Benchmarking algorithms to detect erroneous label values in regression datasets.

Codes to implement method proposed in Hang Zhou, Jonas Mueller, Mayank Kumar, Jane-Ling Wang and Jing Lei (2023). Detecting Errors in Numerical Data via any Regression Model

This repository is only for intended for scientific purposes. To find errors in your own regression data, you should instead use the official cleanlab library.

Simulation

• simulation: Main folder for the simulation results in Section 5.
  • generat_data.py: Run this code can get the data used in our simulation. The data is stored in Data folder.
  • utils.py: Consists of helper functions to generate data, etc.
  • conformal_atg.py: Conformal methods for autogloun package. One can change the hyperparameter to use different regression regressors.
  • conformal_sklearn.py: Conformal methods for sklearn package. The default setting is the Random Forest regressor.
  • Eva_before_removing.py: To get AUPRC before in Table 3.
  • Eva_after_removing.py: To get AUPRC after in Table 3.

Realdata

• realdata: Main folder for the simulation results in Section 6.
  • data_preprocessing: method to pre-preocess the data
  • dataset: folder to store the datasets
  • eva_all: evaluate the results to get Table 4.
  • utils.py: Consists of helper functions to generate data, etc.
  • modeling: This folder contains all the information and code needed to train models on a specific dataset. Each dataset will have a folder named after itself i.e., airquality_co etc. Each dataset folder has two sub-directories:
  • predictions: It stores predictions of different models considered during training.
  • trained_models: Stores the trained models as required.