Product quality evaluation method and device based on artificial intelligence attention mechanism

Abstract

The invention discloses a product quality evaluation method and device based on an artificial intelligence attention mechanism, and belongs to the technical field of artificial intelligence. The method comprises the following steps: constructing a quality evaluation index system of a product; obtaining a plurality of training samples based on the quality evaluation index system; the training sample comprises input information which is acquired for a sample product and corresponds to each evaluation index, and scores for the sample product; the input information corresponding to each evaluation index in the training samples is used as input, the score is used as output, a neural network model based on an attention mechanism is trained through the multiple training samples, the neural network model automatically recognizes more concerned evaluation indexes, and attention weights of different evaluation indexes are adjusted; and performing quality evaluation on a to-be-evaluated product by using the trained neural network model. The attention weights of different evaluation indexes can be automatically adjusted, and objectivity and consistency in the evaluation process of the to-be-evaluated product are achieved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly to a product quality evaluation method and device based on an artificial intelligence attention mechanism. Background Art

[0002] In the fields of medical treatment, rehabilitation, and elderly care, the application of intelligent devices is in a rapid development stage. From wearable health monitoring devices to large-scale rehabilitation training equipment and then to intelligent elderly care service robots, these devices play a crucial role in improving medical efficiency, enhancing rehabilitation effects, and elevating the quality of life for the elderly. However, the quality of these devices is directly related to the health and safety of users. Therefore, it is of great importance to conduct scientific and comprehensive quality evaluations on these devices.

[0003] Traditional product quality evaluation methods mainly rely on manual experience and simple detection means, and mainly test the functionality of products to determine whether the product functions meet the established requirements, and calculate the evaluation results based on the assigned weights to achieve product quality evaluation.

[0004] However, weight assignment is greatly affected by subjective factors, making it difficult to ensure the objectivity and consistency of evaluation results. Therefore, a more objective and accurate evaluation method is needed. Summary of the Invention

[0005] The present invention provides a product quality evaluation method and device based on an artificial intelligence attention mechanism. The technical solutions are as follows: On the one hand, a product quality evaluation method based on an artificial intelligence attention mechanism is provided. The method includes: Construct a quality evaluation index system for the product; the product has a specified function; the quality evaluation index system includes multiple evaluation indexes; Based on the quality evaluation index system, obtain multiple training samples; the training samples include the input information corresponding to each evaluation index obtained for the sample product, and the score given to the sample product; the multiple training samples are obtained from at least multiple different types of sample products, and the multiple different types of sample products all have the specified function; the score is calculated from a first score and at least one second score; the first score is the score calculated according to the set scoring rule using the input information; and the evaluation indexes in the quality evaluation index system are performance indexes that can be physically measured; the second score is the score obtained by evaluating using non-performance indexes; Taking the input information corresponding to each evaluation index in the training samples as the input and the score as the output, so as to train the neural network model based on the attention mechanism using the multiple training samples, so that the neural network model can automatically identify the more concerned evaluation index and adjust the attention weights of different evaluation indexes; Using the trained neural network model to evaluate the quality of the product to be evaluated.

[0006] In a possible implementation manner, the sample product is at least one of a medical device, a rehabilitation device, and an elderly care device; and the specified function is at least one of the functions possessed by the medical device, the rehabilitation device, and the elderly care device.

[0007] In a possible implementation manner, the quality evaluation index system is constructed based on the analytic hierarchy process, and the quality evaluation index system includes evaluation indexes at least two levels.

[0008] In a possible implementation manner, the using the trained neural network model to evaluate the quality of the product to be evaluated includes: Based on the quality evaluation index system, obtaining the input information corresponding to each evaluation index for the product to be evaluated; Inputting the input information corresponding to each evaluation index obtained for the product to be evaluated into the trained neural network model to obtain the output score, and using this score as the quality evaluation result of the product to be evaluated.

[0009] In a possible implementation manner, it further includes: using the test samples and the trained neural network model to determine the magnitude relationship of the weights assigned by the neural network model to multiple evaluation indexes in the quality evaluation index system; wherein, the test samples include the input information corresponding to each evaluation index obtained for the test product; After using the trained neural network model to evaluate the quality of the product to be evaluated, it further includes: if the score output by the neural network model for the product to be evaluated is less than the set score, then based on the magnitude relationship of the weights assigned to multiple evaluation indexes in the quality evaluation index system and the input information of each evaluation index for the product to be evaluated, determining at least one target evaluation index affecting the score, and outputting improvement suggestions for the product to be evaluated according to the at least one target evaluation index.

