Noise frequency hearing loss prediction method and system based on machine learning

Abstract

The invention relates to the technical field of medical information, and discloses a noise-induced frequency hearing loss prediction method and system based on machine learning, and the method comprises the steps: inputting related medical record information into a pre-constructed user classification model, obtaining a classification result, determining correlation information according to environment noise data, and carrying out the prediction of the noise-induced frequency hearing loss. The method comprises the steps of determining environment noise data and correlation information, calculating a noise influence degree value based on the environment noise data and the correlation information, determining membership information according to user work data, calculating a work influence degree value based on the user work data and the membership information, and finally inputting the noise influence degree value, the work influence degree value and a standard influence factor into a pre-constructed loss prediction model. According to the method, the weight of the data of each dimension for hearing loss prediction is comprehensively considered, so that the loss prediction model identifies and quantifies the specific contribution of the data of each dimension to hearing loss, and the prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and more specifically, to a method and system for predicting noise-induced frequency hearing loss based on machine learning. Background Art

[0002] Hearing loss can lead to long-term defects in language cognition, comprehension and social adaptability. Complex noise occupations are one of the main causes of hearing loss. Long-term exposure to dangerous levels of noise can cause permanent hearing damage. Due to the prevalence of occupational noise exposure and the consequences of language frequency hearing loss, early prediction and timely intervention of language frequency hearing loss are needed. In the existing technology, there are few means of early prevention of hearing loss.

[0003] Then, the patent application with publication number CN111223564A provides a noise-induced hearing loss prediction system based on convolutional neural networks. This technical solution collects noise exposure data and workers' personal information, and converts them into time series data that can be processed by convolutional neural networks through a data preprocessing module. The hearing loss prediction module of the convolutional neural network is used to determine whether workers will suffer hearing loss under noise exposure; the Chinese patent with publication number CN111584065A provides a method, device, terminal and medium for predicting noise-induced hearing loss and screening susceptible populations. This patent obtains medical feature data and hearing threshold measurement data from a sample of the population to be tested, compares the predicted notch area value with the actual notch area value, and determines the individual susceptibility of the sample of the population to be tested.

[0004] Although the above-mentioned existing technologies all predict hearing loss through noise data and personal information, the above-mentioned existing technologies do not comprehensively consider the weight of each dimensional data for hearing loss prediction, which results in the prediction model being unable to identify and quantify the specific contribution of each dimensional data to hearing loss, affecting the accuracy of the prediction.

[0005] In view of this, the present invention proposes a noise-induced frequency hearing loss prediction method and system based on machine learning to solve the above-mentioned problem. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for predicting noise-induced frequency hearing loss based on machine learning.

[0007] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, a method for predicting noise-induced frequency hearing loss based on machine learning is provided, comprising: Obtain relevant medical record information, input the relevant medical record information into a pre-built user classification model, and obtain classification results, the classification results including users who are prone to hearing loss and users who are not prone to hearing loss, and the relevant medical record information represents medical record information related to hearing loss; Acquire environmental noise data and user work data, determine correlation information according to the environmental noise data, calculate noise impact value based on the environmental noise data and the correlation information, determine membership information according to the user work data, and calculate work impact value based on the user work data and the membership information; The corresponding standard impact factor is determined based on the classification result, and the noise impact value, work impact value and standard impact factor are input into the pre-built loss prediction model to obtain the loss prediction result.

[0008] Furthermore, the method for determining correlation information according to the environmental noise data includes: Obtain W groups of noise sub-data, traverse the W groups of noise sub-data, count the number of times any two first elements in each group of noise sub-data appear simultaneously, divide the number of simultaneous appearances by W to obtain a correlation factor, combine the S correlation factors to obtain correlation information, and the noise sub-data includes H first elements.

[0009] Furthermore, the environmental noise data includes a target duration, a target decibel value, and a target frequency value. The method for calculating the noise impact value based on the environmental noise data and the correlation information includes: According to the relationship between the target duration, target decibel value and target frequency value, the noise impact degree is obtained by combining logarithmic transformation, inverse trigonometric and hyperbolic function calculation.

