Health Insurance Claims Overview

A health insurance claim refers to a request for payment or reimbursement for medical services received by an insured individual. Claims are submitted by the insured or their healthcare provider to their insurer for benefits or payments.

A study by Change Healthcare revealed that hospitals in the U.S. submitted $3 trillion in medical claims in 2016, with $262 billion of these claims being denied initially.

Roadmap for Implementation

To achieve this, we will follow these steps using Python and machine learning techniques:

1. Import necessary software libraries.
2. Load and import the dataset.
3. Conduct data analysis and exploration.
4. Select relevant features.
5. Divide the data into training and test sets.
6. Normalize the data.
7. Train the model using the training data.
8. Generate predictions on the test data.
9. Evaluate the model's performance.
10. Draw conclusions from the evaluations.

The Program

Objective: Predict medical insurance costs based on policyholder characteristics.

Import Libraries

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import seaborn as sns

from sklearn.preprocessing import StandardScaler

from sklearn.model_selection import train_test_split

from sklearn.linear_model import LinearRegression

from sklearn.ensemble import RandomForestRegressor

from sklearn.svm import SVR

Loading and Importing the Dataset

# Load data on Google Colab

from google.colab import files

uploaded = files.upload()

# Read the dataset into a DataFrame

insurance = pd.read_csv('insurance.csv')

Data Analysis and Exploration

# Display the first five rows of the dataset

insurance.head()

This dataset, sourced from Kaggle, includes records based on Miri Choi's Medical Cost Personal Datasets. The columns are described as follows:

• age: Age of the policyholder.
• sex: Gender of the policyholder (female, male).
• BMI: Body Mass Index, which indicates body weight relative to height, ideally between 18.5 and 25.
• children: Number of dependents covered by the policy.
• smoker: Smoking status of the policyholder (non-smoker = no, smoker = yes).
• region: Geographic area of the policyholder in the U.S. (northeast, northwest, southeast, southwest).
• charges: Total medical costs billed by health insurance (in USD).

# Show data columns and types

insurance.info()

Summary of the Data

• Total records: 1,338
• Variables: 7
• No missing values present.
• Data types include both categorical and numerical.

Statistical Overview

insurance.describe()

Visualizing Relationships

To better understand the data, we visualize the relationships between age, gender, BMI, number of children, smoking status, and region against claim charges.

# Example visualization for age vs claim charges

fig = plt.figure(figsize=(10, 6))

sns.barplot(y='charges', x='age', data=insurance, ci=None)

plt.xlabel("Age", size=12)

plt.ylabel("Claim Charges ($)", size=12)

Key Observations

• Claim charges typically increase with age.
• Males tend to have slightly higher claim charges.
• Higher BMI correlates with increased claim costs, except for the 40–45 BMI group, where costs dip slightly.
• Claims are notably higher when policies cover two or three children.
• Smokers face significantly higher claim charges, while regional variations are minimal, with the Northwest showing the highest charges.

Data Normalization

# Normalize charges to a range between 0 and 1

column = 'charges'

insurance[column] = (insurance[column] - insurance[column].min()) / (insurance[column].max() - insurance[column].min())

Feature Selection

Categorical data needs to be converted into numerical format using one-hot encoding.

from sklearn import preprocessing

le = preprocessing.LabelEncoder()

insurance["sex"] = le.fit_transform(insurance["sex"])

insurance["smoker"] = le.fit_transform(insurance["smoker"])

insurance["region"] = le.fit_transform(insurance["region"])

We will generate a correlation matrix to evaluate the relationships between features and the target variable (charges).

plt.figure(figsize=(8,6))

sns.heatmap(insurance.corr(), annot=True, fmt='.0%', cmap='Blues')

Data Splitting

We will separate the dataset into features (X) and target variable (Y), then split it into training and testing sets.

x_data = insurance.drop('charges', axis=1)

y_data = insurance['charges']

x_training_data, x_test_data, y_training_data, y_test_data = train_test_split(x_data, y_data, test_size=0.25, random_state=42)

Model Training and Evaluation

We will train three different machine learning models: Linear Regression, Random Forest, and Support Vector Machines, evaluating their performance using R² score and RMSE.

Linear Regression Model

# Linear regression model training

model_lr = LinearRegression()

model_lr.fit(x_training_data, y_training_data)

predictions_lr = model_lr.predict(x_test_data)

Calculate R² Score and RMSE

r2 = r2_score(y_test_data, predictions_lr)

rmse = np.sqrt(np.mean((predictions_lr - y_test_data) ** 2))

Random Forest Model

model_rf = RandomForestRegressor(n_estimators=10, random_state=0)

model_rf.fit(x_training_data, y_training_data)

predictions_rf = model_rf.predict(x_test_data)

Support Vector Machine Model

model_sv = SVR()

model_sv.fit(x_training_data, y_training_data)

predictions_sv = model_sv.predict(x_test_data)

Display Results

models = pd.DataFrame({

'Model': ['Linear Regression', 'Random Forest', 'Support Vector Machine'],

'Accuracy Score': [r2_lr, r2_rf, r2_sv]

})

sns.barplot(x='Accuracy Score', y='Model', data=models)

Conclusion

All models effectively predicted insurance claim charges based on policyholder characteristics, with the Random Forest model achieving the highest accuracy at 82%. Future work could focus on expanding the dataset and refining model hyperparameters for improved predictions.

Thanks for engaging with this article! Feel free to share any thoughts or feedback below.

This video titled "Project 11: Medical Insurance Cost Prediction using Machine Learning with Python" provides a detailed overview of the process and techniques for predicting medical insurance costs using machine learning.

In this second video titled "Medical Insurance Cost Prediction Using Linear Regression | Machine Learning Project 7," the linear regression model is explored in the context of predicting medical insurance expenses.