pysyn-data 0.3.2

This package is for generating synthetic data using 4 models i.e Conditional Genrative Adveserial Networks(CTGAN), Gaussian Mixture Model (GMM), Prinicipal Component Analysis (PCA) and Bayesian Network (BN). It also informs the user which model will work best based on the input data characterisitics.

pysyn-data Package

This package is for generating synthetic data using 4 models i.e Conditional Genrative Adveserial Networks(CTGAN), Gaussian Mixture Model (GMM), Prinicipal Component Analysis (PCA) and Bayesian Network (BN). It also informs the user which model will work best based on the input data characterisitics. This package was created in TAVLAB under guidance of Dr. Tavpritesh Sethi (Computational Biology Lab under Dr Tavpritesh Sethi at IIITD) for Clinical Data Synthesis Project. This is the link to the lab's official webpage.

Package Intallation

For installing the pysyn-data package the following command has to be run:

pip install pysyn-data

Functions Usage

generate_data:

import pysyn_data as psd
syn_data_list= psd.generate_data(df=data, discrete_cols=discrete_cols, models=['bn','ctgan','gmm','pca'],req_size=5000,target_class=target_class,do_PCA=False)

The above function generates synthetic data of size 5000 using the above 4 models i.e BN, CTGAN, GMM and PCA. It also takes input of discrete columns if there are any in the input dataset. User has the option of doing PCA before generating the required synthetic data

predict_best_models:

psd.predict_best_models(df=data,discrete_cols=discrete_cols)

This function shows user the best augmentation model which will work according to the input data. The used metrics are number of discrete columns to number of continuous columns ratio, number of features to number of samples ratio, number of samples and number of features individually.