[0010] Furthermore, the method for determining the membership information according to the user work data includes: Based on the user work data, the ratio between each second element and the corresponding standard value is calculated, and all the ratio values ​​are summed up to obtain the total ratio value. The ratio of the ratio value of each second element to the total ratio value is calculated and used as the membership of the second element. The user work data includes F second elements.

[0011] Furthermore, the user work data includes the user's length of service, average working hours, protection frequency value, and maximum contact time. The method for calculating the work influence value based on the user work data and the membership information includes: The work impact value is obtained by calculating the logarithmic transformation and tangent function of the user's length of service, average working hours, protection frequency value and maximum contact time.

[0012] Furthermore, the method for obtaining the standard impact factor includes: The initial impact factor corresponding to the preset classification result is input into the pre-built initial prediction model by inputting the noise impact value, the work impact value and the initial impact factor to obtain the initial prediction result; Obtain real-time detection results, calculate the similarity between the initial prediction results and the real-time detection results, optimize the initial impact factor based on the similarity and the intelligent optimization algorithm, and obtain the standard impact factor.

[0013] Furthermore, the method for constructing the initial prediction model includes: Acquire a sample data set, wherein the sample data set includes a historical noise impact value, a historical work impact value, a historical initial impact factor, and a historical initial prediction result; Divide the sample data set into a sample training set and a sample test set, and build a regression network; The historical noise impact value, the historical work impact value and the historical initial impact factor in the sample training set are used as the input data of the regression network, and the historical initial prediction results in the sample training set are used as the output data of the regression network, and the regression network is trained to obtain an initial regression network for predicting the initial prediction results; The initial regression network is tested using the sample test set, and the initial regression network with an output smaller than the preset error value is used as the initial prediction model.

[0014] Furthermore, the intelligent optimization algorithm is a simulated annealing algorithm, and the method for optimizing the initial influencing factors according to the similarity and the intelligent optimization algorithm includes: S301: Randomly select an initial impact factor as the initial dynamic ratio , preset the initial temperature of simulated annealing and cooling rate α, 1>α>0; S302: Initial dynamic ratio The similarity of As the objective function value; S303: In the current dynamic ratio Add random disturbances to generate new dynamic ratios , the current dynamic ratio The starting point is the initial dynamic ratio ; S304: Calculate new dynamic ratio The corresponding new similarity , the new similarity Similarity to current Compare, current similarity With the current dynamic ratio correspond; S305: Determine whether to update the dynamic ratio. , and returns to S303, the The starting point is ; S306: Repeat the above S303-S305 until < , stop iteration and output the optimal dynamic ratio , the optimal dynamic ratio As a standard impact factor, is the temperature parameter, The preset minimum temperature value.

[0015] Furthermore, the method for determining whether to update the dynamic ratio includes: like , then let = ; like , then randomly generate a probability threshold r between 0 and 1, and determine whether the probability K is greater than the probability threshold r. If so, let = If not, keep unchanged, through the new similarity Similarity to current The difference divided by the temperature parameter , obtain the calculation result, take the negative value of the calculation result as the input of the exponential function, and obtain the probability K.

