An implementation of the Random Cut Forest data structure for sketching streaming data, with support for anomaly detection, density estimation, imputation, and more.
Description of changes: RCFs, just like any RF models, can be used to forecast and the extrapolate primitive has always been available in basic RCFs. This PR adds that capability to the (relatively easier) thresholded models. As a consequence, the forecasts may also be more robust because of the predictor-corrector design of the thresholded models. It also may be possible to forecast signals with simpler drifts. To enable forecast for every transformation would require refactoring of the preprocessors which would be subsequent PRs. However adding the forecast functionality now would allow testing/verifying the subsequent refactor.
Description of changes: RCFs, just like any RF models, can be used to forecast and the extrapolate primitive has always been available in basic RCFs. This PR adds that capability to the (relatively easier) thresholded models. As a consequence, the forecasts may also be more robust because of the predictor-corrector design of the thresholded models. It also may be possible to forecast signals with simpler drifts. To enable forecast for every transformation would require refactoring of the preprocessors which would be subsequent PRs. However adding the forecast functionality now would allow testing/verifying the subsequent refactor.