Learning disorder identification method and device and electronic equipment

Abstract

The embodiment of the invention discloses a learning disorder recognition method and device and electronic equipment, and the method comprises the steps: outputting an interaction interface, and obtaining the learning difficulty degree of a to-be-recognized child in daily life damage, score damage and learning disorder subtype in the interaction interface, and the identity information of the to-be-recognized child; after calculating scores of all the dimensions, inputting the scores and the identity information into a pre-trained target recognition model to obtain an initial judgment result about whether learning obstacles exist or not; when the initial judgment is positive, target user information is judged through subtype truncation values, and classification results of all subtypes are obtained; the target recognition model can be a logistic regression model, a random forest model, a support vector machine or an extreme gradient lifting model. By implementing the application, multi-dimensional information can be automatically collected through the electronic scale, obstacles can be quickly and accurately screened through the machine learning model, subtypes are refined, the evaluation efficiency and objectivity are improved, and the method is suitable for large-scale early intervention of schools and medical institutions.

Description

Technical Field

[0001] This application relates to the field of computer technology, and specifically to a method, apparatus, and electronic device for identifying learning disabilities. Background Technology

[0002] Learning disorder (LD) is a neurodevelopmental disorder, referred to as specific learning disorder (SLD) in the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5). It is characterized by persistent difficulties in reading, writing, numeracy, or mathematical reasoning skills in school-aged children, with current academic skills significantly below the cultural and linguistic average for reading, writing, or mathematics, and a significant impact on their academic achievement, vocational abilities, or daily life. However, this impairment is not caused by a lack of learning opportunities, intellectual disability, traumatic brain injury, or disease, but is based on abnormal biological functions during brain development.

[0003] The prevalence of specific learning disabilities affecting academic areas such as reading, written expression, and mathematics is 5%–15%, with an adult prevalence of approximately 4%, and the detection rate is trending upward. As a neurodevelopmental disorder that affects individuals throughout their lives, learning disabilities typically lead to lower academic achievement, higher dropout rates, and low self-esteem, negatively impacting children and adults in multiple ways, including education, employment, economics, society, and emotional well-being.

[0004] Unfortunately, there is a lack of learning disability prediction tools in China to date. Although there are a few screening scales, different scales are based on different theoretical bases, have different applicable age ranges, and their reliability and validity are not satisfactory. More importantly, they cannot distinguish subtypes according to DSM-5. Therefore, this greatly limits the identification and early intervention of learning disabilities in clinical practice. Summary of the Invention

[0005] This application discloses a method, device, and electronic device for identifying learning disabilities. It can obtain the identity information of the child to be identified, as well as the degree of impairment or difficulty in different dimensions, through an electronic scale. By analyzing the above information through a trained machine learning model, it can automatically determine whether the child to be identified has a learning disability and further determine the specific subtype of learning disability.

[0006] This application discloses a method for identifying learning disabilities, the method comprising: Output the interactive interface; The system obtains the learning disability screening questionnaire responses entered in the interactive interface and the identity information of the child to be identified. The learning disability screening questionnaire is used to collect information on the child to be identified in the following dimensions: the degree of impairment in daily life, the degree of impairment in academic performance, and the degree of learning difficulty corresponding to each of the three learning disability subtypes. The three learning disability subtypes include: reading disability, writing disability, and mathematics learning disability. Calculate the test scores of the child to be identified in each of the dimensions based on the answer information; The target user information of the child to be identified is input into the target recognition model; the target user information includes the scores of the child to be identified on each of the dimensions and the identity information of the child to be identified. Obtain the recognition result output by the target recognition model; When the identification result indicates that the child to be identified has a learning disability, the subtype classification result is determined based on the cutoff values ​​corresponding to the three learning disability subtypes and the test scores of the child to be identified on each of the dimensions. The target recognition models are all pre-trained and are any one of the following machine learning models: logistic regression model, random forest model, support vector machine or gradient boosting model.

[0007] As an optional implementation, the target recognition model is a logistic regression model.

