first image computing power value according to the first computing power characteristic information, and calculate a first result response time according to the first computing power value, the first image computing power value, the task data volume, and the task delay time; a fourth acquisition module, configured to acquire second computing power characteristic information of an unstructured data type, acquire a second computing power value and a second image computing power value according to the second computing power characteristic information, and calculate a second result response time according to the second computing power value, the second image computing power value, the task data volume, and the task delay time; A first calculation module, used to calculate a result response time ratio value according to the first result response time and the second result response time; A first judgment module is used to judge whether the result response time ratio is greater than a preset threshold; If the result response time ratio is greater than a preset threshold, the task characteristic attribute is determined to be a task with a high calculation ratio, and a calculation power model for arranging tasks for complex scenarios is processed; If the result response time ratio is not greater than a preset threshold, the task characteristic attribute is determined to be a task with a high image ratio, and an image computing model is arranged to process the task for complex scenes.

8. The model selection system for complex scene tasks according to claim 7, characterized in that: The first acquisition module includes: A first acquisition unit is used to acquire task data for complex scenarios, acquire a corresponding task data format for complex scenarios according to the task data, and analyze the task data format based on Python to obtain task type information and data format information; A first recognition unit, used to recognize data format information based on a preset image data format to obtain a plurality of image data formats; A second identification unit is used to identify the data format information based on a preset computing power structure data format to obtain a plurality of computing power structure data formats, wherein the preset computing power structure data format is a preset fixed format; The feature information set corresponding to the multiple image data formats and the multiple computing power structure data formats is used as the task data structure feature information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images