RT-Analyzer

This project revolves around the implementation of predictive models capable of determining the RT of analytes based on various chemical features.

This repository contains a set of scripts for predicting Retention Time (RT) in chromatography data. The scripts are designed to handle data preprocessing, model training, and submission of the prediction.

Installation

To install the RT-Analyzer project, follow these steps:

1. Clone the repository:

git clone https://github.com/your-username/RT-Analyzer.git

2. Install the following dependencies :

• pandas
• numpy
• scikit-learn
• keras
• pytorch
• skorch
• xgboost
• rdkit
• matplotlib

By running:

pip install numpy pandas scikit-learn keras pytorch skorch xgboost rdkit matplotlib

Usage : Kaggle submission

To obtain the predictions submitted on Kaggle, run the following file :

python main.py

Main Script Parameters

The main.py script features the following parameters:

• model paths: Path to the submissions files of individual models ('nn.csv', 'keras.csv', 'gb.csv).
• output path: Path to store the submission file (default: 'submission.csv').
• submit: Flag to indicate whether to generate Kaggle submission
• plot: Flag to enable/disable plotting during model training

Testing Individual Models

To test individual models implemented in this project, follow these steps:

1. Open the relevant script file (e.g., models.py, non_linear_models.py, linear_models.py).

2. Uncomment the code section pertaining to the model of interest.

3. Adjust as needed the model testing parameters

# Model testing parameters

submit = False
plot = True
cddd = False
file_path = 'nn.csv'

4. Run the modified script to evaluate the selected model.

Contribution

@melinacherchali @slne18