[0008] As an optional implementation, the three learning disability subtypes each correspond to different cutoff values, and a test score greater than or equal to the cutoff value is considered positive, wherein: The cutoff value for reading difficulties is 36; The cutoff value for writing difficulties is 48; The cutoff value for math learning difficulties is 45.

[0009] As an optional implementation, after calculating the test scores of the child to be identified on each of the dimensions based on the answer information, the method further includes: The test scores of the child to be identified in each of the dimensions are output on the interactive interface.

[0010] As an optional implementation, before inputting the target user information of the child to be identified into the target recognition model, the method further includes: To improve the ability to identify children with learning disabilities, a number of initial sample models are trained one by one using a training dataset to obtain a number of converged sample models. The initial sample models are all binary classification machine learning models. The training dataset includes the test scores of several first sample children on each of the dimensions and the identity information of the first sample children. The performance of each converged sample model in identifying children with learning disabilities was tested using a test dataset, and the performance scores of each converged sample model under the evaluation metrics were obtained. The test dataset includes the scores of several second sample children on each dimension and the identity information of the second sample children. The evaluation metrics include one or more of the following metrics: area under the curve, sensitivity, specificity, positive predictive value, negative predictive value, F2 score, and accuracy. The performance scores of each converged sample model under the evaluation index are compared, and the converged sample model with the best ability to identify children with learning disabilities is determined as the target recognition model.

[0011] As an alternative implementation, the training dataset includes real data collected from real children and generated data; the generated data is generated based on the real data using a synthetic minority class oversampling technique.

[0012] As an optional implementation, the cutoff values ​​corresponding to the three learning disability subtypes are obtained through the following steps: Receiver Operating Characteristic (ROC) curve analysis is performed on the target recognition model, with negative samples from the first and second sample children as the control group, and positive samples of the learning disability subtypes diagnosed by the hospital as the case group, and the value corresponding to the Youden Index reaching its maximum is used as the cutoff value corresponding to each learning disability subtype.

[0013] This application discloses an identification device for children with learning disabilities, the device comprising: An interactive unit is used to output an interactive interface; and to obtain the learning disability screening scale response information filled in the interactive interface and the identity information of the child to be identified; the learning disability screening scale is used to collect information of the child to be identified in the following dimensions: the degree of impairment in daily life, the degree of impairment in academic performance, and the degree of learning difficulty corresponding to the three learning disability subtypes; the three learning disability subtypes include: reading disability, writing disability, and mathematics learning disability. A calculation unit is configured to calculate the test scores of the child to be identified in each of the dimensions based on the answer information; A learning disability identification unit is used to input the target user information of the child to be identified into the target identification model; the target user information includes the scores of the child to be identified in each of the dimensions and the identity information of the child to be identified; and to obtain the identification result output by the target identification model. The subtype identification unit is used to determine the subtype classification result based on the cutoff values ​​corresponding to the three learning disability subtypes and the test scores of the child to be identified on each of the dimensions when the identification result indicates that the child to be identified has a learning disability. The target recognition models are all pre-trained and are any one of the following machine learning models: logistic regression model, random forest model, support vector machine or gradient boosting model.

[0014] As an optional implementation, the device further includes: The training unit is used to train several initial sample models one by one using a training dataset to obtain several converged sample models, targeting the identification ability of children with learning disabilities. The several initial sample models are all binary classification machine learning models. The training dataset includes the scores of several first sample children on each dimension and the identity information of the first sample children. The evaluation unit is used to test the performance of each converged sample model in identifying children with learning disabilities using a test dataset, and to obtain the performance score of each converged sample model under the evaluation index. The test dataset includes the scores of several second sample children on each dimension and the identity information of the second sample children. The evaluation index includes one or more of the following indices: area under the curve, sensitivity, specificity, positive predictive value, negative predictive value, F2 score, and accuracy. A determining unit is used to compare the performance scores of each of the converged sample models under the evaluation index, and to determine the converged sample model with the best ability to identify children with learning disabilities from among the converged sample models as the target recognition model.

