Product repair requesting method and device, storage medium and program product

Abstract

The invention discloses a product repair method and device, a storage medium and a program product, and relates to the technical field of data processing.The product repair method comprises the steps that a product repair list is generated according to product repair information; based on a pre-trained repair-requesting auditing model, auditing the product repair-requesting order to obtain an auditing result, the pre-trained repair-requesting auditing model being obtained by pre-training a preset neural network model based on a back propagation algorithm; according to the audit result, determining whether a maintenance task is generated or not; and if the maintenance task is generated, allocating the maintenance task to maintenance personnel. According to the scheme, the product damage in the product repair information can be accurately judged, and the efficiency of the product repair service and the user satisfaction are effectively improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a product repair reporting method, device, storage medium, and program product. Background Technology

[0002] Existing product repair reporting methods suffer from several problems, impacting user experience and hindering operational efficiency and service quality. First, the review process relies heavily on manual intervention, resulting in slow and inefficient processing. Second, subjective differences in manual review lead to inconsistent review standards, lowering the quality of repair services, negatively affecting user experience, and reducing user satisfaction.

[0003] Therefore, the existing product repair reporting methods suffer from low accuracy and efficiency in product repair review and processing.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a product repair reporting method, device, storage medium, and program product, which aims to solve the technical problem of low accuracy and efficiency in the product repair reporting review and processing of existing product repair methods.

[0006] To achieve the above objectives, this application proposes a product repair reporting method, the method comprising:

[0007] Generate a product repair order based on the product repair information;

[0008] The product repair request is reviewed based on a pre-trained repair review model to obtain the review result. The pre-trained repair review model is obtained by pre-training a preset neural network model based on the backpropagation algorithm.

[0009] Based on the audit results, confirm whether to generate a maintenance task;

[0010] If a maintenance task is generated, the maintenance task is assigned to maintenance personnel.

[0011] In one embodiment, before the step of reviewing the product repair request based on the pre-trained repair review model and obtaining the review result, the method further includes:

[0012] Based on historical product repair data, establish training and validation datasets;

[0013] Effective feature extraction is performed on the training dataset to obtain effective feature data;

[0014] Based on the effective feature data and the backpropagation algorithm, the preset neural network model is trained to adjust the model parameters and obtain the adjusted repair review model.

[0015] Using the validation dataset, the performance of the adjusted repair request review model is evaluated to adjust the model structure and parameters, thereby obtaining a pre-trained repair request review model.

[0016] In one embodiment, the historical product repair data includes several first product repair data sets, and the step of establishing a training dataset and a validation dataset based on the historical product repair data includes:

[0017] Using a preset annotation tool, each first product repair data was annotated multiple times to obtain multiple data annotation results;

[0018] When the multiple data annotation results are the same, a training dataset and a validation dataset are established based on the data annotation results and the first product repair data.

[0019] When the multiple data annotation results are different, the first product repair data is fed back to human for manual re-annotation, and the manual re-annotation result is obtained.

[0020] Based on the manual re-annotation results and the first product repair report data, a training dataset and a validation dataset are established.

[0021] In one embodiment, the step of generating a product repair order based on product repair information includes:

[0022] The product repair information is preprocessed to obtain preprocessed product repair information;

[0023] The image data in the pre-processed product repair information is used to perform image recognition and analysis using a pre-trained repair review model to obtain product damage information;

[0024] A product repair order is generated based on the product damage information and the product repair request information.

[0025] In one embodiment, the step of confirming whether to generate a maintenance task based on the audit result includes:

[0026] When the review result is that the product repair request does not meet the preset product repair standards, the product repair request is marked using the review result and the product repair request is fed back to the user.

[0027] When the review result shows that the product repair request meets the preset product repair standards, a repair task is generated based on the product repair request and repair resources.

[0028] In one embodiment, after the step of assigning the maintenance task to maintenance personnel if a maintenance task is generated, the method includes:

[0029] Collect product repair information within a preset period to construct a second training dataset and a second validation dataset;

[0030] The first repair review model is updated based on the preset incremental learning algorithm and the second training dataset to obtain the second repair review model. The first repair review model is a pre-trained repair review model.

[0031] Using the second verification dataset, the first repair request review model and the second repair request review model are tested and evaluated to obtain the test evaluation results.

[0032] Based on the test evaluation results, a model is selected from the first repair review model and the second repair review model for repair review.

[0033] In one embodiment, after the step of assigning the maintenance task to maintenance personnel if a maintenance task is generated, the method includes:

[0034] Once the maintenance task is completed, collect user feedback information;

[0035] Collect the product repair information and corresponding user feedback information to generate a third training dataset;

[0036] Based on the third training dataset and preset index parameters, the performance of the pre-trained repair review model is evaluated, and the model is optimized based on the evaluation results to obtain a pre-optimized repair review model.

[0037] In addition, to achieve the above objectives, this application also proposes a product repair reporting device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the product repair reporting method as described above.

[0038] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the product repair method described above.

[0039] In addition, to achieve the above objectives, this application also proposes a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the product repair method described above.

