1:00 PM - 1:20 PM
[4C1-OS-27a-04] Comprehensive Classification for Blood Test Data to Extract Prediction Factor Candidates
Keywords:Data Mining, Machine Learning, Medicine
The electric medical record of hospitals stores a huge amount of blood test results and it is expected to estimate the effect of treatments and the cause of diseases by analyzing the results. We have proposed a method to analyze blood test data using Linear Discriminant Analysis (LDA) with a variable selection by non-parametric test. But this method is inconsistent because LDA assumes the data comes from a normal distribution and non-parametric test assumes that the data do not come from any distributions.
In this paper, we propose a method to comprehensively analyze blood test data using general machine learning methods. This method includes a new sampling method and a new cross-validation method in a stepwise variable selection method without nonparametric test.
We show the experimental results for blood test data and confirm the effectiveness.
In this paper, we propose a method to comprehensively analyze blood test data using general machine learning methods. This method includes a new sampling method and a new cross-validation method in a stepwise variable selection method without nonparametric test.
We show the experimental results for blood test data and confirm the effectiveness.