[0010] On the other hand, a product quality evaluation device based on the artificial intelligence attention mechanism is provided, and the device includes: A construction unit for constructing a quality evaluation index system for the product; the product has a specified function; the quality evaluation index system includes multiple evaluation indexes; An acquisition unit, configured to acquire a plurality of training samples based on the quality evaluation index system; the training samples include input information corresponding to each evaluation index obtained for a sample product, and a score given to the sample product; the plurality of training samples are obtained from at least a plurality of different types of sample products, and the plurality of different types of sample products all have the specified function; the score is calculated from a first score and at least one second score; the first score is a score calculated according to a set scoring rule using the input information; and the evaluation indexes in the quality evaluation index system are performance indexes that can be physically measured; the second score is a score obtained by evaluating using non-performance indexes. A training unit, configured to use the input information corresponding to each evaluation index in the training samples as inputs and the scores as outputs, so as to train a neural network model based on an attention mechanism using the plurality of training samples, so that the neural network model can automatically identify the evaluation indexes that it pays more attention to and adjust the attention weights of different evaluation indexes. An evaluation unit, configured to perform quality evaluation on a product to be evaluated using the trained neural network model.

[0011] In a possible implementation manner, it further includes: A first determination unit, configured to use a test sample and the trained neural network model to determine the magnitude relationship of the weights assigned by the neural network model to a plurality of evaluation indexes in the quality evaluation index system; wherein, the test sample includes input information corresponding to each evaluation index obtained for a test product. A second determination unit, configured to, after the evaluation unit finishes execution, if the score output by the neural network model for the product to be evaluated is less than a set score, determine at least one target evaluation index affecting the score based on the magnitude relationship of the weights assigned by the neural network model to a plurality of evaluation indexes in the quality evaluation index system and the input information of each evaluation index for the product to be evaluated, and output an improvement suggestion for the product to be evaluated according to the at least one target evaluation index.

[0012] On the other hand, a computer device is provided, the computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored on the memory to implement the steps of the above-mentioned product quality evaluation method based on an artificial intelligence attention mechanism.

[0013] On the other hand, a computer-readable storage medium is provided, the storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of the above-mentioned product quality evaluation method based on an artificial intelligence attention mechanism.

[0014] On the other hand, a computer program product is provided, including a computer program which, when executed by a processor, implements the steps of the product quality evaluation method based on the artificial intelligence attention mechanism described above.

[0015] The technical solution provided by the present invention can at least bring the following beneficial effects: By constructing a quality evaluation index system for products, the training sample products and the products to be evaluated used for training the neural network model are all evaluated according to the same quality evaluation index system, and the input information of each evaluation index in the quality evaluation index system can be obtained through physical evaluation, thus ensuring the objectivity and consistency of the data source. The multiple training samples are at least obtained from multiple different types of sample products to ensure that the trained neural network model can perform quality evaluation on products with specified functions and different types. For the scores used as outputs in the training samples, they are jointly obtained by a first score and at least one second score. The first score is a score calculated according to a set scoring rule using the input information, while the second score is a score obtained by evaluating using non-performance indicators, that is, the second score is not calculated through the input information but is scored by external factors. When using the neural network model to evaluate new products, it can achieve the scores of joint evaluation of performance indicators and non-performance indicators only based on the input information of performance indicators. In this way, during the process of training the neural network model, the neural network model can map based on the input information and the output scores, automatically identify the more concerned evaluation indicators, and thus can automatically adjust the attention weights of different evaluation indicators, and further achieve objectivity and consistency in the process of evaluating the products to be evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 is a flowchart of a product quality evaluation method based on the artificial intelligence attention mechanism provided by an embodiment of the present invention; Figure 2 is a structural diagram of a product quality evaluation device based on the artificial intelligence attention mechanism provided by an embodiment of the present invention; Figure 3 is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0019] Please refer to Figure 1 , a product quality evaluation method based on an artificial intelligence attention mechanism provided by an embodiment of the present invention, the method includes: Step 100, constructing a quality evaluation index system for the product; the product has a specified function; the quality evaluation index system includes multiple evaluation indexes; Step 102, based on the quality evaluation index system, obtaining multiple training samples; the training samples include input information corresponding to each evaluation index obtained for a sample product, and a score for the sample product; the multiple training samples are obtained from at least multiple different types of sample products, and the multiple different types of sample products all have the specified function; the score is calculated from a first score and at least one second score; the first score is a score calculated according to a set scoring rule using the input information; and the evaluation indexes in the quality evaluation index system are performance indexes that can be physically measured; the second score is a score obtained by evaluating using non-performance indexes; Step 104, using the input information corresponding to each evaluation index in the training samples as input and the score as output to train a neural network model based on the attention mechanism using the multiple training samples, so that the neural network model automatically identifies the more concerned evaluation indexes and adjusts the attention weights of different evaluation indexes; Step 106, using the trained neural network model to evaluate the quality of the product to be evaluated.

