Task processing method and device, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a task processing method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of computers, in particular to the fields of artificial intelligence, cloud technology and the like, and the method can comprise the steps that object data of all objects and a target task are obtained; the target task comprises at least one of a training task, an optimization task or an intention degree rating task; wherein the training task is used for training the initial model to obtain an intention degree rating model, the optimization task is used for optimizing the intention degree rating model, and the intention degree rating task is used for predicting the intention degree corresponding to each object through the intention degree rating model; determining a target path based on the target task; wherein the target path represents each processing node which needs to pass through in sequence for executing the target task and the operation executed by each processing node; and executing the target task according to the target path based on the object data of each object.

Description

Technical Field

[0001] The present application belongs to the field of computer technology and may involve fields such as artificial intelligence and cloud technology. Specifically, the present application relates to a task processing method, device, electronic device, storage medium and program product. Background Art

[0002] With the development of information technology, various industries have also developed accordingly. The development of various industries is inseparable from the participation of user objects. Therefore, predicting the intention of each object and conducting intention rating has become a key issue. For example, predicting the intention of a certain object for a certain product, or the intention of a certain object to click on an advertisement, etc.

[0003] Intention rating is a systematic recommendation project that requires consideration of multiple steps, including data access, feature construction, model training, model deployment, and iterative optimization. In related technologies, different managers are typically responsible for different steps. Consequently, implementation may involve duplication, or data processed in one step may not be suitable for the next. This leads to increased complexity in implementing intention rating and slow overall delivery efficiency. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to provide a method, device, electronic device, storage medium, and program product for task processing that can reduce the complexity of implementing intention rating and improve overall delivery efficiency. To achieve the above purpose, the technical solutions provided by the embodiments of the present application are as follows:

[0005] In a first aspect, a task processing method is provided, comprising:

[0006] Obtaining object data and a target task for each object; the target task includes: at least one of a training task, an optimization task, or an intention rating task; wherein the training task is used to train an initial model to obtain an intention rating model, the optimization task is used to optimize the intention rating model, and the intention rating task is used to predict the intention corresponding to each of the objects using the intention rating model;

[0007] Determine a target path based on the target task; wherein the target path represents each processing node that needs to be sequentially passed through to execute the target task and the operations performed by each processing node;

[0008] The target task is executed based on the object data of each object and according to the target path.

[0009] In a possible implementation, determining a target path based on the target task; and executing the target task based on the object data of each object and according to the target path includes:

[0010] Based on the target task, determining an initial processing node;

[0011] Obtaining a target configuration file corresponding to the initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node; wherein the input data corresponding to the first initial node is object data of each object;

[0012] The following operations are performed repeatedly until the preset conditions are met:

[0013] Using the next processing node as the initial node, and using the input data corresponding to the next node as the input data corresponding to the initial node;

[0014] Obtaining a target configuration file corresponding to the initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node;

[0015] The target configuration file corresponding to any processing node includes node information of the next processing node corresponding to the processing node, and the processing operation corresponding to the processing node;

[0016] The preset condition includes: a configuration file of a processing node corresponding to the current processing operation indicates that the target task is completed.

[0017] In another possible implementation, if a processing node corresponds to at least two configuration files, then:

[0018] The obtaining of the target configuration file corresponding to the initial processing node includes:

[0019] Determining the data type of the input data corresponding to the initial processing node;

[0020] Based on the data type corresponding to the input data, a target configuration file corresponding to the initial processing node is determined from at least two configuration files corresponding to the initial processing node.

[0021] In another possible implementation, determining a target path based on the target task includes:

[0022] Determining a configuration file corresponding to the target task based on the target task and a mapping relationship between the task and the configuration file; wherein the configuration file corresponding to the target task is configured with each processing node required to execute the target task and an operation performed by each processing node;

[0023] Based on the configuration file, the target path is determined.

[0024] In another possible implementation, the method further includes:

[0025] Creating an intention rating system; wherein the intention rating system comprises: a plurality of layer structures; each layer structure comprises at least one processing node;

[0026] Based on the intention rating system, the paths corresponding to the respective tasks are determined. Any path includes the layer structures, processing nodes in each layer structure, and operations performed by each processing node required to execute the task.

[0027] In another possible implementation, the target task includes a training task; the object data of each object is training data; the multiple layer structures include at least one of an application layer, a model layer, a feature layer, or a data layer;

[0028] The executing the target task based on the object data of each object and according to the target path includes:

[0029] The training data is processed through the data interface of the data layer, and parallel processing, data cleaning processing and matching intersection processing are performed in sequence to obtain the output result of the data layer;

[0030] The output result of the data layer is sequentially passed through the feature engineering and sample construction of the feature layer to obtain the output result of the feature layer;

[0031] The output results of the feature layer are sequentially subjected to model library modeling, model evaluation and model optimization of the model layer to obtain the output results of the feature layer;

[0032] Based on the output results of the feature layer and combined with the business logic of the application layer, deployment updates are performed to obtain the intention rating model.

[0033] In another possible implementation, the target task includes a training task; the object data of each object is prediction data; the multiple layer structures include at least one of an application layer, a model layer, a feature layer, or a data layer;

[0034] The executing the target task based on the object data of each object and according to the target path includes:

[0035] The predicted data is processed through the data interface of the data layer, and parallel processing, data cleaning processing and matching intersection processing are performed in sequence to obtain the output result of the data layer;

[0036] The output result of the data layer is sequentially passed through the feature engineering, sample construction and feature detection of the feature layer to obtain the output result of the feature layer;

[0037] The output results of the feature layer are used to predict the intentionality of the application layer to obtain the intentionality rating results corresponding to each object;

[0038] The effect is detected through the application layer based on the intention rating results corresponding to each object.

[0039] In another possible implementation, the target task includes: an optimization task; the object data of each object is retrieved from the data layer;

[0040] The executing the target task based on the object data of each object and according to the target path includes:

[0041] Based on the object data of each object, feature engineering and sample construction are performed through the feature layer to obtain an output result of the feature layer;

[0042] The output result of the feature layer is processed through the model layer for model iteration;

[0043] Based on the results of model iterative processing, the model is updated through the application layer.

[0044] In another possible implementation, the method further includes:

[0045] Get the output of the feature engineering node;

[0046] At least one of feature value evaluation or feature anomaly detection is performed based on the output result of the feature engineering node.

[0047] In a second aspect, a task processing apparatus is provided, comprising:

[0048] A first acquisition module is configured to acquire object data and a target task for each object; the target task includes at least one of a training task, an optimization task, or an intentionality rating task; wherein the training task is configured to train an initial model to obtain an intentionality rating model, the optimization task is configured to optimize the intentionality rating model, and the intentionality rating task is configured to predict the intentionality corresponding to each object using the intentionality rating model;

[0049] A first determining module is configured to determine a target path based on the target task, wherein the target path represents each processing node that needs to be sequentially passed through to execute the target task and an operation performed by each processing node;

[0050] The task execution module is used to execute the target task based on the object data of each object and according to the target path.

[0051] In a possible implementation, when the task execution module determines a target path based on the target task and executes the target task based on the object data of each object and according to the target path, it is specifically configured to:

[0052] Based on the target task, determining an initial processing node;

[0053] Obtaining a target configuration file corresponding to the initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node; wherein the input data corresponding to the first initial node is object data of each object;

[0054] The following operations are performed repeatedly until the preset conditions are met:

[0055] Using the next processing node as the initial node, and using the input data corresponding to the next node as the input data corresponding to the initial node;

[0056] Obtaining a target configuration file corresponding to the initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node;

[0057] The target configuration file corresponding to any processing node includes node information of the next processing node corresponding to the processing node, and the processing operation corresponding to the processing node;

[0058] The preset condition includes: a configuration file of a processing node corresponding to the current processing operation indicates that the target task is completed.

[0059] In another possible implementation, if a processing node corresponds to at least two configuration files, the task execution module, when obtaining the target configuration file corresponding to the initial processing node, is specifically configured to:

[0060] Determining the data type of the input data corresponding to the initial processing node;

[0061] Based on the data type corresponding to the input data, a target configuration file corresponding to the initial processing node is determined from at least two configuration files corresponding to the initial processing node.

