From: Predicting anatomic therapeutic chemical classification codes using tiered learning
3rd level ATC Code Prediction using different initializations and comparison to STP | ||||||||
---|---|---|---|---|---|---|---|---|
Initialization method | Supervised learning | SuperPred | ChEMBL (Perfect 1st level) | STP | ||||
Prediction method | A(3) | CA(3) (N(3)) | A(3) | CA(3) (N(3)) | A(3) | CA(3) (N(3)) | A(3) | CA(3) (N(3)) |
MLP | 63.62% | 68.82% (57.16%) | 63.26% | 70.67% (58.24%) | 72.40% | 83.81% (66.67%) | 72.40% | 73.42% (59.14%) |
Naive Bayes | 56.99% | 50.96% (31.00%) | 66.49% | 68.22% (46.59%) | 69.00% | 78.14% (52.51%) | 57.53% | 60.81% (30.47%) |
Random Forest | 57.35% | 62.60% (48.03%) | 60.93% | 69.06% (54.84%) | 69.00% | 82.20% (62.37%) | 65.23% | 68.46% (50.18%) |
SVM | 64.16% | 67.98% (56.99%) | 67.56% | 72.64% (63.44%) | 76.88% | 86.26% (71.68%) | 64.52% | 70.31% (58.60%) |