[0020] In an embodiment of the present invention, by constructing a quality evaluation index system for products, the training sample products used to train the neural network model and the products to be evaluated are all evaluated according to the same quality evaluation index system, and the input information of each evaluation index in the quality evaluation index system can be obtained through physical evaluation, so as to ensure the objectivity and consistency of the data source. The multiple training samples are obtained from at least multiple different types of sample products to ensure that the trained neural network model can perform quality evaluation on products with specified functions and different types. For the scores used as outputs in the training samples, they are jointly obtained by a first score and at least one second score. The first score is a score calculated according to a set scoring rule using the input information, while the second score is a score obtained by evaluating using non-performance indicators. That is to say, the second score is not calculated through the input information, but is scored by external factors. When using the neural network model to evaluate the quality of new products, it can realize the scores of joint evaluation of performance indicators and non-performance indicators only based on the input information of performance indicators. In this way, during the process of training the neural network model, the neural network model can map based on the input information and the output scores, automatically identify the evaluation indicators that it pays more attention to, and thus can automatically adjust the attention weights of different evaluation indicators, and further achieve objectivity and consistency in the process of evaluating the products to be evaluated.

[0021] The following describes Figure 1 the execution manners of the respective steps shown.

[0022] First, for step 100, construct a quality evaluation index system for the product; the product has a specified function; the quality evaluation index system includes multiple evaluation indicators.

[0023] In an embodiment of the present invention, in order to ensure consistency in the product evaluation process, a quality evaluation index system for the product can be constructed. Among them, the quality evaluation index system is constructed for products with a specified function, and the quality evaluation index system includes multiple evaluation indicators.

[0024] The product quality evaluation method of the embodiment of the present invention is used to evaluate the product quality, especially for the quality evaluation of medical devices, rehabilitation devices, and elderly care devices. Therefore, the specified function that the product has can be at least one of the functions possessed by medical devices, rehabilitation devices, and elderly care devices. For example, the specified function can be a blood oxygen measurement function, a blood pressure measurement function, a heart rate measurement function, etc.

[0025] When constructing the quality evaluation index system, it can be constructed by using industry indicators, standard indicators, national indicators, industry indicators, etc. of the product.

[0026] In one implementation, the quality evaluation index system may be multiple evaluation indexes with a parallel relationship, that is, the multiple evaluation indexes in the quality evaluation index system are at the same level.

[0027] In another implementation, the quality evaluation index system may be constructed based on the analytic hierarchy process. Then, the quality evaluation index system includes evaluation indexes at least two levels.

[0028] Among them, in the quality evaluation of medical devices, by constructing a hierarchical structure model, the quality evaluation goal of the device is decomposed into criterion layers such as functionality, safety, and usability, and then further subdivided into specific index layers, which facilitates the evaluator to make pairwise comparisons of the factors at each level, determine the relative importance weights, and then comprehensively obtain the evaluation results. However, in the traditional method, the analytic hierarchy process also has obvious disadvantages: on the one hand, the construction of its judgment matrix depends on the subjective judgment of experts. Factors such as the knowledge structure, experience level, and personal preferences of experts will affect the accuracy of the judgment matrix, resulting in strong subjectivity in weight assignment; for example, when determining the relative importance of the functionality and safety of medical devices, different experts may give quite different judgments, making the evaluation results lack sufficient objectivity and reliability; on the other hand, the analytic hierarchy process has limited data processing capabilities and is difficult to handle massive and complex multi-source data; in the case of the explosive growth of the data volume of intelligent devices, a large amount of sensor data, user feedback data, etc. generated during the operation of the device, the analytic hierarchy process cannot fully mine the potential information therein and is difficult to comprehensively and accurately reflect the quality status of the device.

[0029] Based on this, in the embodiments of the present invention, combined with the artificial intelligence attention mechanism, the neural network model is used to automatically identify the weights of the evaluation indexes, which not only reduces the calculation amount but also reduces the participation of artificial subjectivity, and improves the objectivity and consistency of product quality evaluation.