[0062] In another possible implementation, when determining the target path based on the target task, the first determining module is specifically configured to:

[0063] Determining a configuration file corresponding to the target task based on the target task and a mapping relationship between the task and the configuration file; wherein the configuration file corresponding to the target task is configured with each processing node required to execute the target task and an operation performed by each processing node;

[0064] Based on the configuration file, the target path is determined.

[0065] In another possible implementation, the apparatus further includes: a creation module and a second determination module, wherein:

[0066] The creation module is used to create an intention rating system; wherein the intention rating system includes: a plurality of layer structures; each layer structure includes at least one processing node;

[0067] The second determination module is used to determine the paths corresponding to each task based on the intention rating system. Any path includes the layer structure, processing nodes in each layer structure, and operations performed by each processing node required to execute the task.

[0068] In another possible implementation, the target task includes a training task; the object data of each object is training data; the multiple layer structures include at least one of an application layer, a model layer, a feature layer, or a data layer;

[0069] When executing the target task based on the object data of each object and according to the target path, the task execution module is specifically configured to:

[0070] The training data is processed through the data interface of the data layer, and parallel processing, data cleaning processing and matching intersection processing are performed in sequence to obtain the output result of the data layer;

[0071] The output result of the data layer is sequentially passed through the feature engineering and sample construction of the feature layer to obtain the output result of the feature layer;

[0072] The output results of the feature layer are sequentially subjected to model library modeling, model evaluation and model optimization of the model layer to obtain the output results of the feature layer;

[0073] Based on the output results of the feature layer and combined with the business logic of the application layer, deployment updates are performed to obtain the intention rating model.

[0074] In another possible implementation, the target task includes a training task; the object data of each object is prediction data; the multiple layer structures include at least one of an application layer, a model layer, a feature layer, or a data layer;

[0075] When executing the target task based on the object data of each object and according to the target path, the task execution module is specifically configured to:

[0076] The predicted data is processed through the data interface of the data layer, and parallel processing, data cleaning processing and matching intersection processing are performed in sequence to obtain the output result of the data layer;

[0077] The output result of the data layer is sequentially passed through the feature engineering, sample construction and feature detection of the feature layer to obtain the output result of the feature layer;

[0078] The output results of the feature layer are used to predict the intentionality of the application layer to obtain the intentionality rating results corresponding to each object;

[0079] The effect is detected through the application layer based on the intention rating results corresponding to each object.

[0080] In another possible implementation, the target task includes: an optimization task; the object data of each object is retrieved from the data layer;

[0081] When executing the target task based on the object data of each object and according to the target path, the task execution module is specifically configured to:

[0082] Based on the object data of each object, feature engineering and sample construction are performed through the feature layer to obtain an output result of the feature layer;

[0083] The output result of the feature layer is processed through the model layer for model iteration;

[0084] Based on the results of model iterative processing, the model is updated through the application layer.

[0085] In another possible implementation, the device further includes: a second acquisition module and a feature processing module, wherein:

[0086] The second acquisition module is used to obtain the output result of the feature engineering node;

[0087] The feature processing module is used to perform at least one of feature value evaluation or feature anomaly detection based on the output results of the feature engineering node.

[0088] In a third aspect, an embodiment of the present application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the task processing method provided by any possible implementation method of the first aspect.

[0089] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the task processing method provided by any possible implementation of the first aspect.

[0090] In a fifth aspect, an embodiment of the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the task processing method provided by any possible implementation of the first aspect.

[0091] The beneficial effects of the technical solution provided in the embodiments of the present application are as follows:

[0092] The embodiments of the present application provide a method, apparatus, device, medium and product for task processing. Compared with related technologies, in the embodiments of the present application, when the object data of each object and a task in the intention rating to be executed are obtained, the target task can be used to determine the processing nodes that need to be passed through in sequence to execute the target task and the operations performed by each processing node. Then, based on the object data of each object and according to the target path, the target task is executed. That is to say, before executing a task of intention rating processing, the processing nodes that need to be passed through in sequence to execute the task and the operations performed by each processing node can be determined in advance, so that the task can be automatically executed based on the determined path. Compared with related technologies, the complexity of the intention rating can be reduced, thereby improving the overall delivery efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments of the present application.

[0094] Figure 1 A schematic diagram of an intention rating provided in an embodiment of the present application;

[0095] Figure 2a A schematic diagram of a business process for intention rating provided in an embodiment of the present application;

[0096] Figure 2bA schematic diagram of another business process for intention rating provided in an embodiment of the present application;

[0097] Figure 3 A schematic diagram of a system architecture for intention rating provided in an embodiment of the present application;

[0098] Figure 4 A flowchart of a task processing method provided in an embodiment of the present application;

[0099] Figure 5 A schematic diagram of the feature engineering configuration provided in the embodiment of this application;

[0100] Figure 6 A schematic diagram of a system architecture for executing a training task based on intention rating in an embodiment of the present application;

[0101] Figure 7 A schematic diagram of a system architecture for performing prediction tasks based on intention rating in an embodiment of the present application;

[0102] Figure 8a A schematic diagram of a system architecture for performing optimization tasks based on intention rating in an embodiment of the present application;

[0103] Figure 8b A flowchart of the prediction and training process for performing intention rating in an embodiment of the present application;

[0104] Figure 9 A schematic diagram of a task processing device structure provided in an embodiment of the present application;

[0105] Figure 10 A schematic diagram of the device structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0106] The following describes the embodiments of the present application in conjunction with the accompanying drawings. It should be understood that the embodiments described below in conjunction with the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions of the embodiments of the present application.

[0107] Those skilled in the art will understand that, unless otherwise stated, the singular forms "a", "an", "said", and "the" used herein may also include plural forms. It should be further understood that the terms "including" and "comprising" used in the embodiments of the present application mean that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements, and / or components, but do not exclude implementation as other features, information, data, steps, operations, elements, components, and / or combinations thereof supported by the present technical field. It should be understood that when we say that an element is "connected" or "coupled" to another element, the element can be directly connected or coupled to the other element, or it can refer to the element and the other element establishing a connection relationship through an intermediate element. In addition, the "connection" or "coupling" used here can include wireless connection or wireless coupling. The term "and / or" used here indicates at least one of the items defined by the term, for example, "A and / or B" can be implemented as "A", or as "B", or as "A and B". When describing multiple (two or more) items, if the relationship between the multiple items is not clearly defined, the multiple items may refer to one, multiple or all of the multiple items. For example, the description of "parameter A includes A1, A2, A3" can be implemented as parameter A including A1 or A2 or A3, and can also be implemented as parameter A including at least two of the three items A1, A2, and A3.

[0108] Before further describing the embodiments of the present application in detail, the nouns and terms involved in the embodiments of the present application are explained. The nouns and terms involved in the embodiments of the present application are subject to the following interpretations.

[0109] Intention rating is an application of recommendation algorithms in real-world scenarios. It specifically refers to the prediction of the strength of an object's intention to engage in a certain behavior, such as whether to click on an ad, participate in an event, or purchase a product.

[0110] Object attribute information, by collecting and analyzing various information about an object, is abstracted into a virtual representation with specific characteristics, expressed as a high-dimensional vector during the modeling process. This helps products better understand the object and provide better services for it.

[0111] The AutoRec model is a standard autoencoder. Its basic principle is to use the co-occurrence matrix in collaborative filtering to autoencode item vectors or user vectors. The autoencoding results are then used to obtain the user's estimated rating for the item, which is then used to rank recommendations.