[0040] One or more technical solutions proposed in this application have at least the following technical effects:

[0041] The product repair reporting method, device, storage medium, and program product proposed in this application specifically involve generating a product repair order based on product repair information; reviewing the product repair order based on a pre-trained repair review model to obtain a review result, wherein the pre-trained repair review model is obtained by pre-training a preset neural network model based on a backpropagation algorithm; determining whether to generate a repair task based on the review result; and assigning the repair task to repair personnel if a repair task is generated.

[0042] This application first generates a product repair request form based on product repair information, and then uses a pre-trained repair review model to automatically review the product repair request form. Based on the review results and the product repair request form, it determines whether a repair task needs to be generated and assigns it to repair personnel for repair. Through the above method, the automatic review of product repair requests is realized, effectively solving the problem of timeliness of repair request review, effectively improving the review speed of product repair requests, shortening user waiting time, and reducing reliance on manual reviewers, thereby reducing the company's labor costs. At the same time, using an updatable repair review model for repair review can effectively improve the accuracy of product repair review and processing. Attached Figure Description

[0043] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a flowchart illustrating the product repair reporting method in Embodiment 1 of this application.

[0046] Figure 2 This is a flowchart illustrating Embodiment 2 of the product repair method provided in this application;

[0047] Figure 3 This is a flowchart illustrating Embodiment 3 of the product repair method provided in this application;

[0048] Figure 4 This is a flowchart illustrating Embodiment 4 of the product repair method provided in this application;

[0049] Figure 5This is a schematic diagram of the hardware operating environment involved in the product repair reporting method in this application.

[0050] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0051] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0052] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0053] The main solution of this application embodiment is as follows: generate a product repair order based on product repair information; review the product repair order based on a pre-trained repair review model to obtain a review result, wherein the pre-trained repair review model is obtained by pre-training a preset neural network model based on the backpropagation algorithm; determine whether to generate a repair task based on the review result; if a repair task is generated, assign the repair task to the repair personnel.

[0054] Technical terms used in this application:

[0055] Artificial Neural Network (ANN) Model: A neural network model is a computational model inspired by the human brain, composed of a large number of neurons (or nodes). These neurons are connected by weights to form a widely parallel and interconnected network, capable of simulating the interactive responses of biological nervous systems to real-world objects. A neural network can be divided into an input layer, hidden layers, and an output layer. The input layer receives external input data, the hidden layers perform nonlinear transformations on the input data, and the output layer generates the final prediction result.

[0056] In a neural network, a neuron receives n input signals from other connected neurons, each with its own weight. These input signals are weighted and summed, then compared to a threshold of the neuron, processed by an activation function, and finally output as the result.

[0057] Backpropagation Algorithm (BP): The backpropagation algorithm is one of the most core and commonly used optimization algorithms in deep learning. It's an algorithm that uses gradient descent to train multi-layer neural networks. Its basic principle is to calculate the gradient of the loss function with respect to the network parameters, thereby updating the parameters and minimizing the loss function.

[0058] Incremental learning is a method that updates an existing model using new data. Unlike traditional batch learning, incremental learning does not require retraining the entire model; instead, it gradually updates the model parameters to adapt to new data.

[0059] Existing product repair reporting methods have several problems, including: the review process relies heavily on manual operation, resulting in slow and inefficient processing; subjective differences in manual review lead to inconsistent review standards, lowering the quality of product repair services, causing a poor user experience, and reducing user satisfaction; with the increasing number of product repair requests, companies have to increase the number of reviewers, thus increasing labor costs and potentially lowering review quality; furthermore, existing methods relying on manual review lack in-depth analysis and utilization of product repair data, causing companies to miss opportunities to use large amounts of product repair data to optimize maintenance processes, predict equipment failures, and improve service quality.

[0060] Therefore, the existing product repair reporting methods have technical problems such as low accuracy and efficiency in product repair review and processing, and also lead to excessive human resource costs for enterprises.

[0061] This application provides a solution that, based on product repair information, first generates a product repair order, and then uses a pre-trained repair review model to automatically review the product repair order. Based on the review results and the product repair order, it determines whether a repair task needs to be generated and assigns it to repair personnel. Through this method, the automatic review of product repair orders is achieved, effectively solving the problem of timeliness in repair order review, significantly improving the review speed of product repairs, shortening user waiting time, and reducing reliance on manual reviewers, thereby lowering the company's labor costs. Furthermore, using an updatable repair review model can effectively improve the accuracy of product repair review and processing.

[0062] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or a product repair reporting system capable of performing the above functions. The following description uses a product repair reporting system as an example to illustrate this embodiment and the subsequent embodiments.

[0063] Based on this, the embodiments of this application provide a product repair reporting method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the product repair reporting method of this application.

[0064] In this embodiment, the product repair reporting method includes steps S110 to S140:

[0065] Step S110: Generate a product repair order based on the product repair information;

[0066] It is understandable that, since the technical solution proposed in this application is actually applied to the review and processing of product repair requests, step S110 is required. First, the product repair system needs to obtain product repair information, and then generate a corresponding product repair order based on the product repair information, so that the repair review model can be used to understand and judge whether the product repair order meets the repair standards and to automate the review and processing of the product repair order.