[0030] Taking the quality evaluation index system including evaluation indexes at three levels as an example, that is, the quality evaluation index system includes first-level indexes, second-level indexes, and third-level indexes. In one implementation, the first-level indexes include at least one or more of the following evaluation indexes: appearance structure, aging adaptation level, functionality, safety, signal quality, environmental adaptability, and user experience. Among them, the appearance structure focuses on the overall design and material selection of the device; the aging adaptation level focuses on the degree of adaptation of the device to the usage habits and physical characteristics of the elderly; the functionality is used to consider the implementation and effect of each function of the device; the safety is used to ensure that the device will not cause harm to the user during use; the environmental adaptability is used to evaluate the stable operation ability of the device under different environmental conditions; the user experience is used to reflect the subjective feelings of the user about the device.

[0031] Preferably, the constructed quality evaluation index system includes at least the evaluation index of the aging adaptation level. Among them, the "aging adaptation level", as the name implies, "the level of whether it is suitable for the elderly to use". Using this evaluation index can focus on the adaptation degree of the device to the usage habits and physical characteristics of the elderly, not only paying attention to the safety of the device, but also being able to evaluate the practical value of the product from the actual needs of the elderly, so as to meet the actual needs of the quality assessment of medical devices, rehabilitation devices, and elderly care devices.

[0032] For example, the secondary indicators for "appearance structure" in the above primary indicators may at least include one or more of the surface, display screen, watch band, switch or button, and logo; the secondary indicators for "aging adaptation level" in the above primary indicators may at least include one or more of perceivability, operability, understandability, compatibility, safety, remote assistance, and emergency application; the tertiary indicators for "perceivability" in the secondary indicators may at least include one or more of font size adjustment, line spacing, contrast, color, and verification code; the tertiary indicators for "operability" in the secondary indicators may at least include one or more of screen display magnification, input keyboard, screen touch interaction, voice broadcast, component focus size, gesture operation, operation time, and floating window.

[0033] Then, steps 102 "Based on the quality evaluation index system, obtain multiple training samples" and 104 "Use the input information corresponding to each evaluation index in the training samples as the input and the score as the output to train the neural network model based on the attention mechanism, so that the neural network model automatically identifies the more concerned evaluation index and adjusts the attention weights of different evaluation indexes" are described simultaneously.

[0034] In the embodiment of the present invention, the quality of the product is evaluated by the neural network model outputting a score for the input information. To ensure the objectivity and consistency of the neural network model trained for product quality evaluation, the training samples used to train the neural network model are also obtained based on the quality evaluation index system.

[0035] Among them, the training sample includes the input information corresponding to each evaluation index obtained for the sample product, and the score given to the sample product; and the multiple training samples are at least obtained from multiple different types of sample products, and the multiple different types of sample products all have the specified function.

[0036] In an embodiment of the present invention, the score given to the sample product in the training sample is calculated from a first score and at least one second score; where: The first score is calculated according to a set scoring rule using the input information; and the evaluation indicators in the quality evaluation index system are performance indicators that can be physically measured; The second score is obtained by evaluating using non-performance indicators.

[0037] The following separately explains the first score and the second score.

[0038] First, the first score is explained.

[0039] Since the evaluation indicators in the quality evaluation index system are all performance indicators that can be physically measured, therefore, the input information corresponding to the evaluation indicators is all objective information that can be directly measured, thus ensuring objectivity in the evaluation process.

[0040] In the embodiments of the present invention, a scoring rule can be preset, and weights are configured for different evaluation indicators based on expert experience to calculate the first score using the input information.

[0041] Among them, the input information can be collected through channels such as sensors, user feedback, and device operation logs. In terms of sensor data collection, sensors with high precision and strong stability can be selected to ensure the accuracy and reliability of the collected data; in terms of user feedback collection, a combination of online and offline methods can be adopted to expand the feedback collection scope and improve the representativeness of the feedback data; in terms of device operation logs, professional data parsing tools can be used to deeply mine the log data and extract valuable information.

