Feature-maximum-dependency-based fusion diagnosis method for COPD

• Published in: Multimedia tools and applications Vol. 79; no. 21-22; pp. 15191 - 15208
• Main Authors: Fang, Youli, Wang, Hong, Wang, Lutong, Di, Ruitong, Song, Yongqiang
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2020; Springer Nature B.V
• Subjects: Algorithms; Analytics; Chronic obstructive pulmonary disease; Computer Communication Networks; Computer Science; Computer simulation; Data management; Data Structures and Information Theory; Dependence; Diagnosis; Feature extraction; Multimedia Information Systems; Regression analysis; Simulated annealing; Special Purpose and Application-Based Systems; Support vector machines; Multi-dimensional feature; Feature-maximum-dependency; Integrated learning; COPD
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-018-6876-6

Chronic Obstructive Pulmonary Disease (COPD) is a chronic lung disease that causes a progressive decline in respiratory function. COPD has become the fourth most lethal disease in the world, and worldwide deaths continue to become more common as a result of COPD. Therefore, it is important to help doctors diagnose COPD more accurately using big data analytics and effective algorithms. In the past, COPD was mainly studied as follows: applying data to determine the impact of a single feature on the disease, such as the effect of FEV1/FVC (forced expiratory volume in the first second/forced vital capacity), and analyzing a case with simple models, such as logistic regression or a support vector machine. Therefore, there are obviously deficiencies in previous studies. First, the impacts of multi-dimensional features on COPD have not been considered comprehensively. Second, there is no fusion of multiple study methods on the diagnosis and prognosis of COPD. Thus, this paper presents a feature-maximum-dependency-based fusion diagnosis method for COPD. First, the MDF-RS (feature maximum dependency-rough set) algorithm is proposed to extract the optimal combination of multi-dimensional features. Second, the integrated model DSA-SVM (direct search simulated annealing-support vector machine) is presented to classify the disease. Finally, the proposed method is experimentally tested. The results show that the algorithms outperform other classic methods.