[0067] Product repair information refers to the product repair request data submitted by the user, which typically includes user information, product model, product purchase time, and product malfunction information. The product repair form organizes the product repair information in a set format to facilitate subsequent review. It typically includes user information, a unique identifier for the product repair form, the status of the product repair request (e.g., pending review, accepted, completed), estimated repair time, estimated repair cost, and product malfunction information from the product repair information. It should be understood that the format of the product repair form is not unique, and this application does not impose a specific limitation.

[0068] In one implementable manner, step S110 includes steps A01 to A03:

[0069] Step A01: Perform data preprocessing on the product repair information to obtain preprocessed product repair information;

[0070] Understandably, in order to improve the data quality of product repair information and make it more suitable for subsequent data processing and analysis, the product repair system can perform data preprocessing operations on the product repair information, including data cleaning and data standardization. For example, it can remove invalid characters or duplicate content and correct typos from the text information in the product repair information, or remove noise points, blurred areas or irrelevant backgrounds from the image data in the product repair information, and finally obtain the preprocessed product repair information.

[0071] Step A02: Use a pre-trained repair review model to perform image recognition and analysis on the image data in the pre-processed product repair information to obtain product damage information;

[0072] Step A03: Generate a product repair order based on the product damage information and the product repair request information.

[0073] In order to effectively utilize the product repair information submitted by users, the product repair system can extract relevant information about product damage from the image data in the product repair information.

[0074] Specifically, the product repair reporting system utilizes a pre-trained repair review model, such as a pre-trained Convolutional Neural Network (CNN), to identify and extract features from image data through multiple convolutional layers, including image edges, textures, and colors. Based on the extracted image features, the pre-trained repair review model performs classification and identification to determine information such as the type, location, and extent of product damage in the image data, i.e., product damage information.

[0075] Then, the product damage information is integrated with the original product repair information, and the integrated information is then filled into a pre-designed product repair form template. This usually involves inserting specific field values ​​into the corresponding positions in the product repair form template, and finally, a product repair form with the information filled in is obtained.

[0076] Step S120: Based on the pre-trained repair review model, the product repair request is reviewed to obtain the review result. The pre-trained repair review model is obtained by pre-training a preset neural network model based on the backpropagation algorithm.

[0077] It should be noted that the pre-trained repair review model is obtained by relevant personnel based on actual product repair needs. In order to make the repair review model output more accurate review results, the backpropagation algorithm is used to pre-train the repair review model. In this embodiment, the repair review model can be a combination of one or more neural network models in machine learning or deep learning models.

[0078] Specifically, the aforementioned product repair requests are input into a pre-trained repair review model. The model then uses preset product repair standards to predict and infer the reasonableness of the repair requests, providing an assessment of their reasonableness and thus obtaining the review result. These preset product repair standards refer to the criteria pre-set by the company based on actual product or related service standards to determine the reasonableness of repair requests. These standards include provisions regarding the scope and date of product warranty, limits on repair costs, the number of repairs required, and the definition of product defects and flaws.

[0079] Step S130: Based on the audit results, confirm whether to generate a maintenance task;

[0080] Specifically, based on the review results obtained from the pre-trained repair review model, it is determined whether it is necessary to generate corresponding repair tasks based on the product repair order, so that the repair tasks can be assigned to relevant repair personnel to resolve the product repair requests made by users.

[0081] In one implementable embodiment, step S130 includes steps B01 to B02:

[0082] Step B01: When the review result is that the product repair request does not meet the preset product repair standards, the product repair request is marked using the review result and the product repair request is fed back to the user.

[0083] Step B02: When the review result is that the product repair order meets the preset product repair standards, a repair task is generated based on the product repair order and repair resources.

[0084] Understandably, when the review result indicates that a product repair request does not meet the preset product repair standards, meaning the user's product repair request is rejected, the product repair system uses this review result to mark the product repair request. Product repair requests with this mark are those whose product repair requests have not been approved. The product repair system needs to send the product repair requests with the review result mark back to the user, allowing the user to supplement or correct the product repair information so that the product repair request can be reviewed again.

[0085] When the review result indicates that the product repair request meets the preset product repair standards, meaning the user's product repair request is approved, the product repair system intelligently matches the most suitable repair personnel based on the product model and specific fault information in the product repair request, combined with factors such as the repair personnel's current working status, professional skills, geographical location, and past experience in handling similar faults, thereby generating a repair task. The repair task includes, but is not limited to, key information such as the associated product repair order, user information, a description of the product fault symptoms, suggested repair steps, and required spare parts.

[0086] Step S140: If a maintenance task is generated, the maintenance task is assigned to maintenance personnel.

[0087] Specifically, when a repair task is generated, the product repair reporting system assigns it to repair personnel based on its priority to resolve the user's product repair request. The priority of a repair task is typically assessed based on multiple dimensions, including the urgency of the product malfunction, its impact on the user, and the time and resources required for repair. For example, for critical product malfunctions affecting daily use, the product repair reporting system will mark them as high-priority tasks to ensure prompt repair.

