Skip to content

Latest commit

 

History

History
43 lines (38 loc) · 2.65 KB

ROADMAP.md

File metadata and controls

43 lines (38 loc) · 2.65 KB
Raw

Roadmap

Authored by: Paulito P. Palmes (ppalmes-ibm)

Release v1.0.0 (Base data structures and ML wrappers)

  • Transformer - abstract class with fit and transform interfaces to be overloaded
  • TSLearner <: Transformer - learners for classification/prediction with fit function for training and transform for prediction
  • Baseline <: TSLearner - returns the mode for classification and usually provides the worst case result
  • CaretLearner <: TSLearner - API wrapper to expose caret regression/classification libs
  • SKLearner <: TSLearner - API wrapper to expose scikitlearn regression/classification libs
  • Identity <: Transformer - identity learner (returns mirror image)
  • Imputer <: Transformer - removes missing values
  • Pipeline <: Transformer - iteratively calls fit! and transform! to the set of transformers in the workflow
  • DateValizer <: Transformer - replace missings with medians grouped by datetime period
  • DateValgator <: Transformer - Aggregate values grouped by datetime period

Release v1.0.1 (Matrify TS for ML workflow)

  • Matrifier <: Transformer - transform vector of values into matrix by sliding windows
  • Dateifier <: Transformer - get the date boundaries in the sliding windows to correspond with matrifier output
  • DateValNNer <: Transformer - nearest neighbor replacement of missing data
  • CSVDateValReader <: Transformer - CSV reader
  • CSVDateValWriter <: Transformer - CSV writer

Release v1.0.6 (Ensemble wrappers, multiformat data readers/writers)

  • RandomForest <: TSLearner - RF regression/classification wrapper
  • PrunedTree <: TSLearner - decision tree regression/classification wrapper
  • Adaboost <: TSLearner - Adaboost regression/classification wrapper
  • DataReader <: Transformer - hdf5/feather/jld/csv multiformat reader
  • DataWriter <: Transformer - hdf5/feather/jld/csv multiformat writer
  • Dockerization - dockerized notebook tutorial and dockerized TSML

Release v1.0.7 (TS auto classification, monotonic data detection/normalization)

  • Statifier <: Transformer - scalar stats for data quality characterization
  • MonotonicFilter <: Transformer - convert monotonic data using finite difference operator
  • TSClassifier <: Transformer - automatic classification of TS data type
  • High-level wrapper for CLI automation and interfacing with other programs in the docker/shell
  • Dockerized branch for Kubernetes deployment

Future Work (Higher-level reasoning and integration APIs)

  • XGBoost wrapper
  • API wrapper for KITT interaction
  • API wrapper for E2D interaction
  • Webserver module to receive/process/output data
  • Higher-level API for parameter optimization during prediction/classification