[0015] This application discloses an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor implements any of the learning disability identification methods disclosed in this application.

[0016] This application discloses a computer-readable storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements any of the learning disability identification methods disclosed in this application.

[0017] Compared with related technologies, the embodiments of this application have the following beneficial effects: By implementing the embodiments of this application, the identity information of the child to be identified, as well as the degree of impairment or difficulty in different dimensions, can be obtained through electronic scales. By analyzing the above information through a trained machine learning model, it can be automatically determined whether the child to be identified has a learning disability, and further determine the specific subtype of learning disability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic flowchart of a method for identifying learning disabilities disclosed in an embodiment of this application; Figure 2 This application discloses a method for training and screening a model for identifying children with learning disabilities. Figure 3 This is a schematic diagram of the structure of an identification device for children with learning disabilities disclosed in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] It should be noted that the terms "comprising" and "having," and any variations thereof, in the embodiments and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0022] This application discloses a method, device, and electronic device for identifying learning disabilities. It can obtain the identity information of the child to be identified, as well as the degree of impairment or difficulty in different dimensions, through an electronic scale. By analyzing the above information through a trained machine learning model, it can automatically determine whether the child to be identified has a learning disability and further determine the specific subtype of learning disability.

[0023] Please see Figure 1 , Figure 1 This is a schematic flowchart of a method for identifying learning disabilities disclosed in an embodiment of this application. Figure 1 The method shown can be executed by electronic devices with computing capabilities, such as personal computers and industrial computers. Figure 1 As shown, the method may include the following steps: 110. Output the interactive interface.

[0024] In this embodiment, the interactive interface is a human-computer interaction interface used to obtain information about the child to be identified. The interactive interface can be a graphical user interface, such as a website webpage, a client window, or a smartphone application page.

[0025] The interactive interface can display a learning disability screening scale, which includes multiple questions. The answers to each question are quantified by scores representing different levels of possible answers. For example, each question includes 5 possible answers, each with a score of an integer in the range [1, 5]. The 5 possible answers are arranged from lowest to highest score as follows: never appeared, occasionally appeared, sometimes appeared, more often appeared, and frequently appeared.

[0026] The scale's questions are divided into multiple dimensions, with each dimension potentially including one or more questions. These dimensions include: the degree of impairment in daily living, the degree of impairment in academic performance, and the degree of learning difficulty corresponding to three learning disability subtypes. The learning disability subtypes include, but are not limited to: dyslexia, dysgraphia, or mathematical learning disability.

[0027] The questions corresponding to the degree of impairment in daily life collect information on the difficulties children experience in daily activities and emotions; the questions corresponding to the degree of impairment in academic performance collect information on the academic difficulties children face at school; the questions corresponding to the degree of reading difficulty collect information on the accuracy, fluency, and ability of children's reading; the questions corresponding to the degree of writing difficulty collect information on the accuracy of children's spelling, grammar and punctuation, and the clarity and organization of their written expression; and the questions corresponding to the degree of difficulty in learning mathematics collect information on children's number sense, arithmetic memory, calculation ability, and mathematical reasoning ability.

[0028] The interactive interface may also include an interface for collecting identity information of the child to be identified, which may include, but is not limited to, grade and gender.

[0029] 120. Obtain the learning disability screening questionnaire answers entered in the interactive interface and the identity information of the child to be identified.

[0030] In this embodiment of the application, the answer information of the learning disability screening scale is filled in by the user of the interactive interface according to the actual situation of the child to be identified. The user of the interactive interface can be the parents, teachers or other close relatives of the child to be identified, or the child to be identified himself, and there is no specific limitation.

[0031] 130. Calculate the test scores of the child to be identified in each dimension based on the answer information.

[0032] In this embodiment of the application, the test score of the child to be identified in each dimension is obtained by scoring the answer information corresponding to each question included in that dimension. The scoring rule can be to calculate the average value or to calculate the cumulative sum, and there is no specific limitation.