[0016] In a second aspect, a noise-induced frequency hearing loss prediction system based on machine learning is provided, which is used to implement the above-mentioned noise-induced frequency hearing loss prediction method based on machine learning, including: User classification module: used to obtain relevant medical record information, input the relevant medical record information into a pre-built user classification model, and obtain classification results. The classification results include users who are prone to hearing loss and users who are not prone to hearing loss. The relevant medical record information represents medical record information related to hearing loss. Data processing module: used to obtain environmental noise data and user work data, determine correlation information according to the environmental noise data, calculate noise impact value based on the environmental noise data and the correlation information, determine membership information according to user work data, and calculate work impact value based on the user work data and the membership information; Loss prediction module: Determine the corresponding standard impact factor based on the classification result, input the noise impact value, work impact value and standard impact factor into the pre-built loss prediction model to obtain the loss prediction result.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention first inputs relevant medical record information into a pre-built user classification model to obtain a classification result, determines correlation information according to environmental noise data, calculates a noise impact value based on the environmental noise data and the correlation information, determines membership information according to user work data, calculates a work impact value based on the user work data and the membership information, and finally inputs the noise impact value, the work impact value and the standard impact factor into a pre-built loss prediction model to obtain a loss prediction result. The present invention comprehensively considers the weight of each dimensional data for hearing loss prediction, so that the loss prediction model identifies and quantifies the specific contribution of each dimensional data to hearing loss, thereby improving the accuracy of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of a method for predicting noise-induced frequency hearing loss based on machine learning in the present invention; Figure 2 It is a structural schematic diagram of a noise-induced frequency hearing loss prediction system based on machine learning in the present invention; Figure 3 It is a schematic diagram of a process of determining correlation information according to environmental noise data in the present invention; Figure 4 It is a schematic diagram of the process of determining membership information according to user work data in the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Example 1 See also Figure 1 As shown, this embodiment discloses a method for predicting noise-induced frequency hearing loss based on machine learning, comprising: S10: Obtain relevant medical record information, input the relevant medical record information into a pre-built user classification model, and obtain classification results, where the classification results include users who are prone to hearing loss and users who are not prone to hearing loss, and the relevant medical record information represents medical record information related to hearing loss; In this embodiment, the relevant medical record information represents medical record information related to hearing loss, for example, the relevant medical record information may be hypertension, chronic otitis media, GJB2 gene mutation or MT-RNR1 gene mutation, etc. Taking hypertension as an example, under the same noise conditions, the hearing loss of people with hypertension is usually more severe, because both noise and hypertension may induce oxidative stress response, aggravating the damage of cochlear hair cells. The cumulative effect of oxidative stress will cause hair cell damage faster, thereby leading to more severe hearing loss. Hypertension can also damage the microvascular circulation of the inner ear, so that the inner ear hair cells and other auditory structures cannot get sufficient blood support, resulting in reduced tolerance to noise, making them more vulnerable to damage in the same noise environment.

[0021] The construction method of the user classification model includes: Obtain Q groups of training data, where Q is a positive integer greater than 1, and the training data includes historical relevant medical record information and historical classification results; Use historical medical record information and historical classification results as sample sets, divide the sample sets into training sets and test sets, and build a classifier; The historical medical record information in the training set is used as input data, and the historical classification results in the training set are used as output data to train the classifier and obtain an initial classifier; The initial classifier is tested using the test set, and a classifier that meets the preset accuracy is output as the user classification model. The classifier is preferably a naive Bayes model or a support vector machine model.

[0022] It should be added that this embodiment can be divided according to preset rules. For example, if the relevant medical record information includes one of hypertension, chronic otitis media, GJB2 gene mutation or MT-RNR1 gene mutation, the user will be classified as a user prone to loss. Of course, other classification rules can also be preset, and this embodiment is not limited to this.

[0023] S20: Acquire environmental noise data and user work data, determine correlation information according to the environmental noise data, calculate a noise impact value based on the environmental noise data and the correlation information, determine membership information according to the user work data, and calculate a work impact value based on the user work data and the membership information; Among them, environmental noise data includes but is not limited to target duration, target decibel value and target frequency value. The target duration refers to the continuous time period in which the user is exposed to a high-noise environment within a certain period of time, and the time period is greater than the preset time threshold, reflecting the user's long-term exposure to high-intensity noise. The longer the duration, the higher the risk of hearing damage. The target decibel value refers to the noise intensity when the user is exposed to a high-noise environment and exceeds the preset decibel threshold. This indicator is used to identify high noise levels that exceed the safety range. The target frequency value refers to the noise frequency when the user is exposed to a high-noise environment and exceeds the preset frequency threshold. When the noise frequency exceeds the preset frequency threshold (usually in the high frequency band of 2kHz to 4kHz, this frequency band is particularly sensitive to the human ear), it means that the frequency component of the noise poses a higher risk to hearing.

[0024] It should be noted that the correlation information represents the degree of overlap between the target duration, target decibel value and target frequency value. That is to say, the correlation information reflects whether these noise characteristics reach a high level at the same time in the same time period. For example, when the target decibel value and target duration are concentrated on a specific center frequency, it means that the frequency, intensity and duration of the noise have a large overlap at this frequency, and the impact on hearing may be more significant.

