A multi-classifier method based on deep learning approach for diseases classification

Accurate prediction of diseases classification is vital in the medical-related field. Misclassification of diseases would be detrimental as it may lead to the misdiagnosis and wrong treatment of patients. In medical fields, the most important performance measurement indicators for diseases cla...

Full description

Bibliographic Details
Main Author: Rosaida binti Rosly (Author)
Corporate Author: Universiti Sultan Zainal Abidin . Faculty of Informatics and Computing
Format: Thesis Book
Language:English
Subjects:

MARC

LEADER 00000cam a2200000 7i4500
001 0000099260
005 20210107.0
008 200921s2020 my eng
040 |a UniSZA   |e rda 
050 0 0 |a RC258   |b .R67 2020 
090 0 0 |a RC258   |b .R67 2020 
100 0 |a Rosaida binti Rosly ,   |e author 
245 0 0 |a A multi-classifier method based on deep learning approach for diseases classification   |c Rosaida binti Rosly 
264 0 |c 2020 
300 |a xviii,257 leaves ;   |c 31cm. 
336 |a text  |2 rdacontent 
337 |a unmediated  |2 rdamedia 
338 |a volume  |2 rdacarrier 
502 |a Thesis (Degree of Doctor of Philosophy) - Universiti Sultan Zainal Abidin,2020 
504 |a Includes bibliographical references (leaves 192-208) 
505 0 |a 1. Introduction -- 2. Literature review -- 3. Research methodology and proposed framework for disease datasets classification -- 4. The proposed single and multi-classifier classification method -- 5. The proposed of deep multi-classifier learning method (DMCL) -- 6. Conclusion 
520 |a Accurate prediction of diseases classification is vital in the medical-related field. Misclassification of diseases would be detrimental as it may lead to the misdiagnosis and wrong treatment of patients. In medical fields, the most important performance measurement indicators for diseases classification are sensitivity, specificity, and accuracy. By using a single classifier, it is impossible to achieve optimal levels of sensitivity and accuracy. Thus, this research proposed a multi-classifier method based deep learning approach that aims to increase the sensitivity and accuracy of diseases classification. Five disease datasets (Breast Cancer Wisconsin, Hepatitis, Pima Indians Diabetes, Parkinson's, and Indian Liver Patient) have been chosen as an application field to examine the proposed multi-classifier approach in achieving high accuracy. Multiple stages are involved in this research, including the generation of predictive methods, selection of methods using single classification, application of fusion classification between different classifiers via a combination of two or more classifiers, followed by the selection of the fusion output that has the highest accuracy before combining it with other classifiers. Lastly, the combination of prediction class by relevant classifier has been imple~ented with the "deep learning method using deep neural networks approach. The classifiers have been chosen to be combined are Sequential Minimal Optimisation (SMO), Naive Bayes (NB), Random Forest (RF), Decision Tree (148), and instance-based learning with parameter k (IBk). The following classifiers are used here because they are used by most researchers that proven bytheir pre-vious performance. The proposed method lias recorded a higher accuracy than single and multi-classifier methods. The result has proved that it has been improved using the proposed method and shows a comparable result from the previous research. The highest improvements of accuracy for datasets were Breast Cancer Wisconsin (96.63%), Hepatitis (92.50%), Pima Indians Diabetes (80.34%), Parkinson (87.69%), and Indian Liver Patient (74.79%). Based -on five disease datasets studied, the combination of relevant classifiers bas~d on deep learning approach has been found can increase the accuracy of the dataset in classifying the different disease. Mostly the dataset was improved in terms of accuracy using the proposed method than other methods such as single ones and multi-classifier method. Overall, our proposed method better in terms of considering accuracy, sensitivity, and specificity. 
610 2 0 |a Universiti Sultan Zainal Abidin   |v Faculty of Informatics and Computing   |x Dissertations 
650 0 |a x 
710 2 |a Universiti Sultan Zainal Abidin .   |b Faculty of Informatics and Computing 
999 |a 1000180273   |b Thesis   |c Reference   |e Tembila Campus