[0042] Taking the aging-friendly level and its secondary indicator "perceivability" as an example, the weight configured for the aging-friendly level is 20%, and the scoring method for each tertiary indicator of "perceivability" can include: Font size adjustment: Score this evaluation indicator from the optional scoring range (3 - 5 points) according to the scoring rule of "the interface font size can be set, adjusted according to the system settings, or there is a font size setting option within the mobile application". Line spacing: Score this evaluation indicator from the optional scoring range (3 - 5 points) according to the scoring rule of "the line spacing of the text within the paragraph is at least 1.3 times, and the paragraph spacing is at least 1.3 times larger than the line spacing, while taking into account the applicable scenarios and display effects of the mobile application". Contrast: Score this evaluation indicator from the optional scoring range (1 - 2 points) according to the scoring rule of "the contrast between text / text image presentation methods, icons and other elements is at least 4.5:1". Color: Score this evaluation indicator from the optional scoring range (1 - 2 points) according to the scoring rule of "the text color is not the only means to distinguish visual elements such as conveying information and indicating actions". Verification code: According to the scoring rule of "if there are verification methods that are not easy for the elderly to understand, such as non-text verification codes, in the mobile application, text verification codes that can be understood by the elderly should be provided", score this evaluation index within the optional scoring range (1-2 points).

[0043] In order to further improve the objectivity of the input information, the optional scoring range for each evaluation index can be two scores. When the scoring rule of the evaluation index is not met, the lowest score is directly used as the score of this evaluation index. When the scoring rule of the evaluation index is met, the highest score is directly used as the score of this evaluation index. For example, the optional scoring range for the evaluation index of contrast is 1 point and 2 points. If the contrast meets its scoring rule, the score is 2 points; otherwise, the score is 1 point.

[0044] After each third-level index is scored based on the corresponding input information, in one implementation, the calculation method of the first score can include: adding up the scores of each third-level index to obtain the score of the corresponding second-level index, then adding up the scores of each second-level index to obtain the score of the corresponding first-level index. Using the scores of each first-level index and their weights, the weighted sum of multiple first-level indexes can be obtained, and this weighted sum is used as the first score.

[0045] Then, the second score is explained.

[0046] In order to evaluate the product from multiple dimensions, in addition to performance indicators, non-performance indicators also need to be used for evaluation. This is because in addition to product performance, market feedback is more important for the quality of the product. Therefore, the non-performance indicators can at least include market feedback evaluation indicators to comprehensively evaluate the product using market feedback. In addition, the non-performance indicators can also include expert evaluation indicators, etc.

[0047] For the market feedback evaluation indicators, it can specifically include at least one of market price, market sales volume, and market score. Based on these non-performance indicators, the corresponding second score of the non-performance indicators can be calculated according to the set calculation rules.

[0048] It can be seen that in order to obtain the score as the output in the training sample, it is necessary to combine not only the input information but also the data outside the input information to jointly obtain the score, so as to realize the quality evaluation for the subsequent products without the support of data outside the input information. For example, for a new product before it is put on the market, since there are no market feedback evaluation indicators, the performance indicators that can be physically measured can be used as the input information to output the score of the comprehensive quality evaluation.

[0049] Then, based on the description of the above embodiments, training samples can be obtained for the sample products.

[0050] To ensure that the trained neural network model can evaluate the quality of different types of products, the multiple training samples used to train the neural network model are at least obtained from multiple different types of sample products, and the multiple different types of sample products all have the specified function.

[0051] When the product quality evaluation method provided by the embodiments of the present invention is applied to medical devices, rehabilitation devices or elderly care devices, the sample products are at least one of medical devices, rehabilitation devices and elderly care devices; and the specified function of the sample products is at least one of the functions possessed by medical devices, rehabilitation devices and elderly care devices.

[0052] It should be noted that if the specified function is multiple functions, in one implementation, the sample product can be a product with at least one of the multiple functions. For example, if the specified functions are blood pressure measurement function and blood oxygen measurement function, the sample product can be a product with only the blood pressure measurement function, or a product with only the blood oxygen measurement function.

[0053] Among them, the types of sample products can include smart watches, bracelets, sports headphones, smart phones, finger clip type blood oxygen meters, medical blood oxygen monitors, etc. Thus, the sample products cover more types, so that the trained neural network model can evaluate the quality of different types of products.

[0054] After obtaining multiple training samples, use the input information corresponding to each evaluation index in the training samples as the input and the score as the output to train the neural network model based on the attention mechanism. The neural network model can deeply analyze and extract features from the input information of different evaluation indexes to learn the mapping relationship between the input information of the evaluation indexes and the score, automatically identify the more concerned evaluation indexes, and then automatically adjust the attention weights of different evaluation indexes. That is to say, although when calculating the first score, the weights of each evaluation index are manually set based on expert experience, the weights of the evaluation indexes in the score obtained by combining the first score and the second score change, and this change can better reflect the true weights of each evaluation index, and the adjusted weights are more objective and accurate.

