690 likes | 1.92k Views
An Integrated Machine Learning Approach to Stroke Prediction. Presenter: Tsai Tzung Ruei Authors: Aditya Khosla , Yu Cao, Cliff Chiung -Yu Lin, Hsu- Kuang Chiu, Junling Hu, Honglak Lee . 國立雲林科技大學 National Yunlin University of Science and Technology. SIGKDD 2010. Outline. Motivation
E N D
An Integrated Machine Learning Approachto Stroke Prediction Presenter: Tsai TzungRuei Authors: AdityaKhosla, Yu Cao, Cliff Chiung-Yu Lin, Hsu-Kuang Chiu, JunlingHu, Honglak Lee 國立雲林科技大學 National Yunlin University of Science and Technology SIGKDD 2010
Outline • Motivation • Objective • Methodology • Experiments • Conclusion • Comments
Motivation • Most previous prediction models have adopted features (risk factors) that are verified by clinical trials or selected manuallyby medical experts. • In the past, high-performance machine learning algorithms such as SVM and logistic regression were not explored.
Objective • To propose a novelautomatic feature selection algorithm that selects robust features based on our proposed heuristic: conservative mean. • To present a margin-based censored regression algorithm that combines the concept of margin-based classifiers with censored regression to achieve a better concordance indexthan the Cox model.
Methodology • Conservative mean feature selection • To consider the varianceacross different folds along with the average of the prediction performance. • To evaluate the performanceof each feature individually.
Age Methodology • Conservative mean feature selection • Left ventricular mass • Calculated hypertension status VECTOR
Methodology • Learning Algorithms for Prediction • Margin-based Censored Regression SVM
Experiments • Data Imputation • Feature Selection
Experiments • Stroke Prediction
Experiments • Identifying risk factors
Conclusion • Contribution • An extensive evaluation of the problems of data imputation, feature selectionand prediction in medical data, with comparisons against the Cox proportional hazardsmodel. • A novelfeature selection algorithm, Conservative Mean feature selection, that outperforms both L 1 regularized Cox model and L 1 regularized logistic regression on the CHS dataset. • A novel risk prediction algorithm, Margin-based Censored Regression, that outperforms the Cox model giventhe same set of features.
Comments • Advantage • The structure of this paper is very clear. • Drawback • …… • Application • classification