Search

CN-121980365-A - Learning disorder recognition method and device and electronic equipment

CN121980365ACN 121980365 ACN121980365 ACN 121980365ACN-121980365-A

Abstract

The embodiment of the application discloses a method and a device for identifying learning disorder and electronic equipment, wherein the method comprises the steps of outputting an interactive interface to obtain learning difficulty degree of a child to be identified in daily life damage, performance damage and learning disorder subtype and identity information thereof; after the score of each dimension is calculated, the score and the identity information are input into a pre-trained target recognition model to obtain a preliminary judgment result of whether learning disorder exists or not, when the preliminary judgment is positive, the target user information is judged through subtype cut-off values to obtain subtype classification results, and the target recognition model can be selected from a logistic regression model, a random forest model, a support vector machine or an extreme gradient lifting model. By implementing the application, multidimensional information can be automatically acquired through the electronic scale, the disorder is rapidly and accurately screened and the subtype is refined through the machine learning model, the evaluation efficiency and objectivity are improved, and the method is suitable for large-scale early intervention of schools and medical institutions.

Inventors

  • LI XIUHONG
  • Gong Ranran
  • WANG DAOSEN

Assignees

  • 中山大学

Dates

Publication Date
20260505
Application Date
20251210

Claims (10)

  1. 1. A method of identifying a learning disorder, the method comprising: outputting an interactive interface; The learning obstacle screening table is used for collecting information of the child to be identified in the following dimensions, namely the degree of damage of daily life, the degree of damage of achievement and the degree of learning difficulty respectively corresponding to three learning obstacle subtypes, wherein the three learning obstacle subtypes comprise reading obstacle, writing obstacle and mathematical learning obstacle; Calculating test scores of the children to be identified in each dimension according to the answer information; Inputting target user information of the child to be identified into the target identification model, wherein the target user information comprises scores of the child to be identified in the dimensions and identity information of the child to be identified; Obtaining a recognition result output by the target recognition model; determining subtype classification results according to the cut-off values respectively corresponding to the three learning disorder subtypes and the test scores of the child to be identified in each dimension when the identification results indicate that the child to be identified has learning disorder; The target recognition model is trained in advance and is any one of a logistic regression model, a random forest model, a support vector machine or an extreme gradient lifting model.
  2. 2. The method of claim 1, wherein the object recognition model is a logistic regression model.
  3. 3. The method according to claim 1 or 2, wherein the three learning disorder subtype packages correspond to different cut-off values, respectively, and a test score greater than or equal to the cut-off value indicates positive, wherein: the cut-off value corresponding to the reading disorder is 36; the cut-off value corresponding to the writing disorder is 48; The cutoff value corresponding to the mathematical learning disorder is 45.
  4. 4. The method of claim 1, wherein after said calculating test scores for said child to be identified in each of said dimensions from said answer information, said method further comprises: And outputting the test scores of the children to be identified in the dimensions on the interactive interface.
  5. 5. The method of claim 1, wherein prior to inputting the target user information of the child to be identified into the target recognition model, the method further comprises: Aiming at learning disorder child recognition capability, training a plurality of initialization sample models one by utilizing a training data set to obtain a plurality of convergence sample models, wherein the plurality of initialization sample models are binary classification machine learning models; the performance of each convergence sample model for identifying the learning disorder children is tested by utilizing a test data set to obtain performance scores of each convergence sample model under evaluation indexes, wherein the test data set comprises scores of a plurality of second sample children on each dimension and identity information of the second sample children, and the evaluation indexes comprise one or more indexes including area under a curve, sensitivity, specificity, positive predictive value, negative predictive value, F2 score and accuracy; and comparing performance scores of the convergence sample models under the evaluation indexes, and determining the convergence sample model with the best learning disability child recognition capability from the convergence sample models as a target recognition model.
  6. 6. The method of claim 5, wherein the training data set comprises real data collected from a real child and generated data based on the real data by a synthetic minority class oversampling technique.
  7. 7. The method according to claim 1 or 5, wherein the cut-off values for the three learning disorder subtypes, respectively, are obtained by: and carrying out subject work characteristic (ROC) curve analysis on the target recognition model, taking negative samples in the first sample children and the second sample children as a control group, taking positive samples of the learning disorder subtype diagnosed by a hospital as a case group, and taking a value corresponding to the maximum about dengue index as a cut-off value corresponding to each learning disorder subtype.
  8. 8. An identification device for learning impaired children, the device comprising: the learning obstacle screening table is used for collecting information of the child to be identified in the following dimensions, namely the degree of damage of daily life, the degree of damage of achievement and the degree of difficulty in learning respectively corresponding to three learning obstacle subtypes, wherein the three learning obstacle subtypes comprise reading obstacle, writing obstacle and mathematical learning obstacle; the calculating unit is used for calculating the test scores of the children to be identified in each dimension according to the answer information; The learning obstacle recognition unit is used for inputting target user information of the child to be recognized into the target recognition model, wherein the target user information comprises scores of the child to be recognized in all dimensions and identity information of the child to be recognized, and a recognition result output by the target recognition model is obtained; The subtype identification unit is used for determining subtype classification results according to the cut-off values respectively corresponding to the three learning disorder subtypes and the test scores of the child to be identified in each dimension when the identification results indicate that the child to be identified has learning disorder; The target recognition model is trained in advance and is any one of a logistic regression model, a random forest model, a support vector machine or an extreme gradient lifting model.
  9. 9. The apparatus of claim 8, wherein the apparatus further comprises: The training unit is used for training a plurality of initialized sample models one by utilizing a training data set aiming at learning obstacle child recognition capability to obtain a plurality of converged sample models, wherein the plurality of initialized sample models are binary classification machine learning models; The evaluation unit is used for respectively testing the performance of each convergence sample model for identifying the children with learning disabilities by using a test data set to obtain the performance score of each convergence sample model under evaluation indexes, wherein the test data set comprises scores of a plurality of second sample children on each dimension and identity information of the second sample children, and the evaluation indexes comprise one or more indexes including area under a curve, sensitivity, specificity, positive predictive value, negative predictive value, F2 score and accuracy; And the determining unit is used for comparing the performance scores of the convergent sample models under the evaluation indexes, and determining the convergent sample model with the best learning disorder child recognition capability from the convergent sample models as a target recognition model.
  10. 10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program which, when executed by the processor, causes the processor to implement the method of any one of claims 1 to 7.

