To Generate an Ensemble Model for Women Thyroid Prediction Using Data Mining Techniques

Area of Science:

• Medical Informatics
• Machine Learning
• Data Mining

Background:

• Thyroid symptom identification is crucial for effective treatment.
• Machine learning offers powerful tools for analyzing complex medical datasets.
• Existing methods may lack the precision needed for accurate thyroid diagnosis.

Purpose of the Study:

• To develop and compare two novel ensemble machine learning techniques for thyroid symptom identification.
• To evaluate the performance of Ensemble-I (Stacking) and Ensemble-II (Bagging + Boosting) models.
• To determine the most effective model for accurate thyroid symptom detection.

Main Methods:

• Proposed two ensemble machine learning techniques: Ensemble-I (Stacking) and Ensemble-II (Bagging + Boosting).
• Applied these techniques to a thyroid dataset to mine hidden patterns.
• Conducted comparative experiments evaluating various performance metrics (ROC, MAE, RMSE, RAE, RRSE).

Main Results:

• Ensemble-II consistently outperformed Ensemble-I across all tested metrics.
• Ensemble-II achieved ROC=(98.79), MAE=(0.31), RMSE=(0.05), RAE=(35.89), and RRSE=(52.67).
• Ensemble-I yielded ROC=(98.80), MAE=(0.89), RMSE=(0.21), RAE=(52.78), and RRSE=(83.71).

Conclusions:

• The Ensemble-II model, combining Bagging and Boosting, is the superior method for thyroid symptom identification.
• This approach offers improved accuracy and reduced error rates in diagnosing thyroid conditions.
• The findings suggest a promising direction for enhancing diagnostic tools in endocrinology.