期刊名称：IAENG International Journal of Computer Science
印刷版ISSN：1819-656X
电子版ISSN：1819-9224
出版年度：2020
卷号：47
期号：1
语种：English
出版社：IAENG - International Association of Engineers
摘要：Thyroid disease is a common high-incidence disease in the field of endocrine. Mastering the disease influence factors plays a vital role in the successful diagnosis of the disease. In this paper, we propose a thyroid disease knowledge discovery and diagnosis framework AR-ANN, which integrates association rule mining and artificial neural network. Two rule generation algorithms (Apriori and Predictive Apriori) are used to investigate the sick and healthy factors which contribute to thyroid disease. These algorithms are also used to select the most frequent features and to reduce the dimensions. After that, we use one of the most classical artificial neural networks, i.e. BP neural network, to diagnose thyroid disease. We use SPSS Statistics to convert the numerical data into nominal data in preprocessing. Analyzing the Top-7 association rules generated by the two algorithms, we know that age and sex are the two most important factors. Thyroid disease have different effects on people of different age intervals, and the elderly from 60 to 90 are the most likely to suffer from thyroid disease. The results also show that 50 to 60 years old is the age interval with the highest recurrence rate of thyroid disease. For gender factor, men have more chance of being free from thyroid disease than women. Thyroid disease knowledge from these rules is used as the attribute input of BP neural network for diagnosing thyroid disease. Two real world thyroid datasets in UCI machine learning repository are applied. The experimental results show that the performance of AR-ANN is better, which also shows the feasibility and practical value of association rule mining algorithm and BP neural network in thyroid disease assistant diagnosis.
