vgramreg

Exploration of regression models for processing square-wave voltammograms to estimate analyte concentration.

This repository contains all the necessary code to replicate the results in the paper titled "Evaluation of multi-feature machine-learning models for analyzing electrochemical signals for drug monitoring"

Prerequisites

• Python 3.8 or later
• Anaconda or Miniconda (optional, can use a python virtualenv instead)

Installation

1. Create a Python3.9 Environent

This can be done using Conda or a python virtualenv.

Using Conda:

conda create -n vgramreg python=3.9
conda activate vgramreg

Using a python virtualenv:

python3.9 -m venv venv
source venv/bin/activate

2. Install Dependencies

pip install -r requirements.txt

Download Dataset

 wget ...

Extract the file and store the dataset in the vgramreg project root directory as shown below:

.
├── ML1_ML2
│ ├── 2024_02_19_ML1
│ └── 2024_02_22_ML2
├── README.md
├── main.py
├── requirements.txt
└── src

Generate Dataset

In a text editor, open the connfiguration file src/config.py, and change DATASET_PATH so that it is set to the full absolute path location to the ML1_ML2 folder, e.g.:

 DATASET_PATH = '/Users/abc/Desktop/Epilepsey/Code/vgramreg/ML1_ML2'

Save and exit your editor. Then, in a bash session where your current working directory is in the vgramreg project root directory, generate Excel spreadsheets of the raw voltammogram files by running the generate_dataset.py script:

 python src/generate_dataset.py

This code creates three .xlsx files in the respective folders (2024_02_19_ML1 and 2024_02_22_ML2). We will use only one file named extracted_features.xlsx.

ML1_ML2
├── 2024_02_19_ML1
│ ├── dataframe_log_NOrecenter_0.006_0_1.04_0.15_0.17extra_features.xlsx
│ ├── extracted_features.xlsx
│ └── stats_log_NOrecenter_0.006_0_1.04_0.15_0.17extra_features.xlsx
|
└──2024_02_22_ML2
├── dataframe_log_NOrecenter_0.006_0_1.04_0.15_0.17extra_features.xlsx
├── extracted_features.xlsx
└── stats_log_NOrecenter_0.006_0_1.04_0.15_0.17extra_features.xlsx

Running the models

python main.py

The code should generate (up to small differences in p-values due to random sampling), the following text output on stdout:

(venv) sramsey-laptop:vgramreg sramsey$ python main.py
######Data Distribution:#########
Training {0: 50, 16: 51, 8: 47}
Testing {8: 31, 16: 34, 0: 34}
#################################
Linear R2 Score Best Feature ['univariate, std(S)', 'univariate, V_at_max(S)', 'univariate, min(dS/dV)', 'vcenter', 'univariate, area(dS/dV)', 'univariate, max(dS/dV)', 'peak curvature']
Linear Percent Error Best Feature ['univariate, std(S)', 'vcenter', 'univariate, area(dS/dV)', 'peak curvature', 'univariate, area(S)', 'univariate, max(dS/dV) - min(dS/dV)', 'univariate, V_at_max(S)']
****************************************************
KNN R2 Score Best Feature ['univariate, max(dS/dV) - min(dS/dV)', 'univariate, V_at_max(S)']
KNN Percent Error Best Feature ['univariate, std(S)', 'univariate, max(S)', 'univariate, min(dS/dV)']
****************************************************
RF R2 Score Best Feature ['univariate, area(dS/dV)', 'univariate, min(dS/dV)', 'univariate, V_at_min(dS/dV)']
RF Percent Error Best Feature ['univariate, max(dS/dV) - min(dS/dV)', 'univariate, V_at_min(dS/dV)']
****************************************************
GP R2 Score Best Feature ['univariate, std(S)', 'univariate, V_at_max(dS/dV)', 'univariate, V_at_max(S)']
GP Percent Error Best Feature ['univariate, std(S)', 'univariate, V_at_max(dS/dV)', 'univariate, V_at_max(S)']
****************************************************
########Paired Permutation Test##############
Model Comparison Observed Diff Diff mean Diff std p value
0 Linear--------- univariate, std(S) 19.010316 3.612422 2.717097 0.0000
1 Linear--------- KNN 14.965660 3.621785 2.659492 0.0003
2 Linear--------- RF 13.785516 3.562176 2.664851 0.0013
3 Linear--------- GP 10.143973 3.289923 2.442589 0.0115

Project Overview

All the graphs are stored in the folder named Output.

Contents of the Output Folder

1. Graphs:

• Contains images in .png format.
• These images are bar charts comparing different models.

2. Feature Selection List:

• A subfolder containing .xlsx files.
• Each file has performance scores for each step in the feature selection process for each model.