[0112] Recommendation algorithms in related technologies primarily rely on traditional or deep learning recommendation models, supplemented by human-developed strategies. The development of traditional recommendation models primarily consists of the following components: 1) collaborative filtering algorithms; 2) logistic regression models; 3) factorization machine models; and 4) combination models. Deep learning recommendation models, with multi-layer perceptrons at their core, offer a variety of different deep learning recommendation models, with development directions including:

[0113] 1) Changing the complexity of the neural network, such as AutoRec and Deep Crossing;

[0114] 2) Changing the feature crossover method, such as Neural Collaborative Filtering (NeuralCF) and Probabipstic Neural Network (PNN);

[0115] 3) Combination models, such as wide and deep models (Wide & Deep Network), deep cross network (Deep & Cross Network), factorization machine (DeepFM);

[0116] 4) Deep FM models, such as Neural Factorization Machine (NFM), Feedforward Neural Network (FNN), and Attentional Factorization Machine (AFM);

[0117] 5) Combined with attention mechanism, such as AFM and Deep Interest Network (DIN);

[0118] 6) Combined with sequence models, such as Deep Interest Evolution Network (DIEN);

[0119] 7) Combined with reinforcement learning, deep residual networks (DRN).

[0120] Related technologies have proposed methods and optimization ideas for solving prediction problems at the model level, and have achieved good performance in most scenarios. However, intention rating is a systematic recommendation project that requires consideration of multiple links, including data access, feature construction, model training, model deployment, and optimization iterations. Related technologies focus on the model training link and partially solve the problem of feature construction. However, there are many other links in the overall intention rating task that are not covered. When faced with multiple modeling requirements, manual completion of operations in other links is required, which is somewhat repetitive and affects overall delivery efficiency.

[0121] Taking a specific application scenario as an example, the application scenario of the embodiment of this application is a data intelligence scenario, which is specifically applicable to scenarios including clue rating, private domain retention, silent insurance customers, etc., and is mainly used for the intention rating of the target purchase / store visit behavior. Figure 1 As shown;

[0122] 1) Input a list of objects to be rated, which contains the unique identification ID of each object to be rated;

[0123] 2) The algorithm performs calculations and outputs the object's intentionality (probability);

[0124] The intention of each object is converted into a rating result, that is, the rating score of each object. This conversion process usually complies with business needs, such as mapping the probability value of 0-1 to 1-5 points in a certain ratio, with 1 point as the lowest intention and 5 points as the highest intention.

[0125] In the embodiment of the present application, the specific method of outputting the object intention degree (probability) by the algorithm is as follows: Figure 2a and Figure 2b As shown, the object opens the client page and enters the list of objects to be rated;

[0126] 1) The client obtains the object list, and the client transmits the object list back to the backend server;

[0127] 2) The backend server uses the object list to match the corresponding object attribute information from the database;

[0128] 3) Inputting object attribute information into the intentionality prediction model;

[0129] 4) The model outputs the intentionality of each object and returns it to the backend server;

[0130] 5) The backend server converts the intention degree into a rating result and returns it to the client and customer;

[0131] 6) If the model requires iterative optimization, the object label can be passed to the model through the backend server to update the model;

[0132] Figure 2a and Figure 2b The difference lies in the intention rating link, that is, the data elements that the algorithm link can receive. Figure 2a The algorithm only receives object attribute information and object labels. Figure 2b The algorithm can receive object ID, object attribute information, and object label. For some business scenarios with sensitive data security requirements, Figure 2a The framework shown can better protect object privacy and can also be extended to federated modeling.

[0133] In reality, as business continues to expand, algorithm developers face a massive demand for modeling. Beyond core algorithm training, they also need to complete numerous interconnected tasks, often with a high degree of repetition. Efficiently building the entire pipeline has become a pressing challenge for algorithm developers.

[0134] Among them, the solutions provided in the embodiments of this application relate to the field of artificial intelligence. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to achieve the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.

[0135] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, pre-trained models, operating / interaction systems, and mechatronics. Pre-trained models, also known as large models or basic models, can be fine-tuned and widely applied to downstream tasks across various AI disciplines. AI software technologies primarily encompass computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0136] The solutions provided in the embodiments of the present application may specifically relate to natural language processing, machine learning / deep learning and other fields in the field of AI. Natural language processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can enable effective communication between humans and computers using natural language. Natural language processing involves natural language, that is, the language people use in daily life, and is closely related to linguistic research; it also involves computer science and mathematics. The pre-training model, an important technology for model training in the field of artificial intelligence, is developed from the large language model (Large Language Model) in the field of NLP. After fine-tuning, the large language model can be widely used in downstream tasks. Natural language processing technology generally includes text processing, semantic understanding, machine translation, robot question answering, knowledge graphs and other technologies.

[0137] Optionally, the solutions provided by the embodiments of the present application may also involve cloud technology. For example, the solutions provided by the embodiments of the present application may be executed by a server or terminal, wherein the server may be a cloud server. The data processing involved in the implementation of the solutions may be implemented based on cloud technology, and the data storage involved in the implementation may use cloud storage. For example, the configuration files corresponding to each processing node may be stored using cloud storage.

[0138] Cloud computing is a computing model that distributes computing tasks across a resource pool consisting of a large number of computers, enabling various application systems to access computing power, storage space, and information services as needed. The network that provides these resources is called the "cloud." To users, these resources appear infinitely scalable and can be accessed at any time, used on demand, expanded at any time, and paid for on a per-use basis.

[0139] As a provider of cloud computing infrastructure (i.e., a cloud vendor), a cloud computing resource pool (referred to as a cloud platform, commonly referred to as an IaaS (Infrastructure as a Service) platform) is established. Various types of virtual resources are deployed in the resource pool for external customers to choose and use. The cloud computing resource pool mainly includes: computing devices (virtualized machines, including operating systems), storage devices, and network devices.

[0140] Based on logical functional divisions, the PaaS (Platform as a Service) layer can be deployed on top of the IaaS (Infrastructure as a Service) layer, and the SaaS (Software as a Service) layer can be deployed on top of the PaaS layer. SaaS can also be deployed directly on top of IaaS. PaaS is a platform for software execution, such as databases and web containers. SaaS is a variety of business software, such as web portals and text messaging tools. Generally speaking, SaaS and PaaS are upper layers relative to IaaS.

[0141] Cloud technology is a general term for network technology, information technology, integration technology, management platform technology, and application technology based on the cloud computing business model. It can form a resource pool that can be used flexibly and conveniently on demand. Cloud computing technology will become a key support. Cloud storage is a new concept that extends and develops from the concept of cloud computing. A distributed cloud storage system (hereinafter referred to as a storage system) uses cluster applications, grid technology, and distributed storage file systems to bring together a large number of different types of storage devices (also called storage nodes) on the network through application software or application interfaces to work together and provide external data storage and business access functions.

[0142] It should be noted that in the optional embodiments of the present application, the object data and other related data of each object involved, when the embodiments in the present application are applied to specific products or technologies, need to obtain the permission or consent of the object, and the collection, use and processing of the relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions. In other words, if the embodiments of the present application involve data related to the object, it needs to be obtained through the authorization and consent of the object, the authorization and consent of the relevant departments, and in compliance with the relevant laws, regulations and standards of the country and region. If personal information is involved in the embodiments, the acquisition of all personal information requires the consent of the individual. If sensitive information is involved, the separate consent of the information subject needs to be obtained. The embodiments also need to be implemented with the authorization and consent of the object.

[0143] In the embodiment of the present application, an overall solution is proposed. The system architecture of the solution includes 4 layers and 14 modules. Figure 3 As shown in the figure, the system architecture consists of four layers: data layer, feature layer, model layer, and application layer. The details are as follows:

[0144] 1) Data Layer

[0145] The data layer is responsible for acquiring, processing, cleaning, and storing data, as well as matching and intersecting it with other data. The data layer consists of four modules: data interface, parallel processing, data cleaning, and matching and intersection.

[0146] 1.1) Data interface: A unified data interface is used to obtain data from external data sources. The data interface needs to be able to efficiently and stably receive data and convert it into a format suitable for subsequent processing;

[0147] 1.2) Parallel processing: Use parallel processing technology to improve data processing efficiency. Functions need to be able to run in parallel on multiple computing nodes and merge the results after processing is completed;

[0148] 1.3) Data cleaning: Clean and verify the acquired data. The cleaning process should be able to effectively remove noise, outliers, and duplicate data, and ensure the accuracy and completeness of the data;

[0149] 1.4) Matching and intersection: Match and intersect the data with other data.

