Tuning of ML models

This repository is useful for tuning machine learning models for both classification and regression using four methods: Bayesian Optimization, Random Search, Grid Search, and Optuna.

Usage

git clone https://github.com/ayjab/tune_ml
cd C:\Users\user\good_stuff_tune
python main.py --method --model --problem --data

Example:

python main.py --method grid_search --model svm --problem classification --data data/iris.csv

The data should be uploaded into the data directory, or you can refer to main.py for specific usage.
The configurations for each model and problem type can be found in the directory named cfg. Feel free to modify the hyperparameters and their ranges or to add a specific model in the models directory, wuth respect to its architecture. A snip of a configuration file:

model:
  name: SVM  
  problem_type: classification  
  parameters:
    C:
      suggest_type: suggest_float
      min: 1e-5
      max: 10
    kernel:
      suggest_type: suggest_categorical
      choices:
        - "sigmoid"
        - "linear"
        - "poly"
        - "rbf"
    gamma:
      suggest_type: suggest_float
      min: 1e-5
      max: 10
optimization:
  n_trials: 10
  cv: 3
  n_iter: 20
  random_state: 42
  n_jobs: -1
  verbose: 5
  timeout: 3600
  show_progress_bar: True