Classification of Thyroid disease using Machine learning Techniques
Machine learning is one of the prominent implementations on a wide range of applications on which our lives are dependent. Progressively the machine learning algorithm decision importance is increasing in the biomedical and health related fields. The data in these fields are very sensitive to error and if any security breach happens, the results are changed and could lead to a serious damage. The dependency on the machine learning algorithm results are increasing. Mostly, we deal with the supervised learning algorithms to evaluate the results. In some recent researches, it is found that the results are being compromised as the training dataset are altered by adding some malicious data. This derives a new class of attacks called poisoning attack or causative attack. In this paper we have proposed a countermeasure for the causative attacks. It may be on the fixed dataset and can also be on the evolved dataset. The attackers might add the malicious data to the training dataset or then make the algorithm to compromise on the result with less efficiency. The proposed algorithm has 96% efficiency and could drop the malicious data from the training data set and could maintain the best accuracy in giving the best results.