Please use this identifier to cite or link to this item:
|Title:||Decision Support for Diagnosing Thyroid Diseases Using Machine Learning|
Thyroid Disease Diagnosis
|Publisher:||3rd International Conference on Advances in Computing and Technology (ICACT ‒ 2018), Faculty of Computing and Technology, University of Kelaniya, Sri Lanka.|
|Citation:||Jayamini, W.K.D. and Weerasinghe, H.D. (2018). Decision Support for Diagnosing Thyroid Diseases Using Machine Learning. 3rd International Conference on Advances in Computing and Technology (ICACT ‒ 2018), Faculty of Computing and Technology, University of Kelaniya, Sri Lanka. p8.|
|Abstract:||Diagnosis of thyroid disorders using two machine learning techniques was studied in this research. Multilayer Perceptron Neural Network with Back-propagation algorithm and Random Forest algorithm were the two algorithms used to build the models for classifying the thyroid diagnosis classes; Hyperthyroidism, Hypothyroidism, Normal. Models were developed with different structures by changing the relevant parameters and the outcomes of the developed models were compared with each other. For developing different neural networks, parameters such as the number of hidden layers, number of neurons in hidden layers and learning rates were changed. For developing different random forest models, parameters such as the number of features per tree and the number of trees in forest were changed. Those models were trained and tested using two different datasets of thyroid diagnosis (Dataset 1 and Dataset 2) which have different attributes that are related to diagnosing thyroid diseases. The models were tested using 10-fold cross-validation while the models were compared and evaluated using the measures Accuracy (%), Mean Absolute Error, Root Mean Squared Error, TP rate, FP rate, Precision and Recall. In diagnosing thyroid disease, both the algorithms performed well. Multilayer Perceptron Neural Network with Backpropagation algorithm performed well for Dataset 1 with an accuracy of 96.7442% and Random Forest algorithm performed well for the Dataset 2 with a mean accuracy level of 98.4915%.|
|Appears in Collections:||ICACT 2018|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.