Machine Learning Approach to Automated Quality Identification of Human Induced Pluripotent Stem Cell Colony Images
Table 5
Results of classification tree, linear discriminant analysis, multinomial logistic regression, and naïve Bayes variants. True positive rates (%) are given in parentheses and accuracy (%) can be found from the last column.
Method/class
Bad
Good
Semigood
ACC
Classification tree
20 (48.8%)
50 (67.6%)
19 (32.8%)
51.4%
Linear discriminant analysis
19 (46.3%)
35 (47.3%)
16 (27.6%)
40.5%
Multinomial logistic regression
17 (41.5%)
32 (43.2%)
19 (32.8%)
39.3%
Naïve Bayes
16 (39.0%)
61 (82.4%)
14 (24.1%)
52.6%
Naïve Bayes with kernel smoothing density estimation and normal kernel
18 (43.9%)
59 (79.7%)
14 (24.1%)
52.6%
Naïve Bayes with kernel smoothing density estimation and box kernel
12 (29.3%)
56 (75.7%)
11 (19.0%)
45.7%
Naïve Bayes with kernel smoothing density estimation and Epanechnikov kernel
13 (31.7%)
57 (77.0%)
11 (19.0%)
46.8%
Naïve Bayes with kernel smoothing density estimation and triangle kernel