[0025] See also Figure 3 , the method for determining correlation information based on environmental noise data includes: Obtain W groups of noise sub-data, traverse the W groups of noise sub-data, count the number of times any two first elements in each group of noise sub-data appear simultaneously, divide the number of simultaneous appearances by W to obtain a correlation factor, combine the S correlation factors to obtain correlation information, and the noise sub-data includes H first elements.

[0026] It is understandable that a person skilled in the art may perform W detections on the user's working environment and obtain W groups of noise sub-data, where the noise sub-data includes H first elements, for example, H is 3, and the noise sub-data may include a target duration, a target decibel value, and a target frequency value. Then, counting the number of times any two first elements appear at the same time refers to calculating in each group of noise data, the situation where any two indicators of the three first elements, namely, the target duration, the target decibel value, and the target frequency value, simultaneously reach a preset high threshold. For example, if the target duration and the target decibel value in a noise sub-data both reach or exceed their respective thresholds, then the situation where "the target duration and the target decibel value appear at the same time" is recorded once. It is not difficult to understand that the number of simultaneous occurrences will only be less than or equal to W, so the correlation factor is a real number between 0 and 1. In this embodiment, combining S correlation factors may be a simple collection of S correlation factors.

[0027] The method for calculating the noise impact value based on environmental noise data and correlation information includes: According to the relationship between the target duration, target decibel value and target frequency value, the noise impact degree is obtained by combining logarithmic transformation, inverse trigonometric and hyperbolic function calculation; Specifically include: NIV= ; Where NIV is the noise impact value, is the target duration, is the target decibel value, is the target frequency value, is the inverse cotangent function, is the hyperbolic cosine function, is the inverse sine function, is a logarithmic function with base e, is the correlation factor between the target duration and the target decibel value, is the correlation factor between the target duration and the target frequency value, is the correlation factor between the target decibel value and the target frequency value.

[0028] It should be added that For example, is the correlation factor between the target duration and the target decibel value, specifically referring to the degree of overlap between the target duration and the target decibel value in the noise environment, that is, It indicates the frequency or probability that the duration and intensity of noise reach or exceed the preset threshold at the same time within a certain period of time. For example, if the noise duration is long and the decibel value is high within a certain period of time, it indicates that the overlap between these two indicators is high. The value of will also be relatively large. The numerical value of the correlation factor can be used to measure the potential risk of hearing damage caused by the combined impact of these two factors in noise exposure. From the above content, it can be seen that the noise impact value is positively correlated with the degree of hearing loss. That is to say, the larger the noise impact value, the more serious the hearing loss caused by the noise to the user.

[0029] In this embodiment, the user work data at least includes the user's length of service, average working hours, protection frequency value and maximum contact time. The average working hours can be the working hours per day or the working hours per week. The protection frequency value can be the frequency of use of earplugs or earmuffs, for example, how many days in a week will earmuffs or earplugs be used. The maximum contact time refers to the longest period of continuous exposure to noise. The membership information represents the proportional factor of each second element in the user's work data.

[0030] See also Figure 4The method for determining membership information based on user work data includes: based on the user work data, calculating the ratio between each second element and the corresponding standard value, and summing up all the ratio values ​​to obtain the total ratio value, calculating the ratio of the ratio value of each second element to the total ratio value, and using it as the membership of the second element, and the user work data includes F second elements.

[0031] It should be noted that, taking the second element as the average working hours as an example, technical personnel in this field will pre-set the standard value corresponding to the average working hours based on expert experience. For example, the standard value can be 8 hours. When it exceeds 8 hours, it means that the user's daily noise exposure time exceeds the safe range. Then the proportion value corresponding to each second element is calculated in turn, and then summed up, and finally the subsequent membership is calculated. The advantage of calculating the membership in this way is that by calculating the proportion value of each second element and normalizing it to the membership, the relative contribution of each second element (such as user's length of service, average working hours, protection frequency, maximum contact time) to the noise exposure risk can be quantified, which makes the impact of each second element more intuitive and easy to analyze and compare. The work data of each user may vary due to position, working environment, protective measures, etc. By presetting standard values ​​and calculating membership, it is possible to adapt to individual differences and achieve more flexible and personalized risk assessment.