[0033] As an optional implementation method, after calculating the test scores of the child to be identified in each dimension, the scores can be output on the interactive interface so that the user can quantitatively understand the specific results of the form filling.

[0034] 140. Input the target user information of the child to be identified into the target recognition model.

[0035] In this embodiment of the application, the target user information includes the scores of the child to be identified in each dimension calculated in step 130 and the identity information of the child to be identified obtained in step 120.

[0036] The target recognition model can be any of the following machine learning models: Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), or eXtreme Gradient Boosting (XGBoost). The target recognition model is a convergent model obtained after training. The training method can refer to any method known in the field, such as supervised learning or unsupervised learning, without specific limitations. After training, the target recognition model can determine whether the child to be identified has a learning disability based on the input target user information and obtain the corresponding recognition result.

[0037] 150. Obtain the recognition results output by the target recognition model.

[0038] In this embodiment of the application, the target recognition model outputs two types of recognition results: one result indicates that the child to be identified has a learning disability; the other result indicates that the child to be identified does not have a learning disability.

[0039] 160. When the recognition results output by the target recognition model indicate that the child to be identified has a learning disability, the subtype classification result is determined based on the cutoff values ​​corresponding to the three learning disability subtypes and the test scores of the child to be identified in each dimension.

[0040] In the embodiments of this application, the cutoff values ​​corresponding to the three learning disability subtypes can be set based on experience, or they can be determined by using Receiver Operating Characteristic (ROC) curve analysis during the training of the target recognition model, with the value corresponding to the maximum Youden index. No specific limitation is made.

[0041] Optionally, the three learning disability subtypes can each correspond to different cutoff values. A test score greater than or equal to the cutoff value is considered positive, meaning that the child to be identified possesses the corresponding learning disability subtype. This application discloses a set of exemplary learning disability subtype cutoff values, which have good performance and can accurately identify different learning disability subtypes. Specifically, the cutoff value for reading disability is 36; for writing disability, it is 48; and for mathematical learning disability, it is 45.

[0042] For example, when the target recognition model outputs a recognition result indicating that the child to be identified has a learning disability, the first total test score for the child in three dimensions—impairment in daily life, impairment in academic performance, and dyslexia—can be calculated. This first total test score is then compared with the cutoff value for dyslexia. If the total test score is greater than 36, the subtype classification result for the child to be identified includes dyslexia. Since the same child may have multiple different learning disability subtypes, a second total test score for the child in three dimensions—impairment in daily life, impairment in academic performance, and writing difficulties—can be calculated. If the second total test score is greater than 48, the subtype classification result for the child to be identified also includes writing difficulties. Similarly, a third total test score for the child in three dimensions—impairment in daily life, impairment in academic performance, and math learning difficulties—can be calculated. If the third total test score is greater than 45, the subtype classification result for the child to be identified also includes math learning difficulties. Finally, if the child is identified as having a learning disability but cannot be classified into any of the above three subtypes, "unspecified type" is output.

[0043] It should be noted that when determining whether a child to be identified has any of the three learning disability subtypes mentioned above, comparisons can be made in parallel or sequentially. The above is an example of sequential comparison, but there is no logically necessary order in which the identification of the three learning disability subtypes is performed, and the above example does not constitute a limitation.

[0044] In summary, the learning disability identification method disclosed in this application can determine whether a child has a learning disability after collecting target user information about the child to be identified, using a target identification model. Furthermore, after determining that a child has a learning disability, it is not necessary to collect information again; the previously collected target user information can be reused to further determine the specific type of learning disability through two or more subtype identification models. Learning disabilities are often complex; a child may have only one learning disability subtype, or several subtypes may be present. Therefore, implementing this application can accurately identify the presence of a learning disability and further refine the identification, facilitating early identification of children with learning disabilities and enabling targeted intervention. In addition, the entire identification process can be completed automatically by electronic devices; users only need to fill in a questionnaire and basic identity information, greatly reducing reliance on human experience and lowering the threshold for learning disability identification.