[0150] 2) Feature layer

[0151] The feature layer extracts useful features from the data and uses them for model training and evaluation. The feature layer consists of three modules: feature engineering, sample construction, and feature detection.

[0152] 1.1) Feature engineering: constructing features based on data. Specific operations include feature selection, feature transformation, feature scaling, etc.

[0153] 1.2) Sample construction: Define positive and negative samples based on business logic. This includes clarifying the definition of positive and negative samples, sample combination, and sample balance;

[0154] 1.3) Feature detection: Regularly check the accuracy and effectiveness of features and promptly detect and fix errors. This includes feature detection, error handling, and feature optimization.

[0155] 3) Model layer

[0156] The model layer completes model training and evaluation, and supports continuous optimization and iteration after launch. The model layer consists of four modules: model library, model evaluation, model optimization, and model iteration;

[0157] 1.1) Model library, which allows users to select and train models suitable for business scenarios, including traditional and deep models, and the ability to combine models;

[0158] 1.2) Model evaluation: evaluate the performance of the model and select the best model from the alternative models;

[0159] 1.3) Model optimization: tuning the model based on the model evaluation results;

[0160] 1.4) Model iteration: Analyze the online effect of the model and perform model iteration.

[0161] 4) Application layer

[0162] Apply the model to actual business scenarios and implement automated intention rating. The application layer consists of three modules: business logic, deployment and update, and effect detection;

[0163] 1.1) Business logic: Define business logic based on actual business scenarios and formulate auxiliary rules to maximize the effectiveness of the model;

[0164] 1.2) Deployment and update, including model deployment and subsequent updates;

[0165] 1.3) Effect detection: detect and analyze the online effect of the model, attribute the effect, and guide feature construction and model iteration.

[0166] The above system architecture can well adapt to the business needs of intention rating. It has been verified in practice that the efficiency of this application for routine modeling tasks is much higher than other existing methods, and the effect is not lower than the average level of existing methods. The machine learning system proposed in the embodiment of this application can complete automated intention rating. The entire framework is highly structured and has strong versatility, plasticity and scalability. In the face of specific scenarios, various variants can be easily created to adapt to various problems. The specific implementation details of the framework can be adjusted according to the actual application scenarios and the experience of algorithm personnel, and are not strictly fixed. In addition, the system can also achieve different levels of performance with different hardware configurations.

[0167] Specifically, the embodiment of the present application provides a method for rating intention, which can be executed by an electronic device, which can be a terminal device or a server, such as Figure 4 As shown, the method may include:

[0168] Step S401: Obtain object data and target tasks of each object.

[0169] Among them, the target task includes: at least one of a training task, an optimization task or an intention rating task; among them, the training task is used to train the intention rating model, the optimization task is used to optimize the intention rating model, and the intention rating task is used to predict the intention corresponding to each object.

[0170] In the embodiment of the present application, a triggering operation of a manager is detected to determine a target task from training tasks, optimization tasks, and intention rating tasks.

[0171] Step S402: Determine the target path based on the target task.

[0172] Among them, the target path characterizes the various processing nodes that need to be passed through in sequence to perform the target task and the operations performed by each processing node. In the embodiment of the present application, the target task to be performed is different, and the various processing nodes that need to be passed through to perform the target task, the order of passing through each processing node, or the operations that need to be performed by each processing node may all be different. Therefore, after determining the target task to be performed, it can be determined which processing nodes need to be passed through in sequence to perform the target task, and which operations need to be performed by each processing node.

[0173] Specifically, in the embodiments of the present application, determining the target path based on the target task may include determining the target path based on the target task and the object data corresponding to each object. That is, in addition to different target tasks, the corresponding target paths may also be different. The target path is also related to the data type of the input data, that is, the type of object data corresponding to each object. In other words, different data types of the input data may result in different nodes passed through, or the processing operations of a processing node may also be different.

[0174] Step S403: Execute the target task based on the object data of each object and according to the target path.

[0175] In the embodiments of the present application, after obtaining the target path, the target task can be automatically executed based on the objects of each object. Specifically, the training task can be automatically executed through each processing node to obtain the intention rating model. Alternatively, the prediction task can be automatically executed through each processing node to obtain the intention rating results corresponding to each object. Alternatively, the optimization task can be automatically executed through each processing node to optimize the intention rating model.

[0176] Specifically, the method of obtaining the object data of each object and the target task can be the acquisition method shown in the above embodiment. In the embodiment of the present application, the object data of each object can be obtained first, and then the target task can be obtained, or the target task can be obtained first, and then the object data of each object can be obtained, or the target task and the object data of each object can be obtained at the same time.

[0177] Specifically, based on the target task, determining the target path, and based on the object data of each object and according to the target path, executing the target task may specifically include: step Sa, step Sb and step Sc, wherein,

[0178] Step Sa: Determine the initial processing node based on the target task.

[0179] As shown in the above intention prediction system architecture, after determining the target task, the initial processing node is determined. The initial processing node is generally a data interface node of the data layer.

[0180] Step Sb: Obtain a target configuration file corresponding to the initial processing node, and based on the target configuration file corresponding to the initial processing node and the input data corresponding to the initial node, perform corresponding operations and determine the next processing node.

[0181] Specifically, a processing node may correspond to at least one configuration file, and each configuration file may be used to indicate the operations performed in the processing node and the next processing node corresponding to the target task. In an embodiment of the present application, the above-mentioned next processing node may be at the same layer as the initial processing node, or at a different layer. For example, the initial processing node is at the data layer, and the next processing node may be at the data layer or at the feature layer. In an embodiment of the present application, the configuration file may be in .json format.

[0182] After determining the initial processing node corresponding to the target task, the target configuration file corresponding to the initial processing node is obtained, the corresponding operation is automatically performed, and after the operation is completed, the output node is transferred to the next processing node. In the embodiment of the present application, the input data corresponding to the first initial node is the object data of each object.

[0183] Step Sc, loop and perform the following operations (operation 1 and operation 2) until the preset conditions are met. In an embodiment of the present application, the preset conditions include: the configuration file of the processing node corresponding to the current processing operation represents the end of the target task. For example, the configuration file may include the node information corresponding to the next processing node. If the configuration file of the processing node corresponding to the current processing operation does not include the node information corresponding to the next processing node, then the target task can be represented as ended; or, the configuration file of the processing node corresponding to the current processing operation carries a target task end identifier, that is, when it is detected that the configuration file of the processing node corresponding to the current processing operation carries a target task end identifier, then the target task is represented as ended.

[0184] Specifically, operations 1 and 2 may be performed as follows:

[0185] Operation 1: Use the next processing node as the initial node, and use the input data corresponding to the next node as the input data corresponding to the initial node.

[0186] Operation 2: Obtain a target configuration file corresponding to the initial processing node, and based on the target configuration file corresponding to the initial processing node and the input data corresponding to the initial node, perform corresponding operations and determine the next processing node.

[0187] The target configuration file corresponding to any processing node includes node information of the next processing node corresponding to the processing node, and the processing operation corresponding to the processing node.

[0188] Furthermore, in an embodiment of the present application, if a processing node corresponds to at least two configuration files, obtaining the target configuration file corresponding to the initial processing node may specifically include: determining the data type of the input data corresponding to the initial processing node; and determining the target configuration file corresponding to the initial processing node from at least two configuration files corresponding to the initial processing node based on the data type corresponding to the input data. As shown above, the initial processing node shown here can execute any processing node corresponding to the target task, that is, the target configuration file can also be determined from at least two configuration files corresponding to the processing node based on the data type of the input data input to a certain processing node. In an embodiment of the present application, for a processing node, the next processing node corresponding to different configuration files corresponding to the processing node may be the same, but the processing operations are different, or the next processing node corresponding to the processing node may be different, but the processing operations are the same, and of course the next processing node corresponding to the processing node and the way of its processing operation may be different.

