wildboar

Added

• Regression shapelet trees
  A new type of shapelet trees has been added.
  wildboar.tree.ShapeletTreeRegressor which allows for constructing shapelet
  trees used to predict real value outputs.

• Regression shapelet forest
  A new tyoe of shapelet forest has been added.
  wildboar.ensemble.ShapeletForestRegressor which allows for constructing
  shapelet forests for predicting real value outputs.

Fixed

• a6f656d Fix bug for strided labels not correctly accounted for
• 321a04d Remove unused property unscaled_threshold

Information

• Stability: beta
• Release data: 2019-05-23

Installation

• pip install -r requirements.txt
• pip setup.py install

Usage

from wildboar.ensemble import ShapeletForestClassifier
c = ShapeletForestClassifier(n_estimators=100, metric="scaled_euclidean")
c.fit(X, y)
c.predict(X_test)

Changelog

• major: add regression shapelet tree: ShapeletTreeRegressor and ShapleletForestRegressor
• fix: striding bug with labels
• fix: bug #1
• fix: issue #2
• fix: remove unused property scaled_threshold

• First public release
• Added: ShapeletTreeClassifier - a generic shapelet tree implementation for constructing random shapelet trees
• Added: Distance computations (for sliding distances):
  • Fast Euclidean distance
  • Fast z-normalized Euclidean distance
  • Fast bounded and z-normalized dynamic time warping distance (with all optimizations from the UCRSuite)
• Added: API for finding the best matching positions and the positions within a given threshold