[0045] To better illustrate the convergence identification model disclosed in the embodiments of this application, the following describes how to obtain this convergence identification model. Please refer to... Figure 2 , Figure 2 This application discloses a method for training and screening a model for identifying children with learning disabilities. Similarly, this method can be executed by a computing-capable electronic device and includes the following steps: 210. To improve the ability to identify children with learning disabilities, several initial sample models are trained one by one using the training dataset to obtain several converged sample models.

[0046] In this embodiment, the initial sample models are all binary classification machine learning models. For example, the aforementioned four machine learning models can be trained one by one, enabling each model to recognize children with learning disabilities. Optionally, the LR model can be a regularized LR model; the SVM can be a linear SVM.

[0047] In this embodiment of the application, data from multiple children with learning disabilities and typically developing children can be collected as a sample dataset through ethically reviewed and legal channels. The inclusion criteria for the sample children during collection may include: 1) Elementary school students in grades 2-6; 2) Normal hearing and vision (or corrected to normal vision); 3) Intelligence Quotient (IQ) > 70.

[0048] Of the collected sample data, 70% can be used as the training dataset, and the remaining 30% can be used as the test dataset. That is, the training and test datasets come from different sample children, referred to below as the first sample children and the second sample children. The training dataset includes the scores and identity information of several first sample children across the aforementioned dimensions; the test dataset includes the scores and identity information of several second sample children across the aforementioned dimensions.

[0049] The aforementioned dimensions include: the degree of impairment in daily living, the degree of impairment in academic performance, and the degree of learning difficulty corresponding to the three learning disability subtypes; the three learning disability subtypes include: dyslexia, dysgraphia, or mathematical learning disability. Identification information collected from the sample children includes: grade level or gender, etc. The above dimensions and identification information have been validated using the Least Absolute Shrinkage and Selection Operator (LASO) and possess a certain degree of representativeness.

[0050] Although the incidence of learning disabilities has reached 5% to 15% in recent years, due to a lack of awareness, diagnosis remains difficult, and collecting data on eligible children is still challenging. A serious imbalance in the ratio of positive samples (children with learning disabilities) to negative samples (normal children) is likely to occur, especially with a very small number of positive samples.

[0051] As an optional implementation, the training dataset may include real data collected from real children and generated data. Real data is collected from real people according to the aforementioned inclusion criteria, while generated data is computer-generated based on real data. Further optionally, the generated data is generated based on real data using the Synthetic Minority Over-sampling Technique (SMOTE). SMOTE synthesizes new minority class samples through interpolation, thereby balancing the class distribution and avoiding overfitting. After generation, the ratio of positive to negative samples in the training dataset can be controlled to be close to 1:1.

[0052] 220. Using the test dataset, test the ability of each converged sample model to identify children with learning disabilities, and obtain the performance score of each converged sample model under the evaluation index.

[0053] In this embodiment of the application, the initial sample model is trained to identify the ability of children with learning disabilities. Therefore, the test is also to test the accuracy of each converged sample model in identifying whether the second sample child has a learning disability.

[0054] In this application embodiment, the evaluation metrics include one or more of the following metrics: Area Under the Curve (AUC), Sensitivity, Specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), F2 score, and Accuracy. Each of these evaluation metrics can be used individually to evaluate the model's performance, and the specific calculation methods are known techniques, which will not be elaborated upon further below.

[0055] 230. Compare the performance scores of each convergent sample model under the evaluation index, and determine the convergent sample model with the best ability to identify children with learning disabilities as the target recognition model.

[0056] In this embodiment, if only the score of each converged sample model under a single evaluation index is calculated, the converged sample model with the highest score can be selected as the convergence identification model according to the order from high to low. If multiple scores of each converged sample model under multiple evaluation indexes are calculated, one converged sample model can be selected from the multiple scores according to a preset rule as the convergence identification model.