[0189] For example, for a certain processing node, the configuration file corresponding to the processing node may include: configuration file 1 and configuration file 2; the processing node may be for performing format conversion processing on the input data. If the input data corresponding to the processing node is the gender attribute of the object, that is, male or female, then it can be determined that the corresponding target configuration file may be configuration file 1, and the input data is format converted through configuration file 1, for example, converting "male" to 0 and converting "female" to 1; if the input data corresponding to the processing node is constellation information, then it can be determined that the corresponding target configuration file may be configuration file 2, so as to convert the input constellation information into corresponding recognizable data.

[0190] It can be seen from the above embodiment that a configuration file can be set for each processing node so that a configuration file is determined from each processing node based on the target task and input data, so as to execute the target task according to the configuration file. Another possible implementation method of the embodiment of the present application is to determine the target path based on the target task, which may specifically include: determining the configuration file corresponding to the target task based on the target task and the mapping relationship between the task and the configuration file; and determining the target path based on the configuration file. In the embodiment of the present application, the configuration files corresponding to each task can be determined in advance, and the configuration file corresponding to any task can be the configuration file corresponding to the entire task. The configuration file corresponding to any task also configures which processing nodes need to be passed through to execute the task, and which operations need to be performed through a certain processing node.

[0191] The configuration file corresponding to the target task is configured with the various processing nodes required to execute the target task and the operations performed by each processing node. For example, for the task of training an intention prediction model, the configuration file can be used to indicate: first pass through the data interface node, parallel processing node, data cleaning node, and matching intersection node of the data layer, then pass through the feature engineering node and sample construction node of the feature layer, then pass through the model library modeling node, model evaluation node, and model optimization node of the model layer, and finally pass through the business logic module of the business layer to obtain the final intention prediction model.

[0192] It should be noted that the configuration file for a certain task can record the processing nodes that need to be passed through to execute the task and the operations that need to be performed by each processing node, etc., or the configuration file for a certain task can also include multiple sub-configurations, and a sub-configuration file can correspond to a certain processing node, that is, the operations actually performed by the processing node when executing the target task can be configured in a certain sub-configuration file.

[0193] Specifically, in the feature engineering processing node, the intention rating modeling involves two types of data: dealer data and lead data, among which the lead data is further divided into training data and prediction data according to the modeling training and prediction stages. Therefore, there are three feature engineering links: dealer feature engineering, training data feature engineering, and prediction data feature engineering. The three paths all include four links: "feature type identification and conversion", "outlier detection and processing", "feature construction", and "feature optimization". Different paths are implemented according to the different configuration files called. Dealer feature engineering calls dealer data configuration, and training and prediction feature engineering calls lead data configuration. In the three types of feature engineering processes, the logic will be solidified to generate a mapping dictionary, which is mainly used for the rapid calling of online applications. Evaluation detection mainly realizes the value evaluation of the features finally output by feature engineering and the detection of feature anomalies. Specifically, Figure 5 As shown. In the embodiment of the present application, feature type identification and conversion are used to call the feature configuration file to pre-read the feature type and perform conversion; outlier detection and processing are used to call the definition of outliers in the feature configuration file to identify and process outliers; feature construction is used to call the feature construction list in the feature configuration file to classify and process different types of features; feature optimization is used to call the list of features that need to be optimized in the feature configuration file to perform targeted optimization on the features that need to be optimized. Specifically, while performing feature engineering, reports on feature value evaluation and feature anomaly detection will be generated, and statistical measurements will be made on feature distribution and feature value respectively. In an embodiment of the present application, the prediction data may be the object data of each object to be rated or predicted for intention, so as to determine the intention of each object towards a target object, or the intention level, through these object data. For example, the intention of each object towards a certain car model, or the intention to enter a certain store, or the intention to click on a certain type of advertisement, etc., can be determined through the prediction data (the object data of each object); during the training process, the clue data (for example, may include: the object data of each object, and their corresponding intentions) is used to train the initial model to obtain a corresponding intention rating model; the dealer data may be dealer-related, including car models, car series, store locations, etc., which can be used as input in the prediction stage to determine the intention of each object. Of course, during the training process, it can also be used as training data to train the corresponding intention rating model.

[0194] Further, in the embodiment of the present application, the object data of each object can be obtained through multiple channels, including channel 1, channel 2, channel 3, channel 4, channel 5, channel 6, channel 7, channel 8, channel 9, and channel 10, as shown in Table 1 below. Table 1 also records the number of the above-mentioned object data obtained based on each channel, the number of positive samples, the number of negative samples, the first proportion (the proportion of the amount of data obtained by one channel to all channels), the positive sample proportion, the mapping number (the number in the mapping dictionary), and the truncation number;

[0195] Table 1

[0196]

[0197] Furthermore, while performing feature engineering, feature value evaluation is also required to obtain the feature value of each feature, as shown in Table 2.

[0198] Table 2

[0199] feature value Feature 1 1.065321 Feature 2 1.7815272 Feature 3 0.1228682 Feature 4 0 Feature 5 0 Feature 6 0.2199949 Feature 7 0 Feature 8 0.0063648 Feature 9 0.0063648 Feature 10 0.0063648 Feature 11 0 Feature 12 0.0063648 Feature 13 0.0063648 Feature 14 0 Feature 15 0.0420753 Feature 16 0.0063648

[0200] Furthermore, the method may also include: creating an intention rating system; based on the intention rating system, determining the paths corresponding to each task. The intention rating system includes: multiple layer structures; each layer structure includes at least one processing node. In the embodiment of the present application, any path includes the layer structures required to execute the task, the processing nodes in each layer structure, and the operations performed by each processing node. In the embodiment of the present application, the intention rating system created can be specifically described in the above embodiment and will not be repeated here.

[0201] Specifically, in determining the path corresponding to any task based on the intention rating system, it can specifically include: based on the intention rating system, determining the various processing nodes that need to be passed through to execute the task, and the configuration files corresponding to each processing node. As shown above, the configuration file corresponding to any processing node is used to indicate the operation performed by the processing node, as well as the next processing node of the processing node. In an embodiment of the present application, when generating a configuration file for a processing node for a certain task, the configuration file for the processing node can be generated based on the requirements of the various processing nodes corresponding to the execution of the task, so as to avoid the data output by the previous processing node being unsuitable for the processing requirements of the subsequent processing node, or the situation where the processing operations of the previous processing node and the subsequent processing node are partially repeated.

[0202] Specifically, as shown above, the intention rating system has multiple layer structures including at least one of the application layer, model layer, feature layer, or data layer; in one possible implementation, the target task includes: a training task; the object data of each object is the training data; based on the object data of each object, and in accordance with the target path, the target task is executed, which may specifically include: passing the training data through the data interface of the data layer, and sequentially performing parallel processing, data cleaning processing, and matching intersection processing to obtain the output result of the data layer; passing the output result of the data layer through the feature engineering and sample construction of the feature layer in sequence to obtain the output result of the feature layer; passing the output result of the feature layer through the model library modeling, model evaluation and model optimization of the model layer in sequence to obtain the output result of the feature layer; based on the output result of the feature layer, and combined with the business logic of the application layer, deployment and update are performed to obtain the intention rating model.

[0203] Specifically, if Figure 6 As shown in the figure, for the training task, the training data is processed in parallel, cleaned, and matched through the data interface of the data layer, and then through the feature engineering and sample construction of the feature layer. Furthermore, it is modeled, evaluated, and optimized through the model library of the model layer, and finally deployed and updated through the business logic of the application layer to obtain the intention rating model.