[0088] It should be understood that when repair personnel provide product repair services to users through on-site service or receiving products for repair by express delivery, the product repair reporting system can track the progress of the repair task in real time, establish an instant feedback mechanism with users, and provide real-time feedback on the repair progress to users, so that users can easily grasp the entire product repair process without worrying about information delays or communication problems. This effectively enhances users' transparency and control over the repair process and improves users' satisfaction with the product repair reporting service.

[0089] This embodiment provides a product repair reporting method, which generates a product repair order based on product repair information; reviews the product repair order based on a pre-trained repair review model to obtain a review result, wherein the pre-trained repair review model is obtained by pre-training a preset neural network model based on the backpropagation algorithm; determines whether to generate a repair task based on the review result; if a repair task is generated, assigns the repair task to repair personnel.

[0090] This application, through the aforementioned scheme, first generates a product repair request form based on product repair information, and then automatically reviews the product repair request form using a pre-trained repair review model. Based on the review results and the product repair request form, it determines whether a repair task needs to be generated and assigns it to repair personnel for repair. Through this method, the automatic review of product repair requests is achieved, effectively solving the problem of timeliness in repair request review, significantly improving the review speed of product repair requests, shortening user waiting time, and reducing reliance on manual reviewers, thereby lowering the company's labor costs. Furthermore, using an updatable repair review model for repair request review can effectively improve the accuracy of product repair review and processing.

[0091] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 In order to use the repair review model to review product repair requests, the repair review model needs to be trained and adjusted in advance. Therefore, before step S120, steps S210 to S240 are also included:

[0092] Step S210: Based on historical product repair data, establish a training dataset and a validation dataset;

[0093] Specifically, the product repair system first needs to collect historical product repair data from the company's after-sales service system, including historical product repair information and corresponding review results. Then, using preset keyword recognition, natural language processing technology, or manual annotation methods, the historical product repair data is annotated with product damage information and review information, thereby generating a product repair dataset with data annotations.

[0094] Then, in order to train and adjust the repair review model, the product repair system needs to divide the labeled product repair dataset into a training dataset and a validation dataset.

[0095] It should be understood that in order for the repair review model to have the ability to review and process repair requests for a variety of products, the product repair system needs to collect historical product repair data in various repair scenarios, covering repair data for household appliances, electronic devices, and other mechanical equipment.

[0096] In one implementable manner, the historical product repair data includes several first product repair data, and step S210 includes steps C01 to C04:

[0097] Step C01: Using a preset annotation tool, annotate each first product repair data multiple times to obtain multiple data annotation results;

[0098] Understandably, to ensure the accuracy of data annotation and to establish a high-quality training dataset that provides the correct learning signal for the repair request review model, each product's repair request data may be annotated multiple times. It should be understood that this application does not specify a particular number of annotations; this number can be set by relevant personnel based on actual needs and standard accuracy requirements.

[0099] Specifically, the historical product repair data collected by the product repair system needs to include several first-level product repair data, which include image data, text information, sound data, etc.

[0100] The product repair reporting system utilizes preset annotation tools to annotate each first product repair report multiple times within historical product repair data, thereby obtaining multiple annotation results for each first product repair report. These preset annotation tools are data annotation tools pre-selected or developed by relevant personnel and can annotate image data, text information, and audio data.

[0101] For example, in this embodiment, the preset annotation tools may include image annotation tools such as VGG Image Annotator or Labelbox, text annotation tools such as Docano or Tagtog, and multi-purpose data annotation tools such as Label Studio. The preset annotation tools can also be data annotation tools developed by developers based on preset keyword recognition technology, natural language processing technology, and image recognition algorithms.

[0102] In this embodiment, a preset annotation tool can be used to annotate the image data in the first product repair report data with product damage information. For example, when the image data covers different types of products, the preset annotation tool needs to be used to identify which type of product it is: household appliance (e.g., refrigerator, washing machine, microwave oven), electronic device (e.g., mobile phone, computer, headphones, etc.), or mechanical equipment, and then annotate the type of product to be repaired in the image data. Then, the preset annotation tool also needs to be used to classify and annotate the type of product damage in the image data, as well as the severity of the product damage. For example, the type of product damage can be classified and annotated using words such as "scratches," "cracks," "broken," and "deformation," and the severity of the product damage can be qualitatively described using words such as "minor," "moderate," and "severe," or quantitatively described using a rating scale of 1 to 10. Then, the image data is annotated using the above qualitative or quantitative descriptions of the severity of product damage.

[0103] In addition, in this embodiment, a preset annotation tool can be used to perform data annotation processing on the text information in the first product repair data to obtain review information. For example, the review results in the product repair data can be used to extract text information to obtain the corresponding review information, such as the review information of text descriptions like "repair required" or "no repair required".

[0104] Step C02: When the multiple data annotation results are the same, establish a training dataset and a validation dataset based on the data annotation results and the first product repair data.

[0105] Step C03: When the multiple data annotation results are different, the first product repair data is fed back to the human operator for manual re-annotation, and the manual re-annotation result is obtained.

[0106] Step C04: Based on the manual re-annotation results and the first product repair report data, establish a training dataset and a validation dataset.

