Machine Learning;Anomaly Detection;Exploratory Data Analysis
발행연도
2019
발행기관
한국보건사회연구원
초록
Machine learning is a subfield of the growing research on artificial intelligence (AI). Specifically, it focuses on the development of algorithms and technology that enable computers to learn independently using data. Machine learning, widely regarded as instrumental in advancements in image processing, video and voice recognition, and Internet searches, has proven to be quite effective as a tool for anomaly prediction and detection. Anomaly detection refers to the process by which one finds instances or data, out of a given pool, that diverge from expected patterns. In this study, we define the concept of anomaly detection, after which we apply the anomaly detection methodology to the given sets of health and welfare policy data to perform an exploratory analysis. We discuss the issues involved in the application of anomaly detection and summarize the policy implications. The data subjected to our analysis include the fludeoxyglucose positron emission tomography (FDG-PET) data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), regarding health, and the results of the Elderly Survey 2017.
목차
Ⅰ. Introduction 1
Ⅱ. Conceptualization 7 1. Anomaly Detection: Definition and Conceptualization 9 2. Factors of Anomaly Detection 11
Ⅲ. Exploratory Analysis of Anomaly Detection in Health Data 15 1. Anomaly Detection in FDG-PET Data for Early Diagnosis of Alzheimer’s Disease 17 2. Analysis Overview 18 3. Analysis Outcomes 20 4. Implications 24
Ⅳ. Exploratory Analysis of Anomaly Detection in Welfare Data 27 1. Data Overview 31 2. Defining Anomalies 32 3. Exploratory Data Analysis 33 4. Implications 42