[0032] The method for calculating the work influence value based on user work data and membership information includes: The work impact value is calculated based on the logarithmic transformation and tangent function of the user's length of service, average working hours, protection frequency value and maximum contact time; Specifically include: WIV= ; Where WIV is the work impact value, The user's length of service, is the average working time, is the protection frequency value, is the maximum contact time, is the membership degree corresponding to the user's length of service, is the membership degree corresponding to the average working hours, is the membership degree corresponding to the maximum contact time, is the membership degree corresponding to the protection frequency value, and e is a natural constant.

[0033] Among them, taking the user's length of service and protection frequency value as an example, the longer the user's length of service, the more serious the user's hearing loss. Conversely, the larger the protection frequency value, the lighter the user's hearing loss. Therefore, from the above content, it can be seen that the work impact value is positively correlated with the severity of the user's hearing loss.

[0034] This embodiment can more accurately assess the comprehensive risk of noise exposure to hearing loss by sequentially combining the user's individual susceptibility and work noise exposure characteristics. This combination sequence enables the system to consider the individual's genetic or health sensitivity to noise when predicting hearing loss, and combine the actual noise level of the working environment to achieve a more targeted risk assessment.

[0035] S30: Determine the corresponding standard impact factor based on the classification result, input the noise impact value, the work impact value and the standard impact factor into a pre-built loss prediction model to obtain a loss prediction result.

[0036] Among them, the methods for obtaining the standard impact factor include: The initial impact factor corresponding to the preset classification result is input into the pre-built initial prediction model by inputting the noise impact value, the work impact value and the initial impact factor to obtain the initial prediction result; Obtain real-time detection results, calculate the similarity between the initial prediction results and the real-time detection results, optimize the initial impact factor based on the similarity and the intelligent optimization algorithm, and obtain the standard impact factor.

[0037] It should be noted that the initial impact factor refers to the influence weight of the noise impact value and the work impact value on the prediction result. For example, when the classification result is a user who is prone to loss, the initial impact factor can be set to 0.65:0.35, the noise impact value corresponds to 0.65, and the work impact value corresponds to 0.35. When the classification result is a user who is not prone to loss, the initial impact factor can be set to 0.5:0.5. The reason for this setting is that for users who are prone to loss, the target duration, target decibel value and target frequency value in the noise impact value may directly cause damage to the inner ear hair cells and auditory nerves. Therefore, high-intensity and long-term exposure to noise pose a greater risk of hearing damage to such users. Users who are not prone to loss have a higher tolerance to noise, so the effects of the noise impact value and the work impact value on their hearing loss risk are relatively balanced.

[0038] It should be noted that both the initial prediction results and the loss prediction results mentioned above can be the risk values ​​of severe hearing damage within a few years, for example, the risk values ​​of severe hearing damage within the next three years. It can be understood that the real-time detection results mentioned above are real-time risk values ​​obtained by technical personnel in this field after experimental statistics. For example, an experiment is conducted on 500 users and predictions are made at the same time. Then, when 80 users have severe hearing damage, the ratio of 80 to 500 can be used as the real-time risk value.

[0039] In this embodiment, the similarity between the initial prediction result and the real-time detection result is calculated, and the cosine similarity between the initial prediction result and the real-time detection result can be calculated. The purpose of calculating the similarity is to optimize the initial impact factor and obtain the standard impact factor to improve the accuracy of the loss prediction model. Specifically, through the optimized standard impact factor, the weights of the noise impact value and the work impact value are made more reasonable, so that the loss prediction result output by the prediction model is closer to the actual hearing loss risk situation.

[0040] The initial prediction model construction method includes: Acquire a sample data set, wherein the sample data set includes a historical noise impact value, a historical work impact value, a historical initial impact factor, and a historical initial prediction result; Divide the sample data set into a sample training set and a sample test set, and build a regression network; The historical noise impact value, the historical work impact value and the historical initial impact factor in the sample training set are used as the input data of the regression network, and the historical initial prediction results in the sample training set are used as the output data of the regression network, and the regression network is trained to obtain an initial regression network for predicting the initial prediction results; The initial regression network is tested using a sample test set, and the initial regression network with an output smaller than a preset error value is used as the initial prediction model. The initial regression network is a deep neural network model.