[0107] Specifically, when multiple data annotation results corresponding to the first product repair report are identical—for example, if the damaged part is the same in multiple data annotation results—then the data annotation results are determined to be identical. The product repair system can then generate a product repair dataset with data annotations based on the data annotation results and the first product repair report data. This dataset is then divided into a training dataset and a validation dataset according to a preset ratio. The preset ratio is either pre-set by relevant personnel based on actual model training needs or set by relevant personnel based on experience.

[0108] When there are discrepancies among multiple data annotation results corresponding to the first product repair report data, such as different annotations of the review information in multiple data annotation results, it is determined that the data annotation results are different. The first product repair report data needs to be fed back to the relevant personnel so that they can manually re-annotate the first product repair report data and obtain the manually re-annotated data annotation results.

[0109] Then, based on the manually re-annotated data and the first product repair report data, a product repair report dataset with data annotations is generated. This dataset is then divided into a training dataset and a validation dataset according to a preset ratio. This process generates the product repair report dataset with data annotations.

[0110] In this embodiment, each first product repair data in the historical product repair data is annotated multiple times. By comparing the results of multiple annotations, the accuracy of the data annotation is improved, so as to establish a high-quality training dataset and validation dataset. This is beneficial to the subsequent model training of the repair review model and improves the accuracy of product repair review processing.

[0111] Step S220: Extract effective features from the training dataset to obtain effective feature data;

[0112] Understandably, in order to improve the training speed and learning ability of the repair review model, the product repair system can effectively extract features from the training dataset, such as text features (e.g., keywords and phrases in the fault description), image features (e.g., image edges, textures, shapes, colors), and other possible features (e.g., product model, purchase date), so that the repair review model can use the deep features of the training dataset for training and model adjustment.

[0113] For example, when a user submits a product repair request regarding a leaking washing machine, the information may include the following table:

[0114] Text description The washing machine is leaking water from the bottom, with water seeping out from the door gap. Image data Clear photos including the leaking part of the washing machine. Video materials A short video showing the leak.

[0115] For the text descriptions in the table above, natural language processing techniques can be used for feature extraction to obtain the keywords "leakage" and "door gap"; for the image data in the table above, image processing algorithms or deep learning models can be used for feature extraction to extract the edge and texture features of the damaged parts of the washing machine; for the video data in the table above, machine learning or deep learning models can be used to identify dynamic leakage patterns.

[0116] In addition, the product repair reporting system needs to encode the washing machine damage type "water leakage" as a numerical label, and also needs to rate the degree of damage of the washing machine as "moderate" according to the product repair standards, and convert it into a quantitative score, so that the above feature data can be used to train the repair review model later.

[0117] Step S230: Based on the effective feature data and the backpropagation algorithm, train the preset neural network model to adjust the model parameters and obtain the adjusted repair review model.

[0118] The product repair reporting system trains the product repair review model's review capabilities based on the divided training dataset. This allows the model to learn the review patterns and features of product repair information that meet preset product repair standards through effective feature data in the training dataset.

[0119] Specifically, a pre-set neural network model is trained using effective feature data from the training dataset, and the parameters of the neural network model, such as the weight and bias parameters of neurons, are adjusted using the backpropagation algorithm to minimize the difference between the model's review results and the actual review results. The pre-set neural network model refers to a suitable neural network model pre-configured by relevant personnel based on actual product repair review needs, used to review product repair information submitted by users. In this embodiment, the pre-set neural network model may include multiple models, such as convolutional neural networks for image data recognition and analysis, and recurrent neural networks (RNNs) or transformers for processing sequence data related to text information.

[0120] In this embodiment, the product repair system first needs to randomly initialize the weight and bias parameters of a preset neural network model before model training begins, providing a starting point for the model learning process. Then, the backpropagation algorithm is used to calculate the loss value through forward propagation. Valid feature data from the training dataset is used as input data, which enters the initialized preset neural network model and is processed layer by layer through weighted summation and activation functions until the output layer, yielding the model's review result. Then, a preset loss function (e.g., mean squared error, cross-entropy) is used to calculate the difference between the model's review result and the actual review result, obtaining the loss value. Next, the gradient of the loss function with respect to the neural network model parameters is propagated backward from the output layer to the input layer, using the chain rule to calculate the gradient at each layer. Finally, the gradient descent algorithm is used to update the neural network model parameters, referring to the following formula:

[0121]

[0122] Where ω′ is the updated neural network model parameter value, which can be a weight parameter or a bias parameter; ω is the current neural network model parameter value; η is the learning rate, which is a hyperparameter used to control the step size of parameter updates in each neural network iteration. The learning rate can be data set by relevant personnel according to actual needs or experience, and is not specifically limited here.

[0123] It is the gradient of the loss function with respect to the parameters of the neural network model.

[0124] Based on the gradient and learning rate calculated using the method described above, update the weights and bias parameters in the neural network model to reduce loss. Repeat the above steps to iterate the model training process multiple times until the value of the loss function converges to a preset threshold or reaches a preset number of iterations.

[0125] It should be understood that the preset threshold can be data set by relevant personnel based on actual needs or experience, and no specific limitations are made here. During the training process, it is also necessary to adjust model hyperparameters, such as learning rate and batch size, to optimize learning efficiency and model performance.

