Omniscence
Omniscence is a multipurpose library for Data Analysis and Preprocessing, focused mainly on assisting the data input pipeline of Machine Learning Projects.
Alpha Version 0.1 is here!
Features:
- Import csv files into Dataframes
- Create Raw Data Distribution Reports
- Perform In-Place Normalization and Filtering
- Perform Data Correlation Heatmap Analysis
- Use Bayesian Gaussian Mixture Model for Unsupervised Classification and Data Grouping.
Installation
Omniscence requires python3,numpy,pandas,seaborn,matplotlib and sklearn to run.
Install all the dependencies from the requirements file in the project.
$ cd Omniscence
$ pip3 install -r requirements.txt
$ pip3 install OmniscenceUsage
import Omniscence as oms
# Initialization
omni_analyzer = oms.OmniAnalyzer(file_dir = The csv file to load,
task = 'classification' / 'regression',
input_size = No. of features without the target,
target = 'the name of the target feature')
# Load data.
omni_analyzer.load()
# Normalize data (optional).
omni_analyzer.normalize('minmax'/'meanstd')
# Report Data.
omni_analyzer.report()
# Create Kernel Density Estimators and plot data distributions.
omni_analyzer.analyze()
# Plot the inter-correlating feature scores in a heatmap
omni_analyzer.heatmap()
# Perform a BIC score search on the best No. of Clusters and then fit a Bayesian Gaussian Mixture Model on the Data performing Unsupervised Classification.
omni_analyzer.unsupervised_classification()
# Denormalize a dataset based on the normalization metrics of the preprocessed dataset
omni_analyzer.denormalize(dataset_to_denormalize)License
Spyros Mouselinos 2018 -MIT
