autor:chen youbing contact Email:youbingchenyoubing@163.com

lab:Xiamen university

Description

=============

this a program for feature selection. the main idea is using minimum Redundancy Maximum Relevance and SVM-forward method to features selection. then, we used features selected by previous method to bulid a classification based on random forest.Finally we used an exhausive search the best model we built in different condition. more deatails see our paper:Predicting Hot Spots in protein interfaces based on feature selection using mRMR combining with SVM Forward

How to use it

DEPENDENCIES

this program depends on the following:

• weka-3.7
• jpython
• scikit-learn
• sckit-feature

Usage

• prepare your features file (csv format) ./scripts/mrmr.csv (for feature selection ) ./scripts/raw_data/all_data.csv (for cross validation) ./scripts/raw_data/train_data.csv ./scripts/raw_data/test_data.csv (for independent validation)

• cd ./scripts/

• ./myrun.sh　

• wait for a little time you can find the result files in ./scripts/result_score

• to analyze the all results use the following command:

• python transferTXT2CSV2.py

ATTENTION

Build model with final features you can find in following directory

+

cd ./scripts/check/feature_select

Reference

Predicting Hot Spots in Protein Interfaces Based on Feature Selection using Mrmr Combining with SVM Forward