[0126] For example, when the training dataset contains an image and description of a damaged smartphone, the repair request review model will use features extracted from the image (such as the texture and shape of the damaged area) and features from the text description (such as "cracked screen") to predict whether the repair request is reasonable, thus confirming whether the review is approved. Through the backpropagation algorithm, the repair request review model learns how to better identify smartphone damage from these extracted features, minimizing prediction errors and thus improving the accuracy of the review.

[0127] Step S240: Using the verification dataset, the performance of the adjusted repair request review model is evaluated to adjust the model structure and model parameters, thereby obtaining a pre-trained repair request review model.

[0128] To further ensure the generalization ability of the pre-trained repair review model, the product repair system can use a validation dataset to evaluate the generalization ability of the repair review model on unseen data. This allows for adjustments to the model structure and parameters, such as weight parameters and bias parameters, to prevent overfitting and enable the repair review model to better adapt to new data.

[0129] Specifically, product repair reporting systems can improve the generalization ability of the repair review model by adjusting the model structure, using regularization techniques, and improving training strategies.

[0130] In this embodiment, regularization techniques are used to prevent model overfitting and improve the generalization ability of the repair review model, such as L1 regularization (Lasso Regression) and L2 regularization (Ridge Regression).

[0131] L1 regularization refers to adding the sum of the absolute values ​​of the weight parameters to the loss function, causing the repair request review model to learn as sparse a set of weight parameters as possible, i.e., having more weight parameters that are zero. The formula for L1 regularization is as follows:

[0132] L L1 =L+λ∑w i

[0133] Among them, L L1 λ is the regularized loss function, which is the original loss function plus the L1 regularization term; L is the loss function, a function that measures the difference between the model's review results and the actual review results; λ is the regularization coefficient, used to control the strength of the regularization term; w i These are model parameters, representing the weight parameters in the neural network.

[0134] L2 regularization adds a sum of squared weights to the loss function to minimize the weight values ​​without making them sparse. The formula for L2 regularization is as follows:

[0135]

[0136] Among them, L L2 λ is the regularized loss function, which is the original loss function plus the L2 regularization term; L is the loss function, a function that measures the difference between the model's review results and the actual review results; λ is the regularization coefficient, used to control the strength of the regularization term; w i These are model parameters, representing the weight parameters in the neural network.

[0137] This embodiment provides a product repair reporting method. It involves establishing a training dataset and a validation dataset based on historical product repair data; extracting effective features from the training dataset to obtain effective feature data; training a preset neural network model using the effective feature data and a backpropagation algorithm to adjust model parameters and obtain an adjusted repair review model; evaluating the performance of the adjusted repair review model using the validation dataset to adjust the model structure and parameters to obtain a pre-trained repair review model; generating a product repair order based on product repair information; reviewing the product repair order based on the pre-trained repair review model to obtain a review result, wherein the pre-trained repair review model is obtained by pre-training a preset neural network model using a backpropagation algorithm; determining whether to generate a repair task based on the review result; and assigning the repair task to repair personnel if a repair task is generated.

[0138] This embodiment uses the above-described scheme to construct a training dataset and a validation dataset based on historical product repair data. Then, it extracts effective feature data from the training dataset, trains a preset neural network model using the backpropagation algorithm, evaluates the model performance using the validation dataset, and continues to adjust the model parameters and structure to obtain a pre-trained repair review model. This model can then be used for automated review of product repair information, effectively improving the speed and accuracy of product repair review.

[0139] Based on the second embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the second embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 After step S140 above, steps S310 to S340 may also be included:

[0140] Step S310: Collect product repair information within a preset period and construct a second training dataset and a second verification dataset;

[0141] To improve the performance of the repair request review model in practical applications, the product repair system needs to continuously collect new product repair request information and conduct online learning or periodic updates to the pre-trained model. Specifically, the product repair system needs to collect product repair request information within a preset period and, following the aforementioned methods for constructing training and validation datasets, use the product repair request information collected within the preset period to construct a second training dataset and a second validation dataset. The preset period refers to data set by relevant personnel based on actual model update needs or experience, and is not specifically limited here.

[0142] It should be understood that the product repair reporting system can also use new product repair information as a second training dataset after receiving it, so that the repair review model can immediately learn online.

[0143] Step S320: Update the first repair review model according to the preset incremental learning algorithm and the second training dataset to obtain the second repair review model, wherein the first repair review model is a pre-trained repair review model.

[0144] It should be noted that the preset incremental learning algorithm refers to the incremental learning algorithm that relevant personnel set up themselves based on the actual model update needs. Incremental learning refers to the ability of a neural network to expand by continuously adding new training samples. Incremental learning algorithms can utilize existing training results, thus avoiding retraining the model from scratch. Common incremental learning algorithms include self-organizing incremental learning neural networks and RBF (Radial Basis Function) models based on resource allocation networks.

[0145] Then, the product repair system uses a preset incremental learning algorithm and a second training dataset to perform secondary model training on the first repair review model, namely the aforementioned pre-trained repair review model, to obtain the incrementally learned repair review model, namely the second repair review model.

