Mobile Application to progressively categorize migraines by means of machine learning and neural networks based on previously created dataset as well as individual user input
Location
Auditorium Hallway
Start Date
28-4-2023 11:20 AM
Department
Computer Science
Abstract
Migraine is a neurological disorder that affects millions of people worldwide. Accurate diagnosis of migraine type is crucial for developing an effective treatment plan. In this study, machine learning algorithms were used to predict migraine types and develop a mobile app that allows users to input their data and obtain their migraine type. The dataset used to train the model consisted of 400 records, and the Synthetic Minority Over-sampling Technique (SMOTE) was used to oversample the minority classes. The top features affecting migraine type, including intensity, visual, character, location, and frequency, were identified using the Predictive Power Score (PPS). Cramer's corrected statistic was applied to reduce the correlation between nominal categorical variables, and the Extra Trees Classifier algorithm was used to achieve the best performance. To determine the optimal number of features, Recursive Feature Elimination with Cross-Validation (RFECV) was used, and the model's hyperparameters were fine-tuned using Grid Search CV. The resulting model achieved a training accuracy of 97% and a testing accuracy of 98%. The mobile app developed based on this model has the potential to improve patient outcomes by providing more accurate and personalized treatment plans. By allowing users to obtain their migraine type from a mobile app, they can seek healthcare that is more tailored to their specific needs. The accurate prediction of migraine type can facilitate the development of more effective and personalized treatment plans, which can improve the quality of life for migraine sufferers. The potential benefits of our model extend beyond personalized treatment plans. By accurately predicting migraine types, our model can contribute to a better understanding of the underlying mechanisms of migraines, leading to new insights and treatments. Additionally, our model can help clinicians identify the most effective treatment options for their patients. In conclusion, our study highlights the potential of machine learning algorithms in healthcare and the importance of developing personalized treatment plans for migraine sufferers. The results of our study can be applied to other neurological disorders to provide accurate and personalized treatment plans for patients.
Faculty Sponsor
Xiwei Wang, Northeastern Illinois University
Mobile Application to progressively categorize migraines by means of machine learning and neural networks based on previously created dataset as well as individual user input
Auditorium Hallway
Migraine is a neurological disorder that affects millions of people worldwide. Accurate diagnosis of migraine type is crucial for developing an effective treatment plan. In this study, machine learning algorithms were used to predict migraine types and develop a mobile app that allows users to input their data and obtain their migraine type. The dataset used to train the model consisted of 400 records, and the Synthetic Minority Over-sampling Technique (SMOTE) was used to oversample the minority classes. The top features affecting migraine type, including intensity, visual, character, location, and frequency, were identified using the Predictive Power Score (PPS). Cramer's corrected statistic was applied to reduce the correlation between nominal categorical variables, and the Extra Trees Classifier algorithm was used to achieve the best performance. To determine the optimal number of features, Recursive Feature Elimination with Cross-Validation (RFECV) was used, and the model's hyperparameters were fine-tuned using Grid Search CV. The resulting model achieved a training accuracy of 97% and a testing accuracy of 98%. The mobile app developed based on this model has the potential to improve patient outcomes by providing more accurate and personalized treatment plans. By allowing users to obtain their migraine type from a mobile app, they can seek healthcare that is more tailored to their specific needs. The accurate prediction of migraine type can facilitate the development of more effective and personalized treatment plans, which can improve the quality of life for migraine sufferers. The potential benefits of our model extend beyond personalized treatment plans. By accurately predicting migraine types, our model can contribute to a better understanding of the underlying mechanisms of migraines, leading to new insights and treatments. Additionally, our model can help clinicians identify the most effective treatment options for their patients. In conclusion, our study highlights the potential of machine learning algorithms in healthcare and the importance of developing personalized treatment plans for migraine sufferers. The results of our study can be applied to other neurological disorders to provide accurate and personalized treatment plans for patients.