[0057] For example, taking convergent sample models including logistic regression (LR), random forest (RF), decision tree (DT), or extreme gradient boosting (XGBoost) models as examples, the table below shows the evaluation index scores of each model in identifying whether a child has a learning disability on the test dataset.

[0058]

[0059] It is evident that the LR model has the highest accuracy and superior sensitivity, NPV, and F2 score. Although its specificity is relatively low, given the balanced performance and low false negative rate of the LR model, it can be selected as an alternative implementation method for determining whether a child has a learning disability.

[0060] As mentioned earlier, the cutoff values ​​for each different learning disability subtype can be determined using ROC curve analysis during the training of the target recognition model. Below are examples of the cutoff values ​​for the three learning disability subtypes.

[0061]

[0062]

[0063]

[0064] In summary, based on the training method disclosed in the embodiments of this application, the best-performing model can be trained from multiple machine learning models to serve as the target recognition model, thereby improving the accuracy of learning disability identification. Furthermore, the ROC curve analysis used during model training to determine the cutoff values ​​corresponding to different learning disability subtypes has statistical significance and demonstrates higher accuracy and better general applicability compared to empirical values.

[0065] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a device for identifying children with learning disabilities, as disclosed in an embodiment of this application. This device can be applied to any of the aforementioned electronic devices. Figure 3 As shown, the device includes: Interaction unit 310 is used to output an interactive interface; and to obtain the learning disability screening scale answer information filled in the interactive interface and the identity information of the child to be identified; The learning disability screening scale is used to collect information on the child to be identified in the following dimensions: the degree of impairment in daily life, the degree of impairment in academic performance, and the degree of learning difficulty corresponding to the three learning disability subtypes; The three learning disability subtypes include: reading disability, writing disability and mathematics learning disability; The calculation unit 320 is used to calculate the test scores of the child to be identified in each dimension based on the answer information; The learning disability identification unit 330 is used to input the target user information of the child to be identified into the target identification model; the target user information includes the scores of the child to be identified in various dimensions and the identity information of the child to be identified; and to obtain the identification result output by the target identification model. The subtype identification unit 340 is used to determine the subtype classification result based on the cutoff values ​​corresponding to the three learning disability subtypes and the test scores of the child to be identified in each dimension when the identification result indicates that the child to be identified has a learning disability. In this model, the target recognition models are all pre-trained and can be any of the following machine learning models: logistic regression model, random forest model, support vector machine or gradient boosting model.

[0066] Optionally, the target recognition model used by the learning disability recognition model 330 is a logistic regression model.

[0067] Optionally, the three learning disability subtypes correspond to different cutoff values. A test score greater than or equal to the cutoff value is considered positive. Specifically, the cutoff value for reading disability is 36; for writing disability, it is 48; and for mathematical learning disability, it is 45.

[0068] Optionally, the interaction unit 310 can also be used to output each test score on the aforementioned interaction interface after the calculation unit 320 calculates the test scores of the child to be identified in each dimension.

[0069] In some possible embodiments, the identification device for children with learning disabilities may further include: Training unit 350 is used to train several initial sample models one by one using the training dataset to obtain several converged sample models. The initial sample models are all binary classification machine learning models. The training dataset includes the scores of several first sample children on various dimensions and the identity information of the first sample children. Evaluation unit 360 is used to test the performance of each converged sample model in identifying children with learning disabilities using the test dataset, and obtain the scores of each converged sample model under the evaluation metrics. The test dataset includes the scores of several second sample children in various dimensions and the identity information of the second sample children. The evaluation metrics include one or more of the following metrics: area under the curve, sensitivity, specificity, positive predictive value, negative predictive value, F2 score, and accuracy. Unit 370 is used to compare the performance scores of each convergent sample model under the evaluation index, and to determine the convergent sample model with the best ability to identify children with learning disabilities as the target recognition model.

[0070] Optionally, the training dataset used by training unit 350 includes real data collected from real children and generated data; the generated data is generated based on real data through synthetic minority class oversampling techniques.