[0204] Take a specific example to introduce the process of executing the training task above. The training data is passed through the data interface node of the data layer to obtain the configuration file a corresponding to the node. The operation is performed based on the configuration file a, and the next processing node is determined to be the parallel processing node of the data layer to obtain the configuration file b corresponding to the node. Parallel processing is performed through the configuration file b, and the next processing node is determined to be the data cleaning processing node to obtain the corresponding configuration file c. Data cleaning processing is performed according to the configuration file c, and the next processing node is determined to be the matching intersection processing node, and the configuration file d corresponding to the node is obtained. Then, according to the configuration file d, the next processing node is determined to be the feature engineering node of the feature layer, and the configuration file e corresponding to the node is obtained. Feature engineering is performed according to the configuration file e, and the next processing node is determined to be the sample construction node, and the corresponding configuration file of the sample construction node is obtained. Configuration file f, execute sample construction according to configuration file f, and determine that the next processing node is the model library modeling node, obtain the configuration file g corresponding to the node, execute modeling processing according to configuration file g, and determine that the next processing node is the model evaluation processing node, obtain the configuration file h corresponding to the node, execute model evaluation processing according to the configuration file h, and determine that the next processing node is the model optimization node, then obtain the configuration file i corresponding to the model optimization node, execute model optimization processing according to the configuration file i, determine that the next processing node is the business logic node, then obtain the configuration file j corresponding to the node, execute business logic processing according to the configuration file j, and determine that the next processing node is the deployment update node, and obtain the configuration file k corresponding to the deployment update node, execute deployment update processing according to the configuration file k to complete the training task and obtain the processed intention prediction model.

[0205] Another possible implementation method of the embodiment of the present application is that the target task includes: a training task; the object data of each object is the prediction data; based on the object data of each object and in accordance with the target path, the target task is executed, which may specifically include: passing the prediction data through the data interface of the data layer, and sequentially performing parallel processing, data cleaning processing, and matching intersection processing to obtain the output result of the data layer; passing the output result of the data layer through the feature engineering, sample construction and feature detection of the feature layer in sequence to obtain the output result of the feature layer; passing the output result of the feature layer through the intentionality prediction of the application layer to obtain the intentionality rating results corresponding to each object; and performing effect detection through the application layer based on the intentionality rating results corresponding to each object.

[0206] Specifically, if Figure 7As shown, the predicted data is processed through the data interface of the data layer, and then parallelized, cleaned, and matched. The data is then processed through feature engineering, sample construction, and feature detection at the feature layer, and then the intention degree is predicted at the application layer, and finally the effect is detected at the application layer.

[0207] The process of executing the prediction task is described using a specific example. The prediction data (the object data corresponding to each object to be predicted for intentionality) is passed through the data interface node of the data layer to obtain the configuration file a1 corresponding to the node. An operation is performed based on the configuration file a1, and the next processing node is determined to be the parallel processing node of the data layer to obtain the configuration file b1 corresponding to the node. Parallel processing is performed using the configuration file b1, and the next processing node is determined to be the data cleaning processing node to obtain the corresponding configuration file c1. Data cleaning processing is performed based on the configuration file c1, and the next processing node is determined to be the matching intersection processing node, and the configuration file d1 corresponding to the node is obtained. Then Perform matching and intersection processing according to the configuration file d1, and determine that the next processing node is the feature engineering node of the feature layer, obtain the configuration file e1 corresponding to the node, perform feature engineering processing, and determine that the next processing node is the sample construction node, and obtain the configuration file f1 corresponding to the node, perform sample construction processing, and determine that the next processing node is the feature detection processing node, obtain the configuration file g1 corresponding to the processing, perform feature detection processing, and determine the intentionality prediction node of the application layer of the next processing node, obtain the configuration file h1 corresponding to the node, perform intentionality prediction processing, and determine that the next processing node is the effect detection node, obtain the configuration file i1 corresponding to the node, and perform effect detection.

[0208] Another possible implementation method of the embodiment of the present application is that the target task includes: an optimization task; the object data of each object is retrieved from the data layer; based on the object data of each object and according to the target path, the target task is executed, which may specifically include: based on the object data of each object, feature engineering and sample construction are performed through the feature layer to obtain the output result of the feature layer; the output result of the feature layer is subjected to model iterative processing through the model layer; based on the model iterative processing result, the model is updated through the application layer.

[0209] Specifically, if Figure 8a As shown in the figure, the data required for optimization is retrieved from the data layer, feature engineering and sample construction are performed in the feature layer, model iteration is completed in the model layer, and the model is updated in the application layer.

[0210] The process of executing the above-mentioned optimization task is introduced using a specific example. The input data is the data retrieved from the application layer. The feature engineering node is performed through the feature layer to obtain the configuration file a2 corresponding to the node, and feature engineering processing is performed. The next processing node is determined to be the sample construction node, and the configuration file b2 corresponding to the node is obtained. Sample construction processing is performed according to the configuration file b2, and the next processing node is determined to be the model iteration processing node of the model layer, and the configuration file b3 corresponding to the processing node is obtained. Model iteration processing is performed according to the configuration file b3, and the next processing node is determined to be the model update node, and the configuration file b4 corresponding to the node is obtained to update the model according to the configuration file b4.

[0211] In another possible implementation of the embodiment of the present application, the method may further include: obtaining the output result of the feature engineering node; and performing at least one of feature value evaluation or feature anomaly detection based on the output result of the feature engineering node.

[0212] Specifically, evaluation detection mainly realizes the value assessment of the features finally output by feature engineering and the anomaly detection of features.

[0213] The following introduces a method of intention rating through a specific implementation method, as shown below:

[0214] 1. Pre-create the system architecture corresponding to the intention rating;

[0215] For the embodiment of the present application, the system architecture corresponding to the intention rating may include: a data layer, a feature layer, a model layer, and an application layer. The data layer includes four processing nodes (also called modules), including: data interface, parallel processing, data cleaning, and matching intersection; the feature layer includes three processing nodes, including: feature engineering, sample construction, and sample detection; the model layer includes four processing nodes, including: model library, model evaluation, model optimization, and model iteration; the application layer may include three processing nodes, including: business logic, deployment update, and effect detection;

[0216] 2. Based on the system architecture corresponding to the intention rating, perform intention rating processing;

[0217] Specifically, in the process of performing intention rating, at least three tasks may be included, for example, a training task, an optimization task, and an intention rating task (also called a prediction task). If the target task is a training task, the training data is processed through the data interface, parallel processing, data cleaning, matching and intersection at the data layer, feature engineering and sample construction are performed at the feature layer, model library modeling, evaluation and optimization are performed at the model layer, and the application layer is deployed in combination with business logic. When the target task is a training task, the input is the training data (in the embodiment of the present application, the training data is the sample data and labels corresponding to each sample object, where the sample data may include: age, gender, etc.), and the output is the intention rating model.

[0218] If the target task is prediction, data passes through the data layer, undergoes feature engineering, sample construction, and feature detection at the feature layer, and then enters the application layer to complete prediction and detection. The input is the prediction data, and the output is the object's intention rating. In this embodiment of the application, the prediction data can be the object data corresponding to each object to be predicted.

[0219] If the target task is an optimization task, combined with the effect detection log, the data required for optimization is retrieved from the data layer, feature engineering and sample construction are performed at the feature layer, model iteration is completed at the model layer, and the model is updated in the application layer.

[0220] Furthermore, the method shown in the embodiment of the present application is highly structured and easy to use, and can be applied to a variety of scenarios, including but not limited to the scenarios defined above, and is basically automated without requiring excessive human intervention.

[0221] Furthermore, if Figure 8bAs shown, the task processing flow based on the above architecture mainly includes a prediction process, that is, a process of predicting the intention of each object, and a process of training to obtain the final intention rating model, specifically including: for the prediction process: first, unified data synchronization is performed, and then the format of the synchronized data is converted, and then field selection is performed on the data after format conversion, and further feature engineering is performed on the data after field selection to obtain a prediction set from the result of feature engineering, and prediction is performed based on the prediction set to obtain the corresponding prediction result, that is, the intention rating result, and feature detection is performed according to the indicators based on the prediction result to detect the prediction result of the model; for the training process: first, unified data synchronization is performed, and then the format of the synchronized data is converted, and then field selection is performed on the data after format conversion, and further feature engineering is performed on the data after field selection, and the training data is divided based on the feature engineering results to train multiple models (Base training model, Auto Xgboost, Single DeepFM, FL DeepFM) to detect the indicators of each model, to select the appropriate model, and perform intention rating or intention prediction.

