Table 2 Capability comparison.
From: Theory and rationale of interpretable all-in-one pattern discovery and disentanglement system
No. | Machine learning tasks | PDD | Unsupervised Learning Algorithm (e.g., K-means) | Supervised Learning Algorithm (e.g., SVM, ANN…) | Association Mining/Classification (DT, Apriori, Association Classification…) |
|---|---|---|---|---|---|
1 | Classification (supervised learning) | YES | NO | YES | YES |
2 | Imbalanced classification (supervised learning) | YES | NO | Partial YES (need sampling/extra processing for dataset) | Partial YES (need sampling/extra processing for dataset) |
3 | Clustering patterns (unsupervised learning) | YES | NO | NO | Partial YES (need extra pattern pruning/pattern clustering) |
4 | Clustering entities (unsupervised learning) | YES | YES | NO | NO |
5 | Interpreting associations with rules or patterns | YES | NO | NO | YES |
6 | Discovery rare patterns | YES | NO | NO | Partial YES (need customized threshold) |
7 | Detecting rare cases, outliers, anomalies | YES | NO | NO | NO |
8 | Detecting mislabels & rectifying classification | YES | NO | NO | NO |
9 | Transparency & explainability of throughput/output | YES | NO | NO | NO |
10 | Knowledge discovery: relating disentangled patterns to known or unknown sources | YES | NO | NO | NO |
11 | Explainable & extendable knowledgebase | YES | NO | NO | NO |
12 | Supporting expert (tracking, monitoring, exporting) | YES | NO | NO | NO |
13 | Feature extraction/selection | Automatic significant AV selection | Need extra process for feature engineering. | Need extra process for feature engineering. | Need extra process for feature engineering. |