[0071] Optionally, the training unit 350 is also used to perform ROC curve analysis on the target recognition model, with negative samples from the first and second sample children as the control group, and positive samples of learning disability subtypes diagnosed by the hospital as the case group, and the value corresponding to the maximum value of the Youden index as the cutoff value corresponding to each learning disability subtype.

[0072] It should be noted that the aforementioned interaction unit 310, calculation unit 320, learning disability identification unit 330, and subtype identification unit 340 can be referred to as the model application part of the identification device for children with learning disabilities; the training unit 350, evaluation unit 360, and determination unit 370 can be referred to as the model training part of the identification device for children with learning disabilities. The model application part and the model training part of this device can be physically arranged in different computing devices, or they can be integrated in the same computing device, without any specific limitation.

[0073] Implementing the aforementioned learning disability identification device allows for the collection of target user information about the child to be identified. This information is then used by a target recognition model to determine whether the child has a learning disability. Furthermore, once a learning disability is identified, the previously collected target user information can be reused to further determine the specific type of learning disability, eliminating the need for further information collection. This facilitates early identification of children with learning disabilities and enables targeted intervention. Moreover, the entire identification process can be automated by the device. Users only need to fill out a questionnaire and provide basic identity information, significantly reducing reliance on human experience and lowering the barrier to learning disability identification.

[0074] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. For example... Figure 4 As shown, the testing equipment may include: Memory 410 containing computer programs; Processor 420 coupled to memory 410; When the computer program stored in the memory 410 is executed by the processor 420, the processor 420 performs any of the learning disability identification methods disclosed in the embodiments of this application.

[0075] It should be noted that the electronic device shown in the figure may also include components not shown, such as a power supply, input buttons, screen, RF circuit, Wi-Fi module, and Bluetooth module, which will not be described in detail in this embodiment.

[0076] This application discloses a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements any of the learning disability identification methods disclosed in this application.

[0077] This application discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute any of the learning disability identification methods disclosed in this application.

[0078] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Those skilled in the art should also recognize that the embodiments described in the specification are optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0079] In the various embodiments of this application, it should be understood that the sequence number of each process does not necessarily imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0080] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they can be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0081] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] If the integrated units described above are implemented as software functional units and sold or used as independent products, they can be stored in a computer-accessible memory. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several requests to cause a computer device (which can be a personal computer, server, or network device, specifically a processor in the computer device) to execute some or all of the steps of the methods described in the various embodiments of this application.

[0083] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0084] The foregoing has provided a detailed description of the learning disability identification method, apparatus, and electronic device disclosed in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there may be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for identifying learning disabilities, characterized in that, The method includes: Output the interactive interface; The system obtains the learning disability screening questionnaire answers entered in the interactive interface and the identity information of the child to be identified. The learning disability screening questionnaire is used to collect information on the child to be identified in the following dimensions: the degree of impairment in daily life, the degree of impairment in academic performance, and the degree of learning difficulty corresponding to the three learning disability subtypes. The three learning disability subtypes include: reading disability, writing disability, and mathematics learning disability. Calculate the test scores of the child to be identified in each of the dimensions based on the answer information; The target user information of the child to be identified is input into the target recognition model; the target user information includes the scores of the child to be identified on each of the dimensions and the identity information of the child to be identified. Obtain the recognition result output by the target recognition model; When the identification result indicates that the child to be identified has a learning disability, the subtype classification result is determined based on the cutoff values ​​corresponding to the three learning disability subtypes and the test scores of the child to be identified on each of the dimensions. The target recognition models are all pre-trained and are any one of the following machine learning models: logistic regression model, random forest model, support vector machine or gradient boosting model.

2. The method according to claim 1, characterized in that, The target recognition model is a logistic regression model.

3. The method according to claim 1 or 2, characterized in that, The three learning disability subtypes each correspond to different cutoff values. A test score greater than or equal to the cutoff value is considered positive. The cutoff value for reading difficulties is 36; The cutoff value for writing difficulties is 48; The cutoff value for math learning difficulties is 45.