[0222] Based on the same principle as the task processing method provided in the embodiment of the present application, the embodiment of the present application provides a task processing device, such as Figure 9 As shown, the task processing device 90 may specifically include: a first acquisition module 91, a first determination module 92 and a task execution module 93, wherein:

[0223] A first acquisition module 91 is configured to acquire object data and a target task for each object; the target task includes at least one of a training task, an optimization task, or an intention rating task; wherein the training task is configured to train an initial model to obtain an intention rating model, the optimization task is configured to optimize the intention rating model, and the intention rating task is configured to predict the intention corresponding to each object using the intention rating model;

[0224] A first determining module 92 is configured to determine a target path based on the target task; wherein the target path represents the processing nodes that need to be sequentially passed through to execute the target task and the operations performed by each processing node;

[0225] The task execution module 93 is used to execute the target task based on the object data of each object and according to the target path.

[0226] In one possible implementation of the embodiment of the present application, the task execution module 93 determines a target path based on the target task; and executes the target task based on the object data of each object and according to the target path, specifically for:

[0227] Based on the target task, determine the initial processing node;

[0228] Obtaining a target configuration file corresponding to an initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node; wherein the input data corresponding to the first initial node is object data of each object;

[0229] The following operations are performed repeatedly until the preset conditions are met:

[0230] Using the next processing node as the initial node, and using the input data corresponding to the next node as the input data corresponding to the initial node;

[0231] Obtaining a target configuration file corresponding to an initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node;

[0232] The target configuration file corresponding to any processing node includes node information of the next processing node corresponding to the processing node, and the processing operation corresponding to the processing node;

[0233] The preset condition includes: the configuration file of the processing node corresponding to the current processing operation indicates that the target task is completed.

[0234] In another possible implementation of the embodiment of the present application, if a processing node corresponds to at least two configuration files, the task execution module 93 is specifically configured to:

[0235] Determine the data type of the input data corresponding to the initial processing node;

[0236] Based on the data type corresponding to the input data, a target configuration file corresponding to the initial processing node is determined from at least two configuration files corresponding to the initial processing node.

[0237] In another possible implementation of the embodiment of the present application, when the first determining module 92 determines the target path based on the target task, it is specifically configured to:

[0238] Based on the target task and the mapping relationship between the task and the configuration file, determining the configuration file corresponding to the target task; wherein the configuration file corresponding to the target task is configured with each processing node required to execute the target task and the operation performed by each processing node;

[0239] Determine the target path based on the configuration file.

[0240] In another possible implementation of the embodiment of the present application, the apparatus 90 further includes: a creation module and a second determination module, wherein:

[0241] A creation module is used to create an intention rating system; wherein the intention rating system includes: a plurality of layer structures; each layer structure includes at least one processing node;

[0242] The second determination module is used to determine the paths corresponding to each task based on the intention rating system. Any path includes the layer structure, processing nodes in each layer structure, and operations performed by each processing node required to execute the task.

[0243] In another possible implementation of the embodiment of the present application, the target task includes: a training task; the object data of each object is training data; the multiple layer structures include at least: at least one of: an application layer, a model layer, a feature layer, or a data layer;

[0244] The task execution module 93 is specifically used to execute the target task based on the object data of each object and according to the target path:

[0245] The training data is passed through the data interface of the data layer, and parallel processing, data cleaning processing and matching and intersection processing are carried out in sequence to obtain the output results of the data layer;

[0246] The output of the data layer is sequentially passed through the feature engineering and sample construction of the feature layer to obtain the output of the feature layer;

[0247] The output results of the feature layer are sequentially passed through the model library modeling, model evaluation and model optimization of the model layer to obtain the output results of the feature layer;

[0248] Based on the output results of the feature layer and combined with the business logic of the application layer, deployment updates are performed to obtain the intention rating model.

[0249] In another possible implementation of the embodiment of the present application, the target task includes: a training task; the object data of each object is prediction data; the multiple layer structures include at least: at least one of: an application layer, a model layer, a feature layer, or a data layer;

[0250] The task execution module 93 is specifically used to execute the target task based on the object data of each object and according to the target path:

[0251] The predicted data is passed through the data interface of the data layer, and then parallelized, cleaned, and matched to obtain the output results of the data layer.

[0252] The output results of the data layer are sequentially passed through the feature engineering, sample construction and feature detection of the feature layer to obtain the output results of the feature layer;

[0253] The output of the feature layer is used to predict the intentionality of the application layer to obtain the intentionality rating results corresponding to each object;

[0254] The effect is detected through the application layer based on the intention rating results corresponding to each object.

[0255] In another possible implementation of the embodiment of the present application, the target task includes: an optimization task; object data of each object is retrieved from the data layer;

[0256] The task execution module 93 is specifically used to execute the target task based on the object data of each object and according to the target path:

[0257] Based on the object data of each object, feature engineering and sample construction are performed through the feature layer to obtain the output results of the feature layer;

[0258] The output results of the feature layer are processed through the model layer for model iteration;

[0259] Based on the results of model iterative processing, the model is updated through the application layer.

[0260] In another possible implementation of the embodiment of the present application, the apparatus 90 further includes: a second acquisition module and a feature processing module, wherein:

[0261] The second acquisition module is used to obtain the output results of the feature engineering node;

[0262] The feature processing module is used to perform at least one of feature value evaluation and feature anomaly detection based on the output results of the feature engineering node.

[0263] It should be noted that the first acquisition module 91 and the second acquisition module can be the same acquisition module or different acquisition modules; the first determination module 92 and the second determination module can be the same determination module or different determination modules, which is not limited in the embodiment of the present application.

[0264] An embodiment of the present application provides a device for processing intention rating tasks. Compared with related technologies, in the embodiment of the present application, when the object data of each object and a task in the intention rating to be executed are obtained, the target task can be used to determine the various processing nodes that need to be passed through in sequence to execute the target task and the operations performed by each processing node. Then, based on the object data of each object and according to the target path, the target task is executed. That is to say, before executing a task of intention rating processing, the various processing nodes that need to be passed through in sequence to execute the task and the operations performed by each processing node can be determined in advance, so that the task can be automatically executed based on the determined path. Compared with related technologies, the complexity of intention rating can be reduced, thereby improving the overall delivery efficiency.

[0265] The devices of the embodiments of the present application can execute the methods provided in the embodiments of the present application, and their implementation principles are similar and have corresponding technical effects. The actions performed by each module in the devices of the embodiments of the present application correspond to the steps in the methods of the embodiments of the present application. For detailed functional descriptions of each module of the device, please refer to the descriptions of the corresponding methods shown above, and will not be repeated here.

[0266] Figure 10 A schematic diagram of the structure of an electronic device applicable to the embodiment of the present application is shown in FIG. Figure 10 As shown, the electronic device may be a server or an object terminal, and the electronic device may be used to implement the method provided in any embodiment of the present application.

[0267] like Figure 10 As shown in FIG, the electronic device 2000 may mainly include at least one processor 2001 ( Figure 10 1 ), memory 2002, communication module 2003 and input / output interface 2004 and other components, optionally, the components can be connected and communicated through bus 2005. It should be noted that, Figure 10 The structure of the electronic device 2000 shown in the figure is merely illustrative and does not constitute a limitation on the electronic devices to which the method provided in the embodiments of the present application is applicable.

[0268] Memory 2002 can be used to store operating systems and application programs, etc. Application programs can include computer programs that implement the methods described in the embodiments of the present invention when called by processor 2001, and can also include programs for implementing other functions or services. Memory 2002 can be ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices that can store information and computer programs, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0269] The processor 2001 is connected to the memory 2002 via the bus 2005 and implements corresponding functions by calling the application program stored in the memory 2002. The processor 2001 can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof, which can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the present disclosure. The processor 2001 can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0270] The electronic device 2000 can be connected to a network via a communication module 2003 (which may include, but is not limited to, components such as a network interface) to communicate with other devices (such as a target terminal or server) via the network to implement data interaction, such as sending data to or receiving data from other devices. The communication module 2003 may include a wired network interface and / or a wireless network interface, etc., that is, the communication module may include at least one of a wired communication module and a wireless communication module.

