oceansen / Virtualis

A tool to make virtual sensors for data quality assessment

Virtualis

A tool to make virtual sensors for data quality assessment

Dependencies

• Python 3.8
  • matplotlib
  • numpy
  • pandas
  • plotly
  • scikit-learn
  • Tensorflow 2.0
  • pyyaml
  • scipy
  • dvc

Experiment pipeline

Stages:

1. Select: Selects usefule features of raw data into a new dataframes.
2. Transform: Transforms dataset (at the moment add zeros). . Split: Split data set into training and test data.
3. Scale: Scale input data.
4. Sequentialize: Split data into input/output sequences.
5. Train: Train model.
6. Evaluate: Evaluate model.

Usage

Run experiment

All stages are defined in the file dvc.yaml, and the parameters to be used are saved in params.yaml.

To run/reproduce an experiment with any given parameters specified in params.yaml, run:

dvc repro

Change dataset

To run experiments with another dataset, just change the content of Depot/data/raw/CNC_Milling_dataset to the files you want to use.