4. The method according to claim 1, characterized in that, After calculating the test scores of the child to be identified on each of the dimensions based on the answer information, the method further includes: The test scores of the child to be identified in each of the dimensions are output on the interactive interface.

5. The method according to claim 1, characterized in that, Before inputting the target user information of the child to be identified into the target recognition model, the method further includes: To improve the ability to identify children with learning disabilities, a number of initial sample models are trained one by one using a training dataset to obtain a number of converged sample models. The initial sample models are all binary classification machine learning models. The training dataset includes the test scores of several first sample children on each of the dimensions and the identity information of the first sample children. The performance of each converged sample model in identifying children with learning disabilities was tested using a test dataset, and the performance scores of each converged sample model under the evaluation metrics were obtained. The test dataset includes the scores of several second sample children on each dimension and the identity information of the second sample children. The evaluation metrics include one or more of the following metrics: area under the curve, sensitivity, specificity, positive predictive value, negative predictive value, F2 score, and accuracy. The performance scores of each converged sample model under the evaluation index are compared, and the converged sample model with the best ability to identify children with learning disabilities is determined as the target recognition model.

6. The method according to claim 5, characterized in that, The training dataset includes real data collected from real children and generated data; the generated data is generated based on the real data using a synthetic minority class oversampling technique.

7. The method according to claim 1 or 5, characterized in that, The cutoff values ​​for the three learning disability subtypes were obtained through the following steps: The target recognition model was subjected to receiver operating characteristic (ROC) curve analysis. Negative samples from the first and second sample children were used as the control group, and positive samples of learning disability subtypes diagnosed by the hospital were used as the case group. The value corresponding to the maximum value of the Youden Index was used as the cutoff value for each learning disability subtype.

8. A device for identifying children with learning disabilities, characterized in that, The device includes: An interactive unit is used to output an interactive interface; and to obtain the learning disability screening scale response information filled in the interactive interface and the identity information of the child to be identified; the learning disability screening scale is used to collect information of the child to be identified in the following dimensions: the degree of impairment in daily life, the degree of impairment in academic performance, and the degree of learning difficulty corresponding to the three learning disability subtypes; the three learning disability subtypes include: reading disability, writing disability, and mathematics learning disability. A calculation unit is configured to calculate the test scores of the child to be identified in each of the dimensions based on the answer information; A learning disability identification unit is used to input the target user information of the child to be identified into the target identification model; the target user information includes the scores of the child to be identified in each of the dimensions and the identity information of the child to be identified; and to obtain the identification result output by the target identification model. The subtype identification unit is used to determine the subtype classification result based on the cutoff values ​​corresponding to the three learning disability subtypes and the test scores of the child to be identified on each of the dimensions when the identification result indicates that the child to be identified has a learning disability. The target recognition models are all pre-trained and are any one of the following machine learning models: logistic regression model, random forest model, support vector machine or gradient boosting model.

9. The apparatus according to claim 8, characterized in that, The device further includes: The training unit is used to train several initial sample models one by one using a training dataset to obtain several converged sample models, targeting the identification ability of children with learning disabilities. The several initial sample models are all binary classification machine learning models. The training dataset includes the scores of several first sample children on each dimension and the identity information of the first sample children. The evaluation unit is used to test the performance of each converged sample model in identifying children with learning disabilities using a test dataset, and to obtain the performance score of each converged sample model under the evaluation index. The test dataset includes the scores of several second sample children on each dimension and the identity information of the second sample children. The evaluation index includes one or more of the following indices: area under the curve, sensitivity, specificity, positive predictive value, negative predictive value, F2 score, and accuracy. A determining unit is used to compare the performance scores of each of the converged sample models under the evaluation index, and to determine the converged sample model with the best ability to identify children with learning disabilities from among the converged sample models as the target recognition model.

10. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to implement the method as described in any one of claims 1 to 7.

Images