MVDA_exploration_tools

Multivariate data analysis (MVDA) exploration tool is a Python library utilizing the scikit-learn library for partial least squares (PLS) and principal components analysis (PCA).

Additionally, an example Jupyter notebook shows how to use the pandas library for row and column labelling prior to PCA and PLS modelling.

Description

One purpose with this code is to name the output loading and score vectors consistent with chemometrics publications and also provide a tools to facilitate plotting of graphs, numbered scatter plots and bar plots based on matplotlib. Another purpose is to deliver a basic example of how to use pandas to keep consistent IDs of rows and columns from the input table(s) to the output results.

Files

• "PCA_and_PLS_by_pandas_12.ipynb" A jupyter notebook that contains an example run using the pandas library for row and column labelling prior to PCA and PLS modelling.

• "PCA_with_pandas_output.ipynb" The PCA part of the previous file splitted out to yield a shorter example.

• "PLS_with_pandas_output.ipynb" A PLS part of the previous combined file splitted out to yield a shorter example.

• "YO-PLS_with_pandas_output.ipynb" Another PLS part put separately to yield a shorter example.

• "MVDA_exploration_tools.py" A python module containing classes based on scikit-learn that provides traditional chemometrics naming conventions for scores and loading vectors from PCA and PLS modelling. A class named Fig based on matplotlib is also present to plot modelling results.

• "data/original_spectra.xlsx" contains spectra in the ultra-violett range 250-450 nm with 2 spectral components varied in an experimental design together with concentration values for the training data for optimal modelling with few observations.

• "data/decimated_spectra.xlsx" is the same spectra heavily decimated on the spectral axis in order to facilitate viewing as tables in the above jupyter notebooks.

Dependencies

• Pandas, and scikit-learn
• Have been tried in Kubuntu 22.04 with Python 3.10 and Windows 10 with Python 3.8 and 3.9, a requirements.txt file exists for the Python 3.10 case.

Installing

• No installation needed

Executing program

• Download the code to a file catalog and run with jupyter lab

Help

This is an alpha version, please use the Github issues mechanism to resolve issues.

Authors

MJosefson, mats.josefson@astrazeneca.com

Version History

• 0.1
  • Initial Release
• 0.1.2
  • Update to fit changes in scikit-learn v. 1.0
• 0.1.6
  • Updates to handle pandas tables also as output
  • Separated jupyter notebook examples showing pandas output
  • More unit tests implemented
• 0.1.8
  • Update to fit changes in scikit-learn v. 1.3
• 0.1.9
  • Switched PCA from sklearn.decomposition.PCA to use the sklearn.decomposition.randomized_svd directly
  • Corrected PCA and PLS modelling and model properties when variables are unit-variance scaled
• 0.1.10
  • Added snv function for use with sklearn.preprocessing.FunctionTransformer to preprocess spectra

License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details