[0055] In addition, when the quality evaluation index system is constructed based on the analytic hierarchy process (AHP), since there are numerous evaluation indexes at all levels involved in the quality evaluation index system, even possibly reaching several hundred, there are complex correlation relationships among evaluation indexes at different levels. And according to the description of the above embodiments, the traditional AHP not only lacks sufficient objectivity and reliability, but also cannot accurately mine which information in the data is more important in the face of a large amount of data, resulting in poor accuracy of the product quality evaluation result. In the embodiments of the present invention, the artificial intelligence attention mechanism is combined, and the neural network model is used to automatically identify the weights of each evaluation index, which not only reduces the calculation amount and reduces the participation of human subjectivity, but also improves the objectivity and consistency of the product quality evaluation.

[0056] During the training process of the neural network model, the backpropagation algorithm can be used to accurately adjust the model parameters. At the same time, various optimization algorithms such as stochastic gradient descent, Adagrad, and Adadelta can be flexibly used to dynamically adjust the learning rate according to the training effect and convergence situation of the model, continuously optimize the model performance, and ensure the accuracy, stability, and generalization ability of the model.

[0057] Finally, for step 106, the trained neural network model is used to evaluate the quality of the product to be evaluated.

[0058] Among them, the product to be evaluated has a specified function.

[0059] Specifically, based on the quality evaluation index system, the input information corresponding to each evaluation index can be obtained for the product to be evaluated, and the input information corresponding to each evaluation index is input into the trained neural network model to obtain the output score, and this score is used as the quality evaluation result of the product to be evaluated.

[0060] It can be seen that even if the product to be evaluated only has the input information of performance indexes, the neural network model can be used to obtain a comprehensive score that also includes non-performance indexes.

[0061] Furthermore, for the trained neural network model, the test sample and the trained neural network model can be used to determine the magnitude relationship of the weights assigned by the neural network model to multiple evaluation indexes in the quality evaluation index system; among them, the test sample includes the input information corresponding to each evaluation index obtained for the test product. Specifically, methods such as the position perturbation method, SHAP (interpretability of machine learning models) method, and gradient change can be used to determine several evaluation indexes with the highest assigned weights.

[0062] The determination process is described by the position perturbation method. The input information corresponding to each evaluation index in the test sample is input into the trained neural network model to obtain the output score; the input information of two evaluation indexes in the test sample is exchanged, and then the exchanged input information is input into the trained neural network model to obtain the score after information exchange; if the difference between the score after information exchange and the score before information exchange exceeds the set threshold, it is determined that at least one of the two evaluation indexes for which information exchange is performed has a higher assigned weight. Based on this, the magnitude relationship of the assigned weights of multiple evaluation indexes in the quality evaluation index system can be determined by the position perturbation method.

[0063] Further, after determining the magnitude relationship of the weights assigned by the neural network model to multiple evaluation indexes in the quality evaluation index system, if the score output by the neural network model for the product to be evaluated is less than the set score, it indicates that the product still needs to be improved in some performances. Based on the magnitude relationship of the weights assigned to multiple evaluation indexes in the quality evaluation index system and the input information of each evaluation index, at least one target evaluation index affecting the score is determined, and improvement suggestions are output for the product according to the at least one target evaluation index.

[0064] The following uses an actual case to illustrate the embodiments of the present invention.

[0065] Intelligent elderly care mattress case.

[0066] A quality evaluation index system is constructed for the sleep monitoring function, and the construction method is implemented based on the analytic hierarchy process. Among them, the first-level indexes include appearance structure, aging-friendly level, functionality, safety, environmental reliability, and user experience; the second-level indexes are further divided into sleep monitoring function, health warning function, etc. under functionality; the aging-friendly level is further divided into whether the mattress height is convenient for the elderly to get on and off the bed, whether the operation is simple and easy to understand, etc.; the third-level indexes are further divided into the sleep stage recognition accuracy rate, heart rate monitoring error, etc. under the sleep monitoring function.

[0067] Multiple training samples are obtained for different models of intelligent elderly care mattresses. Among them, for the input information of each evaluation index, the acquisition methods can include: collecting data such as the user's sleep posture and turning times through the pressure sensors on the mattress; collecting information such as the comfort and operation convenience of the mattress through user feedback; obtaining data such as the failure records and operation duration of the mattress from the device operation logs; and calculating the first score using the obtained input information, and then calculating the second score for non-performance indexes, and taking the sum of the first score and the second score as the score of the training sample.

