Talib Ghanim Muslim, Ögr. Üyesi Yusuf Erkan Yenice
Hospitals are reporting various sorts of thyroid disorder. In this paper, we used a big dataset which involves a test records of 2800 subjects along with the subject diagnoses. Machine learning approach such as Feed Forward Neural Network is used to learn over the data and hence to predict the subject diagnosis. Three techniques are developed over this study; namely: plain model, weight freezing model and Particle Swarm Algorithm model. The last model is outperformed as PSO algorithm is proven a noteworthy performance in optimizing the training process of the neural network. However, performance of this model is studied in terms of accuracy, MSE, MAE, RMSE and time and epochs. However, the Feed Forward Neural Network optimized by Particle Swarm Algorithm model is realized with optimum efficiency as prediction accuracy of 89.4 % is observed.
ML, Metrics, Learning, Training, Testing, Accuracy.