Skip to content

πŸ” Evaluating Sampling Techniques for Healthcare Insurance Fraud Detection in Imbalanced Dataset.

License

Notifications You must be signed in to change notification settings

joanitolopo/eval-sampling-methods

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

4 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

eval-sampling-methods

Pull Requests Welcome GitHub license

This repository contains code for evaluating sampling method, especially for imbalanced dataset problem. You can cite the research paper in below section.

Getting Started

Clone the Repository

To get started with this project, you'll need to clone this repository to your local machine. Open a terminal and run the following command:

git clone https://github.com/joanitolopo/eval-sampling-methods.git

Setting up the Environment

  • Change directory cd to eval folder
cd /eval-sampling-methods
  • Run this command to install the environment. Make sure you have conda or miniconda package
conda env create -f requirements.yml

Usage

To use the provided script and run the application, follow these steps:

  1. Open a terminal or command prompt.

  2. Navigate to the project directory where the script is located:

    cd /path/to/your/project/directory

Command-Line Arguments

You can customize the behavior of the application using the following command-line arguments:

Argument Description Value
--file Specify the path to your CSV file containing data. path
--sampling Choose the sampling method. (default: "ros") ["ros", "rus", "smt", "iht"]
--ratio Set the sampling ratio. (default: 0.5) 0.5
--plot Enable this flag to generate plots. (optional) optional
--custom_param_path Provide the path to a JSON file with custom hyperparameter values to be tuned. (optional) json file

Example Usage

To run the application with custom parameters and generate plots, you can use the following command:

python main.py --file example_data.csv --sampling smote --ratio 0.6 --plot --custom_param_path example_custom_params.json

Research Paper

If you find this code or research useful in your work, we kindly request that you cite the following paper: http://journal.uad.ac.id/index.php/JITEKI/article/view/25929/pdf_182.

@article{JITEKI25929,
	author = {Joanito Agili Lopo and Kristoko Dwi Hartomo},
	title = {Evaluating Sampling Techniques for Healthcare Insurance Fraud Detection in Imbalanced Dataset},
	journal = {Jurnal Ilmiah Teknik Elektro Komputer dan Informatika},
	volume = {9},
	number = {2},
	year = {2023},
	keywords = {Healthcare Insurance; Imbalanced Dataset; Oversampling; XGBoost; Fraud Detection; Undersampling},
	abstract = {Detecting fraud in the healthcare insurance dataset is challenging due to severe class imbalance, where fraud cases are rare compared to non-fraud cases. Various techniques have been applied to address this problem, such as oversampling and undersampling methods. However, there is a lack of comparison and evaluation of these sampling methods. Therefore, the research contribution of this study is to conduct a comprehensive evaluation of the different sampling methods in different class distributions, utilizing multiple evaluation metrics, including π΄π‘ˆπΆπ‘…π‘‚πΆ, 𝐺 βˆ’ π‘šπ‘’π‘Žπ‘›, 𝐹1π‘šπ‘Žπ‘π‘Ÿo, Precision, and Recall. In addition, a model evaluation approach be proposed to address the issue of inconsistent scores in different metrics. This study employs a real-world dataset with the XGBoost algorithm utilized alongside widely used data sampling techniques such as Random Oversampling and Undersampling, SMOTE, and Instance Hardness Threshold. Results indicate that Random Oversampling and Undersampling perform well in the 50% distribution, while SMOTE and Instance Hardness Threshold methods are more effective in the 70% distribution. Instance Hardness Threshold performs best in the 90% distribution. The 70% distribution is more robust with the SMOTE and Instance Hardness Threshold, particularly in the consistent score in different metrics, although they have longer computation times. These models consistently performed well across all evaluation metrics, indicating their ability to generalize to new unseen data in both the minority and majority classes. The study also identifies key features such as costs, diagnosis codes, type of healthcare service, gender, and severity level of diseases, which are important for accurate healthcare insurance fraud detection. These findings could be valuable for healthcare providers to make informed decisions with lower risks. A well-performing fraud detection model ensures the accurate classification of fraud and non-fraud cases. The findings also can be used by healthcare insurance providers to develop more effective fraud detection and prevention strategies.},
	issn = {2338-3070},
	url = {http://journal.uad.ac.id/index.php/JITEKI/article/view/25929},
	pages = {223--238}
}

Releases

No releases published

Packages

No packages published

Languages