[0146] Specifically, in this embodiment, the online gradient descent method is used to incrementally learn the first repair request review model using the second training dataset. The model parameters are updated according to the following formula:

[0147]

[0148] Among them, W t The model parameters at time t, where η is the learning rate, and x... t+1 These are the features of the new data points, i.e., the features of the second training dataset; y t+1 These are the labels of the new data points, i.e., the labels of the second training dataset.

[0149] Step S330: Using the second verification dataset, test and evaluate the first repair request review model and the second repair request review model to obtain the test and evaluation results;

[0150] Specifically, the product repair reporting system uses the second verification dataset to conduct A / B group testing on the first and second repair reporting review models. That is, the second verification dataset is randomly divided into two datasets, A and B, to verify the performance of the first repair reporting review model on the dataset A, and at the same time to verify the performance of the second repair reporting review model on the dataset B. By comparing the performance of the first and second repair reporting review models, the test evaluation results can be obtained.

[0151] For example, product repair requests submitted by users in Group A are processed by the second repair request review model. Product repair requests submitted by users in Group B are processed by the first repair request review model. The A / B group test results show that the second repair request review model performs better in accurately identifying the extent of product damage, eliminating the need for further manual review. The first repair request review model can only determine that an item is damaged, but cannot accurately identify the degree of damage.

[0152] Step S340: Based on the test evaluation results, select a model from the first repair review model and the second repair review model to review the repair request.

[0153] When the test evaluation results show that the first repair review model performs better than the second repair review model, the first repair review model can be used for subsequent product repair review processing.

[0154] When the test evaluation results show that the second repair review model performs better than the first repair review model, the second repair review model can be used for subsequent product repair review processing.

[0155] This embodiment provides a product repair reporting method. By collecting product repair information within a preset period, a second training dataset and a second verification dataset are constructed. Then, the second training dataset and an incremental learning algorithm are used to enable the repair review model to learn and update the new data online. The performance of the repair review model after incremental learning is monitored using the second verification dataset, thereby further improving the accuracy of product repair review and providing users with a better repair service experience.

[0156] Based on the second and / or third embodiments of this application, in the fourth embodiment of this application, the content that is the same as or similar to the above-described embodiments two and / or three can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 Following step S140, steps S410 to S430 are also included:

[0157] Step S410: After the maintenance task is completed, collect user feedback information;

[0158] Step S420: Collect the product repair information and corresponding user feedback information to generate a third training dataset;

[0159] Specifically, after the repair task assigned to the repair personnel is completed, the product repair reporting system collects corresponding user feedback information in the form of a questionnaire, including information such as satisfaction with the repair service, satisfaction with the review results, and problems existing in the repair service.

[0160] Collect product repair reports that have already been repaired, along with corresponding user feedback. Use the user feedback as a label and integrate it with the original product repair reports to form a third training dataset.

[0161] Step S430: Based on the third training dataset and preset index parameters, the performance of the pre-trained repair review model is evaluated, and the model is optimized based on the evaluation results to obtain the pre-optimized repair review model.

[0162] Then, based on the negative product repair feedback from users in the third training dataset, the reasons and patterns of errors predicted and reviewed by the repair review model are identified. Next, the pre-trained repair review model is evaluated using preset index parameters, and the model is optimized based on the evaluation results to obtain a pre-optimized repair review model.

[0163] The preset indicator parameters include accuracy, precision, recall, and F1 score, which are used to evaluate the performance of the repair request review model. The indicator parameters are calculated according to the following formula:

[0164]

[0165] In this dataset, TP represents the number of truly positive samples predicted in the third training dataset, TN represents the number of truly negative samples predicted in the third training dataset, FP represents the number of incorrectly predicted positive samples predicted in the third training dataset, and FN represents the number of incorrectly predicted negative samples predicted in the third training dataset. Accuracy is the accuracy of the repair request review model, Precision is the precision of the repair request review model, Recall is the recall of the repair request review model, and F1 is the F1 score, which is the harmonic mean of precision and recall, used to comprehensively consider both precision and recall. A truly positive prediction means that the model's review result matches the actual review result, and the review result shows that the product repair information meets the preset product repair standards, and the product repair request is approved. An incorrectly predicted positive sample means that the model's review result does not match the actual review result, and the review result shows that the product repair information meets the preset product repair standards, and the product repair request is approved.

[0166] For example, the third training dataset contains a product repair request with negative user feedback, where the user is dissatisfied with the repair review result and reports: "The minor blockage in the washing machine was not identified as requiring repair." This negative product repair request will be marked as a "missed" case. By analyzing similar negative product repair cases in the third training dataset, the repair review model may need to be optimized in recognizing minor damage, such as by enhancing the sensitivity of the image recognition algorithm or readjusting the model's decision threshold. After adjusting the sensitivity or decision threshold, the model's performance is re-evaluated using preset index parameters, and further optimization is performed based on the evaluation results to obtain a pre-optimized repair review model.

[0167] This embodiment provides a product repair reporting method. By collecting user feedback information after the repair task is completed and combining it with the corresponding product repair information, a third training dataset is generated. Then, based on the third training dataset and preset index parameters, the pre-trained repair review model is optimized to obtain a pre-optimized repair review model, which further improves the accuracy of product repair review and provides users with a better repair service experience.

