This project uses Data Science and AI technologies to create a diagnosis process for chronic kidney disease from a dataset of patients. I has been realized at UTBM during the semester of spring 2023 as a part of the UV DS50 taught by J.Gaber and S.Iovleff.
BEN SALAH Youssef BOUKHTAM Mohamed CHENG Ling Yi HOU Xinyu LAATAR Nisrine AUBRY Youri
This project consists in a machine learning prediction pipeline. Each step of the pipeline is represented by one notebook. Steps generate a some files, which are used in the following steps.
File name | Description |
---|---|
EDA | Exploratory Data Analysis of the dataset. Specific to CKD |
data_preparation | Preprocess the dataset into a clean dataset |
hyperparameter_tuning | For a set of learning models, optimize its parameters for the dataset |
feature_selection | Select the best features to use for training the dataset |
hand_feature_selection | Computation of a prediction model made from topic research. Specific to CKD |
model_comparison | Compare the models. Creates various comparison graphs |
The pipeline can be adapted to any dataset. To do so, the file has to be placed in a csv file at path Data/{dataset_name}/original_dataset.csv
. All of the files generated by the pipeline will then be generated in the same directory. By default, the project contains 2 datasets with all of their generated files :
File name | Generated during step | Used for steps |
---|---|---|
original_dataset.csv | not generated by the pipeline | data_preparation, EDA, feature_selection |
clean_dataset.csv | data_preparation | hyperparameter_tuning, model_comparison |
tuned_hyperparameters.json | hyperparameter_tuning | model_comparison |
important_features.json | feature_selection | model_comparison |
lda_dataset.csv | feature_selection | model_comparison |
pca_dataset.csv | feature_selection | model_comparison |
feat_selec_dataset.csv | feature_selection | model_comparison |