Skip to content

Latest commit

 

History

History
56 lines (50 loc) · 2.14 KB

README.md

File metadata and controls

56 lines (50 loc) · 2.14 KB
Raw

Stochastic logistic sigmoid regression

SLR

Architecture

INSTALLATION

  • sklearn
  • numpy
  • cvxopt
  • pandas
  • Run on command: 
    pip install -r requirements.txt

DATASETS

Download datasets : Link

TRAINING

  • K-Means clustering algorithm: 
    python test.py --state train --path_data <path to dataset> --n_clus <number of cluters> --alpha <alpha> --option Kmean --threshold <not default> --max_loop<max training loop>
  • Approximate q-quantiles algorithm: 
    python test.py --state train --path_data <path to dataset> --n_clus <number of cluters> --alpha <alpha> --option Quantile --threshold <not default> --max_loop<max training loop>

EVALUATION

  • Evaluate accuracy, f1 score, precision score, recall socre with number of clusters (level-quantiles) and time training: 
    python test.py --state eval --path_data <path to dataset> --alpha <alpha> --min_iters <start iter> --max_iters <end iter>
  • Beside, we also provide custom procedure for evaluating accuracy, f1 score, precision score, recall socre with alpha in file model.py. That is 
    def weight_eval_alpha()

ISSUES

  • Data in together group K-Mean (level-quantiles) is a element. You'll see this Warning, to disable that Warning you can add -W ignore before function.py
    Architecture

RESULTS

  • Framingham Heart Disease datasets
    acc-h f1-h pre-h re-h
  • German Credit datasets acc-g f1-g pre-g re-g

TIME TRAINING

  • Framingham Heart Disease datasets
    time-h
  • German Credit datasets time-g

NOTE

If you don't select threshold, output model is accuracy of original logistic regression which is provided by sklearn library

python test.py --state train --path_data <path to dataset> --n_clus <number of cluters> --alpha <alpha> --option Kmean(or Quantile)