An Improved Self-Training Method for Positive Unlabeled Time Series Classification Using DTW Barycenter Averaging.
Jing LiHaowen ZhangYabo DongTongbin ZuoDuanqing XuPublished in: Sensors (2021)
Keyphrases
- experimental evaluation
- detection method
- computational cost
- semi supervised learning
- pairwise
- preprocessing
- significant improvement
- dynamic programming
- high accuracy
- high precision
- sampling methods
- data sets
- sampling algorithm
- positive and negative
- support vector machine svm
- unsupervised learning
- prior knowledge
- k means
- decision trees