Predictive-Analysis-and-Fraud-Detection-for-Insurance-Claims

Capstone-Project-on-Insurance-Claims

Overview

Machine learning in insurance claim analysis involves employing algorithms to analyze claim data. These algorithms detect fraudulent claims by identifying irregular patterns and behaviors. They also assess risk levels associated with claims, predict claim severity, and estimate claim frequency based on historical data and relevant factors. By leveraging machine learning, insurers can make data-driven decisions, improving efficiency, reducing costs, and effectively managing risk. This enables insurers to streamline the claims process, detect fraudulent activities, and allocate resources more efficiently, ultimately benefiting both insurance companies and policyholders.

Requirements

• Python 3

• Libraries: NumPy, pandas, Sklearn, etc.

• Jupyter Lab

Datasets Used

• Claims1.csv

• cust_demographics (1).xlsx

• Python Script (Code)

• Ppt presentation

Features

• Data preprocessing: Clean and prepare the transactional data for analysis.

• Supervised learning: Train classification models to classify transactions as fraudulent or legitimate.

• Model evaluation: Assess the performance of the models using relevant metrics such as precision, recall, and F1-score.

Balancing an unbalanced dataset

import matplotlib.pyplot as plt
plt.figure(figsize = (3,3))
Cust_data["fraudulent"].value_counts().plot(kind="bar",color=["red","green"])

import matplotlib.pyplot as plt
fig = plt.figure(figsize = (3,3))
balanced_sample.fraudulent.value_counts().plot(kind='bar', color= ['red','green'])
plt.title('Distribution of data based on the Fraudulent of our balanced dataset')
plt.show()

Result

Model evaluation

Compariosion of All The Accuracy of Each Model

models = ['LogisticRegression', 'DecisionTreeClassifier','RandomForestClassifier', 'KNeighborsClassifier', 'GaussianNB',] 
accuracy_values = [LogisticRegression_Train_Accuracy, DecisionTreeClassifier_Train_Accuracy,RandomForestClassifier_Train_Accuracy, KNeighborsClassifier_Train_Accuracy, GaussianNB_Train_Accuracy] 
plt.figure(figsize=(18, 5))
# # Plot the bar graph
bars = plt.bar (models, accuracy_values, color=['red', 'blue', 'green', 'orange','black'])
#Add accuracy values on top of each bar
plt.bar_label(bars, labels=[f"{acc:.2f}" for acc in accuracy_values])
#Add Labels and title
plt.xlabel('Models')
plt.ylabel('Accuracy')
plt.title('Accuracy Comparison for Different Models')
#Show the plot
plt.show()

Results

Manoj Gaikwad
https://www.linkedin.com/in/manojgaikwad13/
manojgaik2000@gmail.com
[Social Media Links (optional)]

Thank you for reading!