This presentation is actually focusing on how to handle insurance claims by the way whether it's fake or not .From thousands of data how to make a good algorithm for prediction using machine learning techniques.
1. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Review spam detection and Fraudulent insurance claims
Harsh Data - 2017CS10
Thejajeto Lousa - 2017CS09
Babu Pallam - 2017CS18
MNNIT Allahabad
November 15, 2017
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
2. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Introduction
Introduction
Fraudulent Insurance claims.
Why did we focus on it?
Future enhancements.
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
3. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Literature Review
Literature Review
Fraud detection of insurance claims : Only one approach
ie.Mechine learning.
There are lot of Data Mining Algorithms.
Eg : Naïve Bayes Classifier Algorithm, K Means
Clustering Algorithm, Support Vector Machine
Algorithm, Linear Regression, Logistic
Regression, Artificial Neural Networks, Decision
Trees, Nearest Neighbours
Supervised Learning could be a best approach because ; we
have a big database.
Neural network can do learning and make prediction accurate.
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
4. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Application of Neural network
Application of Neural network
Figure: How it works
Neural network can train the dataset using learning algorithms
like gradient descent[1]
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
5. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Optimize data
Principle Component Analysis
Selecting Principal Components
Optimize data
Mapping to the form that is familiar to algorithms.
Reducing dimensionality using PCA.
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
6. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Optimize data
Principle Component Analysis
Selecting Principal Components
Overview of PCA
The main goal of a PCA analysis is to identify patterns in data.
PCA aims to detect the correlation between variables,by the
way it attempts to reduce the dimensionality.
The desired goal is to reduce the dimensions of a
dd-dimensional dataset by projecting it onto a
(k)(k)-dimensional subspace (where k<dk<d) in order to
increase the computational efficiency while retaining most of
the information.
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
7. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Optimize data
Principle Component Analysis
Selecting Principal Components
Selecting Principal Components
Obtain the Eigenvectors and Eigenvalues from the covariance
matrix or correlation matrix, or perform Singular Vector
Decomposition.
Sort eigenvalues in descending order and choose the kk
eigenvectors that correspond to the kk largest eigenvalues
where kk is the number of dimensions of the new feature
subspace (k≤dk≤d)/. Construct the projection matrix WW
from the selected kk eigenvectors.
Transform the original dataset XX via WW to obtain a
kk-dimensional feature subspace YY.
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
8. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Classification of Dataset
classification of dataset
Auto insurance
Figure: Output of auto insurance claim dataset
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
9. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Classification of Dataset
classification of dataset
Classification of insurance claim dataset
Figure: Output of insurance claim dataset accuracy
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
10. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Future works
Future works
Future is DEEP LEARNING
Reduce False positives
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa
11. Introduction
Literature Review
Neural Network
Building a Algorithm for Spam Classification
Result and Conclusions
Future works
Future works
William H Wolberg, W Nick Street, and OL Mangasarian.
Machine learning techniques to diagnose breast cancer.
Cancer letters, 77(2-3):163–171, 1994.
Review spam detection and Fraudulent insurance claims Harsh Data - 2017CS10 Thejajeto Lousa - 2017CS09 Babu Pa