Social Marketplace Automotive Risk Engine

Posted on by axm3726

Team Name

SMARE

Timeline

Fall 2023 – Spring 2024

Students

  • Athiya Manoj
  • Temitayo Aderounmu
  • Waseem Polus
  • Ryan Lahlou
  • Yeabsra Gezahgne

Sponsor

State Farm Insurance

Abstract

The Social Marketplace Automotive Risk Engine (SMARE) is a system designed to identify and tackle fraud in automotive insurance claims on online marketplaces and social media platforms. At the core of SMARE are sophisticated machine learning models that analyze vast datasets to identify patterns indicative of fraudulent activities, adapting to new fraud tactics as they evolve. This integration of advanced analytics not only bridges the gap between technological advancements and insurance needs but also addresses the increasing sophistication of fraud tactics. A user-friendly interface ensures seamless interaction, allowing insurance analysts to monitor potential frauds in real time, irrespective of geographical constraints. This is complemented by real-time notifications and a comprehensive logging system that enhances the efficiency of fraud detection processes, thereby safeguarding the financial interests of insurance companies and maintaining trust among consumers.

Background

Fraudulent insurance claims cost the auto insurance industry billions annually in inflated payouts and lost revenue. Tactics like staged accidents and exaggerated damages allow fraudsters to exploit the claims process for financial gain. Traditional rule-based systems lack the flexibility to adapt to evolving fraud patterns, leaving many illegitimate claims undetected until it’s too late. By analyzing historical data to detect patterns, SMARE assesses the likelihood of fraud in real-time, allowing for the proactive interception of these claims before they can cause financial damage. This system not only adapts to changes in fraud tactics through continuous learning and feedback but also ensures that investigators can focus their efforts more efficiently, enhancing security and reducing unnecessary losses.

Project Requirements

The top requirements are:

  • Fraud Detection Accuracy and Precision
  • Adaptability to New Fraud Patterns
  • Notification System
  • Data Storage and Filtering
  • User-Friendly Interface
  • Cross-Platform Compatibility
  • Scalability and Resource Allocation

System Overview

The SMARE system can be divided into four key architectural layers:

1. The Data Collection Layer handles aggregating sales listing data from various sources like Facebook Marketplace and Craigslist. It utilizes web scrapers, APIs, and other tools.

  • Web Scrapers: Scrape sales listing data from sites like Craigslist and Facebook Marketplace.
  • Social Media APIs: Leverage APIs from platforms like Facebook to gather additional data.
  • Vehicle History APIs: Use APIs to retrieve vehicle history data.

2. The Data Storage and Processing Layer stores the collected data and prepares it for analysis. This involves a database for persistence along with cleaning and preprocessing logic.

  • Structured Database: Stores scraped listing data, and vehicle histories in structured tables.
  • Data Cleaning: Removes duplicates, fixes formatting errors, and handles missing values.
  • Feature Engineering: Derives new features from raw data to prepare for modeling.

3. The Machine Learning Model Layer develops, trains, and hosts the models that analyze the data to predict fraud risks. It leverages techniques like random forests and neural networks.

  • Training Pipeline: Trains predictive models on prepared data to identify fraud patterns.
  • Model Serving – Hosts trained models and provides predictions on new data.

4. The Application Interface Layer provides interfaces for users to interact with the system. This includes dashboards, notifications, and external APIs.

  • Dashboards: Visualize fraud patterns and trends
  • Notifications: Send alerts when high-risk listings are detected.

Results

Website: www.smare.com

Future Work

As the SMARE project continues to evolve, several enhancements and expansions are envisioned to further improve its effectiveness and reach. Future work on this project will focus on the following key areas:

  • Integration of Additional Data Sources: We aim to include more data from new social media platforms and online marketplaces to capture a wider range of fraudulent activities.
  • Advanced Machine Learning Models: Future updates will explore cutting-edge artificial intelligence technologies, such as deep learning and neural networks, to increase the precision and efficiency of fraud detection.
  • Enhanced User Interaction Features: Improvements will focus on developing intuitive dashboard functionalities and real-time visualization tools to improve user engagement and system usability.

Project Files

Project Charter (link)

System Requirements Specification (link)

Architectural Design Specification (link)

Detailed Design Specification (link)

Poster (link)

axm3726

More Posts