[0041] The intelligent optimization algorithm is a simulated annealing algorithm. The methods for optimizing the initial influencing factors according to the similarity and the intelligent optimization algorithm include: S301: Randomly select an initial impact factor as the initial dynamic ratio , preset the initial temperature of simulated annealing and cooling rate α, 1>α>0; In this example, an initial solution and parameters for controlling the annealing process are determined to provide a basis for subsequent optimization. Random selection of the initial dynamic ratio can avoid being limited to a specific solution at the beginning of the optimization, thereby having better exploratory potential. The setting of and cooling rate α also determines the “annealing” speed of the search process, ensuring that the entire optimization process has sufficient global search capability.

[0042] S302: Initial dynamic ratio The similarity of As the objective function value; S303: In the current dynamic ratio Add random disturbances to generate new dynamic ratios , the current dynamic ratio The starting point is the initial dynamic ratio ; Among them, a new candidate solution is formed to enrich the diversity of solutions. By adding random perturbations, the simulated annealing algorithm can jump out of the local optimum and find a possible global optimal solution. This process generates candidate solutions that are similar to but slightly different from the current solution by fine-tuning the current solution, further expanding the solution space.

[0043] S304: Calculate new dynamic ratio The corresponding new similarity , the new similarity Similarity to current Compare, current similarity With the current dynamic ratio correspond; Among them, by comparing the objective function values ​​of the new and old solutions, we can determine whether the new solution is better, and then decide whether to accept the new solution. The advantage of doing this is to ensure that each step moves in the direction of minimizing the objective function, so that the optimization process gradually approaches the optimal solution.

[0044] S305: Determine whether to update the dynamic ratio. , and returns to S303, the The starting point is ; This embodiment ensures that the algorithm gradually converges to the optimal solution while exploring the solution space through a gradual cooling process, thereby balancing the capabilities of global exploration and local development.

[0045] S306: Repeat the above S303-S305 until < , stop iteration and output the optimal dynamic ratio , the optimal dynamic ratio As a standard impact factor, is the temperature parameter, The preset minimum temperature value.

[0046] Generate new dynamic ratio The methods include: ; in, is a random disturbance.

[0047] Methods for determining whether to update the dynamic ratio include: like , then let = ; like , then randomly generate a probability threshold r between 0 and 1, and determine whether the probability K is greater than the probability threshold r. If so, let = If not, keep unchanged, the probability K is: K=exp ; Among them, exp It is an exponential function with e as the base, where e is a natural constant; using the temperature parameter The rate of change of the control probability K varies with the temperature parameter The decrease of , the probability of accepting a worse state will also decrease, which helps to jump out of the local optimal solution.

[0048] The negative value of the calculated result is to ensure that the new similarity Similarity than the current Time difference (i.e. > ), the value of the exponential function will be less than 1, thereby reducing the probability of accepting this worse state. Specifically: If the new state is better than the current state ( ), the difference is negative, and after taking the negative value, it becomes positive, and the value of the exponential function will be greater than 1, indicating that the probability of accepting this better state is higher; If the new state is worse than the current state ( > ), the difference is positive, and becomes negative after negation. The value of the exponential function will be less than 1, indicating that the probability of accepting this worse state is low.

[0049] This allows the early stages (temperature parameters higher) have a greater probability of accepting a worse solution to explore a wider solution space, while in the later stages (temperature parameter When the value is lower, it is more inclined to accept better solutions and gradually converge to a better solution.

[0050] It should be added that the training process of the loss prediction model mentioned above is the same as the training process of the initial prediction model, except that the historical initial influencing factors are replaced by historical standard influencing factors, and the historical initial prediction results are replaced by historical loss prediction results. This embodiment does not go into details about this.

[0051] In this embodiment, the noise impact value and the work impact value are first calculated, and then the similarity between the initial prediction result and the real-time detection result is calculated through an intelligent optimization algorithm, the initial impact factor is optimized to obtain the final standard impact factor, and finally the loss prediction result is obtained based on the noise impact value, the work impact value and the standard impact factor. This process improves the closeness of the loss prediction result to the actual situation and avoids the deviation of relying solely on theoretical models.