[0168] This application provides a product repair reporting device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the product repair reporting method in the above embodiment.

[0169] The following is for reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing product repair devices according to embodiments of this application. The product repair devices in these embodiments may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The product repair equipment shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0170] like Figure 5As shown, the product repair device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the product repair device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the product repair device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows product repair devices with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0171] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0172] The product repair reporting device provided in this application, employing the product repair reporting method described in the above embodiments, can solve the technical problem of low accuracy and efficiency in the review and processing of product repair reports using existing methods. Compared with the prior art, the beneficial effects of the product repair reporting device provided in this application are the same as those of the product repair reporting method provided in the above embodiments, and other technical features of this product repair reporting device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0173] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0174] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0175] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the product repair method described in the above embodiments.

[0176] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0177] The aforementioned computer-readable storage medium may be included in the product repair equipment; or it may exist independently and not assembled into the product repair equipment.

[0178] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the product repair device, the product repair device causes the device to: generate a product repair order based on the product repair information; review the product repair order based on a pre-trained repair review model to obtain a review result, wherein the pre-trained repair review model is obtained by pre-training a preset neural network model based on a backpropagation algorithm; determine whether to generate a repair task based on the review result; and if a repair task is generated, assign the repair task to a repair person.

[0179] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0180] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0181] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0182] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described product repair reporting method. This solves the technical problem of low accuracy and efficiency in the product repair review and processing of existing product repair methods. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the product repair reporting method provided in the above embodiments, and will not be repeated here.

[0183] This application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the product repair method described above.

[0184] The computer program product provided in this application can solve the technical problem of low accuracy and efficiency in the product repair review and processing of existing product repair methods. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the product repair methods provided in the above embodiments, and will not be repeated here.

[0185] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent scope of this application.

Claims

1. A product repair reporting method, characterized in that, The method includes: Generate a product repair order based on the product repair information; The product repair request is reviewed based on a pre-trained repair review model to obtain the review result. The pre-trained repair review model is obtained by pre-training a preset neural network model based on the backpropagation algorithm. Based on the audit results, confirm whether to generate a maintenance task; If a maintenance task is generated, the maintenance task is assigned to maintenance personnel.

2. The method as described in claim 1, characterized in that, Before the step of reviewing the product repair request based on the pre-trained repair review model and obtaining the review result, the method further includes: Based on historical product repair data, establish training and validation datasets; Effective feature extraction is performed on the training dataset to obtain effective feature data; Based on the effective feature data and the backpropagation algorithm, the preset neural network model is trained to adjust the model parameters and obtain the adjusted repair review model. Using the validation dataset, the performance of the adjusted repair request review model is evaluated to adjust the model structure and parameters, thereby obtaining a pre-trained repair request review model.

3. The method as described in claim 2, characterized in that, The historical product repair data includes several first product repair data sets. The step of establishing a training dataset and a validation dataset based on the historical product repair data includes: Using a preset annotation tool, each first product repair data was annotated multiple times to obtain multiple data annotation results; When the multiple data annotation results are the same, a training dataset and a validation dataset are established based on the data annotation results and the first product repair data. When the multiple data annotation results are different, the first product repair data is fed back to human for manual re-annotation, and the manual re-annotation result is obtained. Based on the manual re-annotation results and the first product repair report data, a training dataset and a validation dataset are established.

4. The method as described in claim 1, characterized in that, The step of generating a product repair order based on product repair information includes: The product repair information is preprocessed to obtain preprocessed product repair information; The image data in the pre-processed product repair information is used to perform image recognition and analysis using a pre-trained repair review model to obtain product damage information; A product repair order is generated based on the product damage information and the product repair request information.

5. The method as described in claim 2, characterized in that, The step of confirming whether to generate a maintenance task based on the audit results includes: When the review result is that the product repair request does not meet the preset product repair standards, the product repair request is marked using the review result and the product repair request is fed back to the user. When the review result shows that the product repair request meets the preset product repair standards, a repair task is generated based on the product repair request and repair resources.

6. The method as described in claim 2, characterized in that, Following the step of assigning a maintenance task to maintenance personnel if a maintenance task is generated, the method includes: Collect product repair information within a preset period to construct a second training dataset and a second validation dataset; The first repair review model is updated based on the preset incremental learning algorithm and the second training dataset to obtain the second repair review model. The first repair review model is a pre-trained repair review model. Using the second verification dataset, the first repair request review model and the second repair request review model are tested and evaluated to obtain the test evaluation results. Based on the test evaluation results, a model is selected from the first repair review model and the second repair review model for repair review.

7. The method as described in claim 2, characterized in that, Following the step of assigning a maintenance task to maintenance personnel if a maintenance task is generated, the method includes: Once the maintenance task is completed, collect user feedback information; Collect the product repair information and corresponding user feedback information to generate a third training dataset; Based on the third training dataset and preset index parameters, the performance of the pre-trained repair review model is evaluated, and the model is optimized based on the evaluation results to obtain a pre-optimized repair review model.

8. A product repair reporting device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the product repair method as described in any one of claims 1 to 7.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the product repair method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the product repair method as described in any one of claims 1 to 7.

Images