Machine Learning-Enabled Dyslexia Detection from Dytective Gaming Participants Datasets
Disciplines
Artificial Intelligence and Robotics | Computational Neuroscience
Abstract (300 words maximum)
For our project, we specifically focused on Dyslexia, and how we can organize data of Dyslexic patients and non-Dyslexic patients based on data from previous research papers. We developed a Logistic Regression model, which is a model that processes the data points and their sets as numerical values, and the model will use one of the sets as a dependent variable to compare the rest of the data sets to find a correlation between them. This model organized the data from a study that tested the use of games to help diagnose patients with Dyslexia, and the model was tested on how well it was able to identify Dyslexia from the current data provided from the paper due to the paper labeling each point as Dyslexia or not. The model used 80% of the data from the paper as training to understand the correlation between the independent variables to the dependent variable, while 20% of the data was used to test the accuracy of the model. The model was able to output an accuracy of 78.88% of the data from the testing set, and the missing percent could be attributed to certain points with the data due to some of the patients of the study being marked as possibly having Dyslexia. Thus, the use of Machine Learning techniques has provided great results for diagnosing disorders based on the data that is provided to the models which helps reduce the risk of misdiagnosis.
Academic department under which the project should be listed
CCSE - Computer Science
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Machine Learning-Enabled Dyslexia Detection from Dytective Gaming Participants Datasets
For our project, we specifically focused on Dyslexia, and how we can organize data of Dyslexic patients and non-Dyslexic patients based on data from previous research papers. We developed a Logistic Regression model, which is a model that processes the data points and their sets as numerical values, and the model will use one of the sets as a dependent variable to compare the rest of the data sets to find a correlation between them. This model organized the data from a study that tested the use of games to help diagnose patients with Dyslexia, and the model was tested on how well it was able to identify Dyslexia from the current data provided from the paper due to the paper labeling each point as Dyslexia or not. The model used 80% of the data from the paper as training to understand the correlation between the independent variables to the dependent variable, while 20% of the data was used to test the accuracy of the model. The model was able to output an accuracy of 78.88% of the data from the testing set, and the missing percent could be attributed to certain points with the data due to some of the patients of the study being marked as possibly having Dyslexia. Thus, the use of Machine Learning techniques has provided great results for diagnosing disorders based on the data that is provided to the models which helps reduce the risk of misdiagnosis.