Bioinformatics stuff that I am working on

This is a simple script to extract important features from a dataset via entropy. This script creates multiple decision trees with entropy gain formula. In a project, using this algorithm, I have extracted most important 20 genes from a dataset containing about 60.000 thousand genes. Using those 20 genes commonly used classfication algorithms could identify each sample up to '95 >' accuracy. Any number of labels and features can be used with this script.

Output of the script is a csv file containing features and their total number of occurrences within all created trees.

To run the script, clone the repo with git, or just download it directly.

git clone https://github.com/kcanakdag/Bioinformatics

Install the required packages with pip:

pip3 install -r requirements.txt

After installing the required packages you can test the script by running the following command;

python rf_features.py -h

-i is a str, the input parameter, it is required and should be a path to a .csv dataset.

-min is an int, the amount of minimum samples required to classify a dataset in decision tree. Lower values may lead to overfitting. Default is 10.

-nf is an int, Number of features randomly selected with each decision tree training. So if you have a lot of features, randomly selecting a number of them shortens the time required to create decision trees greatly. Default is 100

-tt is an int, the train-test split size in percentage. Default is 20.

-n is an int, number of trees generated. default is 20

-acc is an int as accuracy cutoff for each tree in percentage. default is 90

-o is a str, the name of the output file-path.

python ./FeatureSelection/RandomForest/rf_feature.py -i ./test_datasets/iris.csv -n 500 -o ./rf_output.csv