[0271] The electronic device 2000 can be connected to the required input / output devices, such as a keyboard, a display device, etc., through the input / output interface 2004. The electronic device 2000 itself can have a display device, and can also be connected to other external display devices through the interface 2004. Optionally, a storage device, such as a hard disk, can also be connected through the interface 2004, so that data in the electronic device 2000 can be stored in the storage device, or data in the storage device can be read, and data in the storage device can also be stored in the memory 2002. It can be understood that the input / output interface 2004 can be a wired interface or a wireless interface. Depending on the actual application scenario, the device connected to the input / output interface 2004 can be a component of the electronic device 2000, or it can be an external device connected to the electronic device 2000 when needed.

[0272] Bus 2005, used to connect the various components, may include a path for transmitting information between the components. Bus 2005 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Depending on their function, bus 2005 may be categorized as an address bus, a data bus, a control bus, or the like.

[0273] Optionally, for the solution provided in the embodiment of the present application, the memory 2002 can be used to store a computer program for executing the solution of the present application, and be run by the processor 2001. When the processor 2001 runs the computer program, the actions of the method or device provided in the embodiment of the present invention are implemented.

[0274] Based on the same principle as the method provided in the embodiment of the present application, the embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the corresponding content of the aforementioned method embodiment can be implemented.

[0275] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the corresponding content of the aforementioned method embodiment can be implemented.

[0276] It should be noted that the terms "first," "second," "third," "fourth," "1," "2," etc. (if any) in the specification and claims of this application and the accompanying 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 terms used in this manner are interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than that shown or described in the drawings.

[0277] It should be understood that, although each operation step is indicated by arrows in the flowchart of the embodiment of the present application, the order of implementation of these steps is not limited to the order indicated by the arrows. Unless otherwise clearly stated herein, in some implementation scenarios of the embodiment of the present application, the implementation steps in each flowchart can be performed in other orders according to demand. In addition, some or all of the steps in each flowchart can include multiple sub-steps or multiple stages based on actual implementation scenarios. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage in these sub-steps or stages can also be executed at different times respectively. Under different scenarios at the execution time, the execution order of these sub-steps or stages can be flexibly configured according to demand, and the embodiment of the present application does not limit this.

[0278] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program that has a predetermined function and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0279] The above description is only an optional implementation method for some implementation scenarios of this application. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the solution of this application, the use of other similar implementation methods based on the technical ideas of this application also falls within the protection scope of the embodiments of this application.

Claims

1. A task processing method, characterized in that: include: Get the object data and target tasks of each object; The target task includes: at least one of a training task, an optimization task, or an intention rating task; wherein the training task is used to train the initial model to obtain an intention rating model, the optimization task is used to optimize the intention rating model, and the intention rating task is used to predict the intention corresponding to each of the objects using the intention rating model; Determine a target path based on the target task; wherein the target path represents each processing node that needs to be sequentially passed through to execute the target task and the operations performed by each processing node; The target task is executed based on the object data of each object and according to the target path.

2. The method according to claim 1, characterized in that The step of determining a target path based on the target task and executing the target task based on the object data of each object and according to the target path includes: Based on the target task, determining an initial processing node; Obtaining a target configuration file corresponding to the initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node; wherein the input data corresponding to the first initial node is object data of each object; The following operations are performed repeatedly until the preset conditions are met: Using the next processing node as the initial node, and using the input data corresponding to the next node as the input data corresponding to the initial node; Obtaining a target configuration file corresponding to the initial processing node, and performing corresponding operations based on the target configuration file corresponding to the initial processing node and input data corresponding to the initial node, and determining a next processing node; The target configuration file corresponding to any processing node includes node information of the next processing node corresponding to the processing node, and the processing operation corresponding to the processing node; The preset condition includes: a configuration file of a processing node corresponding to the current processing operation indicates that the target task is completed.

3. The method according to claim 2, characterized in that If a processing node corresponds to at least two configuration files, then: The obtaining of the target configuration file corresponding to the initial processing node includes: Determining the data type of the input data corresponding to the initial processing node; Based on the data type corresponding to the input data, a target configuration file corresponding to the initial processing node is determined from at least two configuration files corresponding to the initial processing node.

4. The method according to claim 1, wherein Determining a target path based on the target task includes: Determining a configuration file corresponding to the target task based on the target task and a mapping relationship between the task and the configuration file; wherein the configuration file corresponding to the target task is configured with each processing node required to execute the target task and an operation performed by each processing node; Based on the configuration file, the target path is determined.

5. The method according to claim 1, wherein The method further comprises: Creating an intention rating system; wherein the intention rating system comprises: a plurality of layer structures; each layer structure comprises at least one processing node; Based on the intention rating system, the paths corresponding to the respective tasks are determined. Any path includes the layer structures, processing nodes in each layer structure, and operations performed by each processing node required to execute the task.

6. The method according to claim 5, characterized in that The target task includes: a training task; the object data of each object is training data; the multiple layer structures include at least: at least one of: an application layer, a model layer, a feature layer, or a data layer; The executing the target task based on the object data of each object and according to the target path includes: The training data is processed through the data interface of the data layer, and parallel processing, data cleaning processing and matching intersection processing are performed in sequence to obtain the output result of the data layer; The output result of the data layer is sequentially passed through the feature engineering and sample construction of the feature layer to obtain the output result of the feature layer; The output results of the feature layer are sequentially subjected to model library modeling, model evaluation and model optimization of the model layer to obtain the output results of the feature layer; Based on the output results of the feature layer and combined with the business logic of the application layer, deployment updates are performed to obtain the intention rating model.

7. The method according to claim 5, characterized in that The target task includes: an intention rating task; the object data of each object is prediction data; the multiple layer structures include at least: at least one of: an application layer, a model layer, a feature layer, or a data layer; The executing the target task based on the object data of each object and according to the target path includes: The predicted data is processed through the data interface of the data layer, and parallel processing, data cleaning processing and matching intersection processing are performed in sequence to obtain the output result of the data layer; The output result of the data layer is sequentially passed through the feature engineering, sample construction and feature detection of the feature layer to obtain the output result of the feature layer; The output results of the feature layer are used to predict the intentionality of the application layer to obtain the intentionality rating results corresponding to each object; The effect is detected through the application layer based on the intention rating results corresponding to each object.

8. The method according to claim 5, characterized in that The target task includes: an optimization task; the object data of each object is retrieved from the data layer; The executing the target task based on the object data of each object and according to the target path includes: Based on the object data of each object, feature engineering and sample construction are performed through the feature layer to obtain an output result of the feature layer; The output result of the feature layer is processed through the model layer for model iteration; Based on the results of model iterative processing, the model is updated through the application layer.

9. The method according to any one of claims 6 to 8, characterized in that: The method further comprises: Get the output of the feature engineering node; At least one of feature value evaluation or feature anomaly detection is performed based on the output result of the feature engineering node.

10. A task processing device, characterized in that: include: A first acquisition module is used to acquire object data and target tasks of each object; The target task includes: at least one of a training task, an optimization task, or an intention rating task; wherein the training task is used to train the initial model to obtain an intention rating model, the optimization task is used to optimize the intention rating model, and the intention rating task is used to predict the intention corresponding to each of the objects using the intention rating model; A first determining module is configured to determine a target path based on the target task, wherein the target path represents each processing node that needs to be sequentially passed through to execute the target task and an operation performed by each processing node; The task execution module is used to execute the target task based on the object data of each object and according to the target path.

11. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the task processing method according to any one of claims 1 to 9 when running the computer program.

12. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method for task processing according to any one of claims 1 to 9 is implemented.

13. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the task processing method according to any one of claims 1 to 9 are implemented.

Images