Fraud Management is inevitable to insurance companies as this informs the level of risk covered which in turn affects the premiums being charged. Over the years, with evolution of fraud detection methods, perpetrators of fraud have also been evolving their practices to avoid detection. This means that, as insurance premiums are increasing so is the level of fraud.
Insurers need to embrace big data techniques that can quickly and accurately access, link and analyse massively increasing volume, velocity and granular data. Big data helps insurers have a superior understanding of their entire business environment and their client behaviours in all dimensions which informs fraud management techniques.
Machine learning has been a major contributor for detecting different types of financial fraud through its diverse methods. The methodology will discuss in detail the technical approach to be applied to a Kenyan motor insurers data set using modern big data technology such as Hadoop. Fraud-blocking and monitoring capabilities by executing ultra-efficient and multi- dimensional analytics over big data, producing the industry's lowest percentage of false positives. Using machine learning technology, the study adapts to fraudsters' schemes in real time and detects fraudulent transactions at the very first instance. Machine learning techniques allow for improved predictive accuracy, enabling loss control units to achieve higher coverage with low false positive rates.
In this paper, multiple machine learning techniques for fraud detection are presented. The impact of feature engineering, feature selection and parameter tweaking are explored with the objective of achieving superior predictive performance. The deep learning model offered a suite to apply in the insurance claims fraud.