ML-Tools

Prerequisites

• Python 3.x

Installation

Clone the repository and install the dependencies,

git clone https://github.com/6ixGODD/ML-EnsembleHub.git
cd ML-EnsembleHub
pip install -r requirements.txt

Usage

Execute the script with command line options,

python main.py  --data <path_to_data> 
                --cfg <path_to_config> 
                --save-dir <path_to_save_dir> 
                --name <name_of_experiment> 
                --save 
                --plot

in english:

• --data - specify the path to the data
• --cfg - specify the path to the config
• --save-dir - set the directory to save the results
• --name - define the name of the experiment
• --save - enable result saving (metrics, models)
• --plot - enable plot

Configuration

YAML configure the experiment:

shuffle: <bool>
random_state: <int>

preprocessing:
    method: <method_name>/null
    <method_name>:
        <param_name>: <param_value>

classifiers:
    method: 
      - <method1_name>
      - <method2_name>
      - ...
    <method1_name>:
        <param_name>: <param_value>
    <method2_name>:
        <param_name>: <param_value>
    ...

feature_selection:
    method: <method_name>/null
    <method_name>:
        <param_name>: <param_value>

model_selection:
    method: <method_name>
    <method_name>:
        <param_name>: <param_value>

• shuffle - shuffle the data
• random_state - random state
• preprocessing - preprocessing method / disable(null)
• classifiers - list of classifiers
• feature_selection - feature selection method / disable(null)
• model_selection - model selection method

DataSet

The data must be in the following format:

label	feature1	feature2	...
0/1	value1	value2	...
...	...	...	...

Example

Execute the script with command line options,

python main.py  --data data/credit.csv --cfg configs/credit.yml --save-dir output --name credit --save --plot

Result

• metrics：output/<name_of_experiment>/metrics/metrics.csv
• plots：output/<name_of_experiment>/plots/
• models：output/<name_of_experiment>/models/
• log：output/<name_of_experiment>/log.txt

:)