bcocco/openmore

Python libraries (NumPy-based) to perform model order reduction, clustering and data analysis for combustion and aerothermochemistry data

OpenMORe is a collection of Python modules for Model-Order-Reduction, clustering and classification.

Implemented techniques:

Model Order Reduction:

• Principal Component Analysis (PCA)
• Local PCA (LPCA)
• Kernel PCA (KPCA)
• Feature selection via PCA
• Outlier removal via PCA
• Data sampling

Clustering:

• Local PCA (via Vector Quantization, unsupervised)
• FPCA (via conditioning vector, supervised)
• Spectral Clustering (unsupervised)

Utilities:

• Multivariate data preprocessing
• Varimax Rotation
• Clustering solution evaluation
• Fast algorithm for SVD

Requirements: in order to use OpenMORe on your devices, the following requirements must be satisfied:

• Python version >= 3.6
• Numpy must be installed
• Scipy must be installed
• Matplotlib must be installed
• Pandas must be installed
• Latex must be installed (for the plots' labels)

Installation: if the libraries requirements are satisfied, clone or download the repo. After that, go to the OpenMORe folder from your terminal (where the file setup.py is located) and type: python setup.py install.

Test: it is possible to check if the installation process was successful running the tests. To do that, just type:

• python -m unittest tests/test_PCA.py
• python -m unittest tests/test_sampling.py
• python -m unittest tests/test_clustering.py

If the tests have positive response, you should get a message like:

___________________________
Ran 4 tests in 0.113s

OK
___________________________

Use: if the tests are successful, you can now use OpenMORe. In the "examples" folder there are some pre-set cases, organized according to the final purpose (e.g., clustering, dimensionality-reduction, variables-selection, others). They are also fully commented to describe the required dictionary inputs. In the “data/reactive_flow” folder, there is a collection of data (from a CFD simulation of a turbulent reacting jet) to run the examples and test the code, while in "data/dummy_data" you can find relatively simple data sets to test the scripts' functionality. A detailed description of all the classes and functions is available in the detailed documentation and in the source code.

Documentation: the official documentation is available in /OpenMORe/Documentation. It is strongly suggested to read it before using the software.

For any question or problem regarding the code you can write to me at the following address: [email protected]