HEALTHCARE FRAUD DETECTION USING DEEP LEARNING AND ML ALGORITHMS

Name : Aditya Sahani(B22CS003) and Shikhar Dave(B22CH032)

Models Used : Isolation Forest, LocalOutlierFactor and AutoEncoders

Dataset : https://www.kaggle.com/datasets/tamilsel/healthcare-providers-data/code

Colab Link : https://colab.research.google.com/drive/151wlpfzLhi2N5yZ1TipKTq302WTXLm2g#scrollTo=OWzs5eJua6jk&uniqifier=1

This project focuses on detecting anomalies in healthcare data using machine learning techniques. The dataset used for this project is sourced from Kaggle.

Contributors

• Aditya Sahani(B22CS003)
• Shikhar Dave(B22CH032)

Technology and Models Used

• Isolation Forest
• Local Outlier Factor
• Autoencoders

Project Overview

Healthcare fraud detection is a critical area where machine learning and data analysis can play a significant role in identifying fraudulent activities and preventing losses. In this project, we aim to detect anomalies in healthcare provider data using various machine learning models and techniques.

Data Preprocessing

• Loaded the dataset and dropped unnecessary columns.
• Removed null values and duplicated rows.
• Removed commas from the numerical data for consistency.

Exploratory Data Analysis (EDA)

• Conducted exploratory data analysis to gain insights into the data distribution and identify potential anomalies.
• Visualized histograms, density plots, and box plots to analyze the distribution of features.

Isolation Forest

• Used the Isolation Forest algorithm to detect anomalies in the preprocessed dataset.
• Visualized anomalies using histograms, density plots, and scatter plots.

Local Outlier Factor (LOF)

• Implemented the Local Outlier Factor algorithm to identify anomalies based on local density deviation.
• Visualized the relationship between the number of neighbors and the number of outliers.

Autoencoder

• Utilized autoencoders for anomaly detection by reconstructing input data and comparing it with original data.
• Scaled the data using MinMaxScaler for optimal performance.
• Trained the autoencoder model to learn the underlying patterns in the data.
• Determined a threshold for anomaly detection based on reconstruction errors.
• Evaluated the model's performance and counted the detected anomalies.