[0052] This embodiment first inputs the relevant medical record information into a pre-built user classification model to obtain a classification result, determines the correlation information according to the environmental noise data, calculates the noise impact value based on the environmental noise data and the correlation information, determines the membership information according to the user work data, calculates the work impact value based on the user work data and the membership information, and finally inputs the noise impact value, the work impact value and the standard impact factor into the pre-built loss prediction model to obtain a loss prediction result. In this way, this embodiment comprehensively considers the weight of each dimensional data for hearing loss prediction, so that the loss prediction model identifies and quantifies the specific contribution of each dimensional data to hearing loss, thereby improving the accuracy of the prediction.

[0053] Example 2

[0054] See also Figure 2 As shown, based on the same inventive concept, this embodiment discloses a noise-induced frequency hearing loss prediction system based on machine learning. For details not provided in this embodiment, please refer to the description of the relevant parts in Embodiment 1. The system includes: User classification module: used to obtain relevant medical record information, input the relevant medical record information into a pre-built user classification model, and obtain classification results. The classification results include users who are prone to hearing loss and users who are not prone to hearing loss. The relevant medical record information represents medical record information related to hearing loss. The construction method of the user classification model includes: Obtain Q groups of training data, where Q is a positive integer greater than 1, and the training data includes historical relevant medical record information and historical classification results; Use historical medical record information and historical classification results as sample sets, divide the sample sets into training sets and test sets, and build a classifier; The historical medical record information in the training set is used as input data, and the historical classification results in the training set are used as output data to train the classifier and obtain an initial classifier; The initial classifier is tested using the test set, and a classifier that meets the preset accuracy is output as the user classification model. The classifier is preferably a naive Bayes model or a support vector machine model.

[0055] Data processing module: used to obtain environmental noise data and user work data, determine correlation information according to the environmental noise data, calculate noise impact value based on the environmental noise data and the correlation information, determine membership information according to user work data, and calculate work impact value based on the user work data and the membership information; The method for determining correlation information based on environmental noise data includes: Obtain W groups of noise sub-data, traverse the W groups of noise sub-data, count the number of times any two first elements in each group of noise sub-data appear simultaneously, divide the number of simultaneous appearances by W to obtain a correlation factor, combine the S correlation factors to obtain correlation information, and the noise sub-data includes H first elements.

[0056] Methods for determining membership information based on user work data include: Based on the user work data, the ratio between each second element and the corresponding standard value is calculated, and all the ratio values ​​are summed up to obtain the total ratio value. The ratio of the ratio value of each second element to the total ratio value is calculated and used as the membership of the second element. The user work data includes F second elements.

[0057] Loss prediction module: Determine the corresponding standard impact factor based on the classification result, input the noise impact value, work impact value and standard impact factor into the pre-built loss prediction model to obtain the loss prediction result.

[0058] The above formulas are all dimensionless and numerical calculations. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technical personnel in this field according to actual conditions.

[0059] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0060] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0061] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0062] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0063] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0064] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0065] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0066] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting noise-induced frequency hearing loss based on machine learning, characterized in that: include: Obtain relevant medical record information, input the relevant medical record information into a pre-built user classification model, and obtain classification results, the classification results including users who are prone to hearing loss and users who are not prone to hearing loss, and the relevant medical record information represents medical record information related to hearing loss; Acquire environmental noise data and user work data, determine correlation information according to the environmental noise data, calculate noise impact value based on the environmental noise data and the correlation information, determine membership information according to the user work data, and calculate work impact value based on the user work data and the membership information; The corresponding standard impact factor is determined based on the classification result, and the noise impact value, work impact value and standard impact factor are input into the pre-built loss prediction model to obtain the loss prediction result.

2. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 1, characterized in that: The method for determining correlation information according to environmental noise data comprises: Obtain W groups of noise sub-data, traverse the W groups of noise sub-data, count the number of times any two first elements in each group of noise sub-data appear simultaneously, divide the number of simultaneous appearances by W to obtain a correlation factor, combine the S correlation factors to obtain correlation information, and the noise sub-data includes H first elements.

3. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 2, characterized in that: The environmental noise data includes a target duration, a target decibel value, and a target frequency value. The method for calculating the noise impact value based on the environmental noise data and the correlation information includes: According to the relationship between the target duration, target decibel value and target frequency value, the noise impact degree is obtained by combining logarithmic transformation, inverse trigonometric and hyperbolic function calculation.

4. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 1, characterized in that: The method for determining membership information according to user work data comprises: Based on the user work data, the ratio between each second element and the corresponding standard value is calculated, and all the ratio values ​​are summed up to obtain the total ratio value. The ratio of the ratio value of each second element to the total ratio value is calculated and used as the membership of the second element. The user work data includes F second elements.

5. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 2, characterized in that: The user work data includes user length of service, average working hours, protection frequency value and maximum contact time. The method for calculating the work impact value based on the user work data and membership information includes: The work impact value is obtained by calculating the logarithmic transformation and tangent function of the user's length of service, average working hours, protection frequency value and maximum contact time.

6. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 1, characterized in that: The method for obtaining the standard impact factor includes: The initial impact factor corresponding to the preset classification result is input into the pre-built initial prediction model by inputting the noise impact value, the work impact value and the initial impact factor to obtain the initial prediction result; Obtain real-time detection results, calculate the similarity between the initial prediction results and the real-time detection results, optimize the initial impact factor based on the similarity and the intelligent optimization algorithm, and obtain the standard impact factor.

7. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 6, characterized in that: The method for constructing the initial prediction model includes: Acquire a sample data set, wherein the sample data set includes a historical noise impact value, a historical work impact value, a historical initial impact factor, and a historical initial prediction result; Divide the sample data set into a sample training set and a sample test set, and build a regression network; The historical noise impact value, the historical work impact value and the historical initial impact factor in the sample training set are used as the input data of the regression network, and the historical initial prediction results in the sample training set are used as the output data of the regression network, and the regression network is trained to obtain an initial regression network for predicting the initial prediction results; The initial regression network is tested using the sample test set, and the initial regression network with an output smaller than the preset error value is used as the initial prediction model.

8. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 7, characterized in that: The intelligent optimization algorithm is a simulated annealing algorithm. The method for optimizing the initial influencing factors according to the similarity and the intelligent optimization algorithm includes: S301: Randomly select an initial impact factor as the initial dynamic ratio , preset the initial temperature of simulated annealing and cooling rate α, 1>α>0; S302: Initial dynamic ratio The similarity of As the objective function value; S303: In the current dynamic ratio Add random disturbances to generate new dynamic ratios , the current dynamic ratio The starting point is the initial dynamic ratio ; S304: Calculate new dynamic ratio The corresponding new similarity , the new similarity Similarity to current Compare, current similarity With the current dynamic ratio correspond; S305: Determine whether to update the dynamic ratio. , and returns to S303, the The starting point is ; S306: Repeat the above S303-S305 until < , stop iteration and output the optimal dynamic ratio , the optimal dynamic ratio As a standard impact factor, is the temperature parameter, The preset minimum temperature value.

9. The method for predicting noise-induced frequency hearing loss based on machine learning according to claim 8, characterized in that: The method for determining whether to update the dynamic ratio includes: like , then let = ; like , then randomly generate a probability threshold r between 0 and 1, and determine whether the probability K is greater than the probability threshold r. If so, let = If not, keep unchanged, through the new similarity Similarity to current The difference divided by the temperature parameter , obtain the calculation result, take the negative value of the calculation result as the input of the exponential function, and obtain the probability K.

10. A noise-induced frequency hearing loss prediction system based on machine learning, which is used to implement a noise-induced frequency hearing loss prediction method based on machine learning according to any one of claims 1 to 9, characterized in that: include: User classification module: used to obtain relevant medical record information, input the relevant medical record information into a pre-built user classification model, and obtain classification results. The classification results include users who are prone to hearing loss and users who are not prone to hearing loss. The relevant medical record information represents medical record information related to hearing loss. Data processing module: used to obtain environmental noise data and user work data, determine correlation information according to the environmental noise data, calculate noise impact value based on the environmental noise data and the correlation information, determine membership information according to user work data, and calculate work impact value based on the user work data and the membership information; Loss prediction module: Determine the corresponding standard impact factor based on the classification result, input the noise impact value, work impact value and standard impact factor into the pre-built loss prediction model to obtain the loss prediction result.

Images