v1.7.4: Small performance improvement

Removed unnecessary thread pool analyzing different decision classes.

v1.7.3: Fixed bug in expert rules pruning

• Fixed bug in expert rules pruning.
• Two-sided intervals counted as a single condition instead of two.

Small fixes

v1.7.1

Some unnecessary stuff removed. Documentation update.

v1.7.0 Multiple speed improvements

• Classification:
  • Significantly faster growing (two orders of magnitude for sets with >100k instances), faster pruning,
  • Added approximate mode (approximate_induction parameter). Note: this is an experimental feature - the results may change in future releases.
• Regression:
  • Mean-based growing set as default (few times faster then median, non-significant impact on accuracy).
• Survival:
  • Faster growing and pruning (few fold improvement).

v1.5.2: Fast mean-based regression rules

Changes from the previous release:

• Added fast induction of regression rules based on mean instead of median (mean_based_regression),
• Added boolean parameter for disabling apriori precision control (control_apriori_precision).

v1.4.8: Contrast set updates

• Updated criteria in regression and survival contrast sets.
• Some other fixes.

v1.4.5 - Bugfixes in default parameter values

Bugfixes in default parameters for standard rule induction.

v1.4.4 - Contrast set-related updates

Contrast set-related updates:

• Fixed time measurement.
• Metainduction with multiple minsupp_all values.
• Update of three data sets so they can be properly read in scipy.
• Toy example results updated
• Parameter names consistent with contrast set manuscript.

v1.4.0 - Contrast set mining

Many new features and improvements:

• Induction of contrast sets for clasiffication, regression, and survival data (contrast_attribute tag in XML data set description),
• Penalties for reusing attributes, rewards for covering new examples (penalty_strength and penalty_saturation parameters),
• Complementary conditions (complementary_conditions parameter ),
• Multiple covering passes (max_passes_count parameter),
• Possibility to ignore selected attributes (ignore tag in XML data set description),
• Model can be saved in tabular form with some useful statistics (model_csv tag in XML data set description),
• Parameter min_rule_covered renamed to mincov_new, automatic decision if value is absolute ( >= 1) or relative ( < 1),
• Several other paremeters added (mincov_all, max_neg2pos),
• Improved selection of best condition in growing,
• Very verbose mode.

v1.3.13

Version verification temporarily removed.