Description

Learning disorder recognition method and device and electronic equipment Technical Field The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for identifying learning disabilities, and an electronic device. Background Learning disorders (learning disorder, LD) are a type of neurological disorder, and in the diagnostic and statistical handbook, mental disorders, 5 th edition (DSM-5) specific learning disorders (SPECIFIC LEARNING disorders, SLD) refer to the fact that children at the school age stage continue to have difficulty in reading, writing, calculating or mathematical reasoning skills, and the current academic skills are far below the average level of reading, writing or mathematics required by their culture and language, and have a significant impact on their academic achievement, professional ability or daily life, but the injury is not caused by a mere lack of learning opportunities, mental retardation, acquired brain trauma or disease, but is based on a biological abnormal function during brain development. The prevalence rate of specific learning disorders affecting the fields of reading, written expression, mathematics and the like is 5% -15%, the prevalence rate of adults is about 4%, and the detection rate is in an increasing trend. As a disease affecting the life-long neural development disorder, learning disorder usually brings about a negative influence such as a lower academic achievement, a higher discontinue one's studies rate, a low self-esteem, and the like, and adversely affects children and adults in education, employment, economy, society, emotion, and the like. Unfortunately, the domestic learning disorder prediction tools are deficient, although a small amount of screening scales exist, however, the theoretical basis established by different scales is different, the applicable age ranges are different, the credibility is not satisfactory, and more importantly, subtypes cannot be distinguished according to DSM-5, so that the identification and early intervention of clinical learning disorders are greatly limited. Disclosure of Invention The embodiment of the application discloses a method, a device and electronic equipment for identifying learning disorders, which can acquire identity information of a child to be identified in an electronic scale mode, damage or difficulty degree under different dimensions, automatically judge whether the child to be identified has the learning disorders or not through analysis of a trained machine learning model on the information, and further judge specific learning disorder subclasses. The embodiment of the application discloses a learning disorder identification method, which comprises the following steps: outputting an interactive interface; The learning obstacle screening table is used for collecting information of the child to be identified in the following dimensions, namely the degree of damage of daily life, the degree of damage of achievement and the degree of learning difficulty respectively corresponding to three learning obstacle subtypes, wherein the three learning obstacle subtypes comprise reading obstacle, writing obstacle and mathematical learning obstacle; Calculating test scores of the children to be identified in each dimension according to the answer information; Inputting target user information of the child to be identified into the target identification model, wherein the target user information comprises scores of the child to be identified in the dimensions and identity information of the child to be identified; Obtaining a recognition result output by the target recognition model; determining subtype classification results according to the cut-off values respectively corresponding to the three learning disorder subtypes and the test scores of the child to be identified in each dimension when the identification results indicate that the child to be identified has learning disorder; The target recognition model is trained in advance and is any one of a logistic regression model, a random forest model, a support vector machine or an extreme gradient lifting model. As an alternative embodiment, the object recognition model is a logistic regression model. As an optional implementation manner, the three learning disorder subtype packages respectively correspond to different cut-off values, and a test score greater than or equal to the cut-off value is indicated as positive, wherein: the cut-off value corresponding to the reading disorder is 36; the cut-off value corresponding to the writing disorder is 48; The cutoff value corresponding to the mathematical learning disorder is 45. As an alternative embodiment, after said calculating a test score of said child to be identified in each of said dimensions according to said answer information, said method further comprises: And outputting the test scores of the children to be identified in the dimensions on the int