[0068] Train a neural network model based on the artificial intelligence attention mechanism, and then for the intelligent elderly care mattress to be evaluated, input its input information into the trained neural network model to obtain a comprehensive score. This comprehensive score corresponds to an excellent quality rating, and improvement suggestions for optimizing the mattress edge design are given.

[0069] Please refer to Figure 2 , an embodiment of the present invention provides a product quality evaluation device based on the artificial intelligence attention mechanism. The device includes: A construction unit 200 for constructing a quality evaluation index system for the product; the product has a specified function; the quality evaluation index system includes multiple evaluation indexes; An acquisition unit 202 for obtaining multiple training samples based on the quality evaluation index system; the training samples include the input information corresponding to each evaluation index obtained for the sample product, and the score given to the sample product; the multiple training samples are obtained from at least multiple different types of sample products, and multiple different types of sample products all have the specified function; the score is calculated from a first score and at least one second score; the first score is the score calculated according to the set scoring rule using the input information; and the evaluation indexes in the quality evaluation index system are physical evaluation performance indexes; the second score is the score obtained by evaluating using non-performance indexes; A training unit 204 for using the input information corresponding to each evaluation index in the training sample as the input and the score as the output to train the neural network model based on the attention mechanism using the multiple training samples, so that the neural network model can automatically identify the more concerned evaluation indexes and adjust the attention weights of different evaluation indexes; An evaluation unit 206 for using the trained neural network model to evaluate the quality of the product to be evaluated.

[0070] In an embodiment of the present invention, the sample product is at least one of a medical device, a rehabilitation device, and an elderly care device; and the specified function is at least one of the functions possessed by a medical device, a rehabilitation device, and an elderly care device.

[0071] In an embodiment of the present invention, the quality evaluation index system is constructed based on the analytic hierarchy process, and the quality evaluation index system includes evaluation indexes at least at two levels.

[0072] In an embodiment of the present invention, the evaluation unit is specifically configured to: based on the quality evaluation index system, obtain the input information corresponding to each evaluation index for the product to be evaluated; input the input information corresponding to each evaluation index obtained for the product to be evaluated into the trained neural network model to obtain the output score, and use this score as the quality evaluation result of the product to be evaluated.

[0073] In an embodiment of the present invention, the device may further include: The first determination unit is configured to use the test sample and the trained neural network model to determine the magnitude relationship of the weights assigned by the neural network model to multiple evaluation indexes in the quality evaluation index system; wherein, the test sample includes the input information corresponding to each evaluation index obtained for the test product. The second determination unit is configured to, when the score output by the neural network model for the product to be evaluated is less than the set score, determine at least one target evaluation index affecting the score based on the magnitude relationship of the weights assigned to multiple evaluation indexes in the quality evaluation index system and the input information of each evaluation index for the product to be evaluated, and output an improvement suggestion for the product to be evaluated according to the at least one target evaluation index.

[0074] It should be noted that: the product quality evaluation device based on the artificial intelligence attention mechanism provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the product quality evaluation device based on the artificial intelligence attention mechanism provided in the above embodiment and the embodiment of the product quality evaluation method based on the artificial intelligence attention mechanism belong to the same concept. For the specific implementation process, please refer to the method embodiment and will not be elaborated here.

[0075] An embodiment of the present application further provides a computer device. Please refer to Figure 3 , this computer device includes a processor and a memory. At least one instruction, at least one program, a code set or an instruction set is stored in the memory, and at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the product quality evaluation method based on the artificial intelligence attention mechanism provided in each of the above method embodiments.

[0076] An embodiment of the present application further provides a computer-readable storage medium. At least one instruction, at least one program, a code set or an instruction set is stored on this computer-readable storage medium, and at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the product quality evaluation method based on the artificial intelligence attention mechanism provided in each of the above method embodiments.

[0077] An embodiment of the present application further provides a computer program product, which includes a computer program. The processor of the computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the product quality evaluation method based on the artificial intelligence attention mechanism described in any one of the above embodiments.

