FAE
Feature Analysis Explorer (FAE) can help researchers develop a classification model with comparison among different methods. This project was inspired on the Radiomics, and provides a GUI to help analyze the feature matrix, including feature matrix pre-process, model development, and results visualization.
If you publish any work which uses this package, I will appreciate that you could give the following link (https://github.com/salan668/FAE)
Welcome any issues and PR.
Release
The Windows 64 version was release here https://drive.google.com/open?id=1htts7YsfaxKtN1NeDcNU4iksXfjr_XyK
(Alternative link is: https://pan.baidu.com/s/1ha66TajeoT6dA-a4Qdt8fA)
Getting Started
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.
Pre-install
The below modules must be installed first to make the FAE work.
- imbalanced-learn=0.6.2
- matplotlib=3.2.0
- numpy=1.18.1
- pandas=1.0.1
- pdfdocument=3.3
- pillow=7.0.0
- PyQt5=5.14.1
- PyQtGraph=0.10.0
- pyradiomics=3.0
- reportlab=3.5.34
- scikit-learn=0.22.2
- scipy=1.4.1
- seaborn=0.10.0
- statsmodels=0.11.1
Installing
Just clone it by typing in:
git clone https://github.com/salan668/FAE.git
Architecture of Project
- DataContainer
- DataContainer. The structure to contain the data, which also includes methods like saving/loading.
- DataSeparate. Including functions to separate data into training part and testing part.
- Feature Analysis
- DataBalance, Sample the cases to make the binary-labels balance.
- Normalization. Normalize the data to avoid the scale effect of different features.
- DimensionReduction. Reduce the feature dimension, including PCA.
- Classifier. Map the features onto the labels.
- CrossValidation. Estimate model by using cross-validation on the training data set.
- FeatureSelector. Select the sub-features from the feature matrix.
- Pipelines. The class to estimate the model with different feature selected method and classifier.
- Image2Feature
- RadiomicsFeatureExtractor. This class help extract features from image and ROI with batch process. This class should be more "smart" in the future.
- Visulization.
- DrawDoubleLine. This function helps draw double-y plot. e.g. plot accuracy and error against the number of iterations.
- DrawROCList. This function helps draw different ROC curves. AUC will be calculated automaticly and labeled on the legend.
- FeatureRelationship. This function helps draw the distribution of the values of different features. I can only show at most 3 features in one figure.
- FeatureSort. This function helps draw the features and the weights of them on the classifiction model.
- PlotMetricVsFeatureNumber. This function helps draw the AUC / Accuracy / other metrics against the number of chosen features. This can help find the adaptive number of the features.
- Report
- To Generate the report with PDF format.
Document
TODO
Author
License
This project is licensed under the GPL 3.0 License - see the LICENSE.md file for details
Acknowledge
- Contributor:
- Demo data support.
