Skip to main content
Welcome to this guide on how to generate a synthetic healthcare claims dataset for your project. In this tutorial, you’ll learn how to set up your development environment, create realistic synthetic data with anomalies, and export it to a CSV file for further analysis or model training.

​
Step 1: Installing Required Packages

Begin by creating a new file named requirements.txt in your VS Code editor. Add the following package list to the file: 
mlflow
pandas
numpy
bentoml
Save the file and open your terminal. Then run the command below to install all required packages: 
$ pip3 install -r requirements.txt
Collecting mlflow (from -r requirements.txt (line 1))
  Downloading mlflow-2.18.0-py3-none-any.whl.metadata (29 kB)
Requirement already satisfied: pandas in /home/codespace/.local/lib/python3.12/site-packages (from -r requirements.txt (line 2)) (2.2.3)
Requirement already satisfied: numpy in /home/codespace/.local/lib/python3.12/site-packages (from -r requirements.txt (line 3)) (2.1.1)
Collecting bentoml (from -r requirements.txt (line 4))
  Downloading bentoml-1.3.15-py3-none-any.whl.metadata (16 kB)
Collecting mlflow-skinny==2.18.0 (from mlflow->-r requirements.txt (line 1))
  Downloading mlflow_skinny-2.18.0-py3-none-any.whl.metadata (10 kB)
Downloading mlflow_skinny-2.18.0-py3-none-any.whl (30 kB)
The command above installs all dependencies mentioned in your requirements.txt file. Once the installation is complete, you are ready to move on to generating synthetic data.

​
Step 2: Generating Synthetic Health Claims Data

Create a new Python file named synthetic_health_claims.py. This script generates a synthetic dataset with both normal claims and anomalous claims to simulate outlier events.

​
How the Script Works

  • Imports necessary libraries and sets a random seed for reproducibility.
  • Generates 1,000 normal claim records with fields such as claim_id, claim_amount, num_services, patient_age, provider_id, and days_since_last_claim.
  • Introduces 50 anomalous entries with significantly higher claim amounts and additional service counts.
  • Combines, shuffles, and exports the dataset to a CSV file.

​
Code Implementation

import pandas as pd
import numpy as np

# Set random seed for reproducibility
np.random.seed(42)

# Generate synthetic normal claim data
num_samples = 1000
data = {
    'claim_id': np.arange(1, num_samples + 1),
    'claim_amount': np.random.normal(1000, 250, num_samples),
    'num_services': np.random.randint(1, 10, num_samples),
    'patient_age': np.random.randint(18, 90, num_samples),
    'provider_id': np.random.randint(1, 50, num_samples),
    'days_since_last_claim': np.random.randint(0, 365, num_samples),
}

# Convert normal data to DataFrame
df = pd.DataFrame(data)

# Introduce anomalies (e.g., very high claim amounts)
num_anomalies = 50
anomalies = {
    'claim_id': np.arange(num_samples + 1, num_samples + num_anomalies + 1),
    'claim_amount': np.random.normal(10000, 2500, num_anomalies),  # Much higher amounts
    'num_services': np.random.randint(10, 20, num_anomalies),
    'patient_age': np.random.randint(18, 90, num_anomalies),
    'provider_id': np.random.randint(1, 50, num_anomalies),
    'days_since_last_claim': np.random.randint(0, 365, num_anomalies),
}

# Convert anomalies to DataFrame
df_anomalies = pd.DataFrame(anomalies)

# Combine and shuffle the dataset
df = pd.concat([df, df_anomalies]).reset_index(drop=True)
df = df.sample(frac=1).reset_index(drop=True)

# Save the dataset to CSV
df.to_csv('synthetic_health_claims.csv', index=False)
print("Synthetic data generated and saved to 'synthetic_health_claims.csv'.")
To run the script, execute the following command in your terminal: 
$ python3 synthetic_health_claims.py
Synthetic data generated and saved to 'synthetic_health_claims.csv'.
This script creates a balanced dataset containing both normal and anomalous data points, ideal for training and testing anomaly detection models.

​
Step 3: Preparing Data for Model Experiment

After generating the initial synthetic dataset, you might want to simulate a different testing scenario by modifying the dataset. In this step, we introduce a smaller set of anomalies (5 records) and omit the num_services field to tailor the dataset for a specific model experiment.

​
Code Implementation for Model Experiment

import pandas as pd
import numpy as np

# Assumption: The normal data (df) and the variable num_samples (e.g., num_samples = 1000)
# Introduce a smaller set of anomalies for the experiment
num_anomalies = 5
anomalies = {
    'claim_id': np.arange(num_samples + 1, num_samples + num_anomalies + 1),
    'claim_amount': np.random.normal(10000, 2500, num_anomalies),  # Significantly higher amounts
    'patient_age': np.random.randint(10, 20, num_anomalies),
    'provider_id': np.random.randint(1, 50, num_anomalies),
    'days_since_last_claim': np.random.randint(0, 365, num_anomalies)
}

# Convert new anomalies to a DataFrame
df_anomalies = pd.DataFrame(anomalies)

# Combine with the existing data
df = pd.concat([df, df_anomalies]).reset_index(drop=True)

# Save the updated dataset to CSV
df.to_csv('synthetic_health_claims.csv', index=False)
print("Synthetic data for the experiment generated and saved to 'synthetic_health_claims.csv'.")
Run the updated script with the command below: 
$ python3 synthetic_health_claims.py
Synthetic data for the experiment generated and saved to 'synthetic_health_claims.csv'.
The modified dataset now includes an updated anomaly configuration, which is useful for various experimental setups in model validation.
In this guide, we demonstrated how to create a comprehensive synthetic dataset with both normal and anomalous healthcare claims. The resulting CSV file, synthetic_health_claims.csv, can now be used in your data analysis or machine learning projects. For more insights on data preparation and anomaly detection, explore our related articles and documentation.

Watch Video