[0078] For the convenience of description, when describing the above system or device, it is divided into various modules or units according to functions for separate description. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0079] From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

[0080] Finally, it should also be noted that in this article, relational terms such as first, second, third, and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0081] The above are only the preferred embodiments of the present application. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A product quality evaluation method based on the artificial intelligence attention mechanism, characterized in that, The method includes: Constructing a quality evaluation index system for a product; the product has a specified function; the quality evaluation index system includes multiple evaluation indexes; Based on the quality evaluation index system, obtaining multiple training samples; the training samples include input information corresponding to each evaluation index obtained for a sample product, and a score given to the sample product; the multiple training samples are obtained from at least multiple different types of sample products, and the multiple different types of sample products all have the specified function; the score is calculated from a first score and at least one second score; the first score is a score calculated using the input information according to a set scoring rule; and the evaluation indexes in the quality evaluation index system are performance indexes that can be physically measured; the second score is a score obtained by evaluating using non-performance indexes; Using the input information corresponding to each evaluation index in the training samples as input and the score as output to train a neural network model based on the attention mechanism using the multiple training samples, so that the neural network model automatically identifies the evaluation indexes it pays more attention to and adjusts the attention weights of different evaluation indexes; Using the trained neural network model to conduct quality evaluation on the product to be evaluated.

2. The method according to claim 1, wherein The sample product is at least one of a medical device, a rehabilitation device, and an elderly care device; and the specified function is at least one of the functions possessed by a medical device, a rehabilitation device, and an elderly care device.

3. The method according to claim 1, wherein The quality evaluation index system is constructed based on the analytic hierarchy process, and the quality evaluation index system includes evaluation indexes at least at two levels.

4. The method according to claim 1, wherein The using the trained neural network model to conduct quality evaluation on the product to be evaluated includes: Based on the quality evaluation index system, obtaining the input information corresponding to each evaluation index for the product to be evaluated; Inputting the input information corresponding to each evaluation index obtained for the product to be evaluated into the trained neural network model to obtain the output score, and using this score as the quality evaluation result of the product to be evaluated.

5. The method according to any one of claims 1-4, characterized in that It further includes: using the test samples and the trained neural network model to determine the magnitude relationship of the weights assigned by the neural network model to multiple evaluation indexes in the quality evaluation index system; wherein, the test samples include the input information corresponding to each evaluation index obtained for the test product; After using the trained neural network model to conduct quality evaluation on the product to be evaluated, it further includes: if the score output by the neural network model for the product to be evaluated is less than the set score, then based on the magnitude relationship of the weights assigned to multiple evaluation indexes in the quality evaluation index system and the input information of each evaluation index for the product to be evaluated, determining at least one target evaluation index affecting the score, and outputting an improvement suggestion for the product to be evaluated according to the at least one target evaluation index.

6. A product quality evaluation device based on an artificial intelligence attention mechanism, characterized in that, The device includes: A construction unit for constructing a quality evaluation index system for a product; the product has a specified function; the quality evaluation index system includes multiple evaluation indexes; An acquisition unit, configured to acquire a plurality of training samples based on the quality evaluation index system; the training samples include input information corresponding to each evaluation index obtained for a sample product, and a score given to the sample product; the plurality of training samples are obtained from at least a plurality of different types of sample products, and the plurality of different types of sample products all have the specified function; the score is calculated from a first score and at least one second score; the first score is a score calculated according to a set scoring rule using the input information; and the evaluation indexes in the quality evaluation index system are physical performance indexes that can be measured; the second score is a score obtained by evaluating using non-performance indexes; A training unit, configured to use the input information corresponding to each evaluation index in the training samples as input and the score as output, so as to train a neural network model based on the attention mechanism using the plurality of training samples, so that the neural network model automatically identifies the evaluation indexes that it pays more attention to and adjusts the attention weights of different evaluation indexes; An evaluation unit, configured to perform quality evaluation on the product to be evaluated using the trained neural network model.

7. The device according to claim 6, characterized in that, It further includes: A first determination unit, configured to use the test samples and the trained neural network model to determine the magnitude relationship of the weights assigned by the neural network model to the plurality of evaluation indexes in the quality evaluation index system; wherein, the test samples include input information corresponding to each evaluation index obtained for the test product; A second determination unit, configured to, after the evaluation unit finishes execution, if the score output by the neural network model for the product to be evaluated is less than a set score, then based on the magnitude relationship of the weights assigned to the plurality of evaluation indexes in the quality evaluation index system and the input information of each evaluation index for the product to be evaluated, determine at least one target evaluation index that affects the score, and output an improvement suggestion for the product to be evaluated according to the at least one target evaluation index.

8. A computer device, characterized in that, The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored on the memory to implement the steps of the method according to any one of claims 1-5 above.

9. 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, it implements the steps of the method according to any one of claims 1-5.

10. A computer program product, characterized in that, It includes a computer program, and when the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-5.

Images