Actuarial Service System - Background

Actuarial science is a field based almost entirely on “crunching numbers” and using formulas to make predictions. Actuaries crunch numbers for insurance policy makers in order to predict the risk of of their insurance policies. According to the Bureau of Labor Statistics, employment of actuaries is predicted to go up by 22% between 2016 and 2026. Currently, there are no public software solutions that provide actuarial services.

Objective

Actuaries are paid a median income of above $90K per year. By creating a software that could do the same work as an actuary, insurance companies could then lower their rates to more affordable levels for the general population.

Methodology

The calculations and predictions made by actuaries can be done using Python, R, or MATLAB. The team will continue to look into the best course of action as far as which language and tools will provide the ideal solution. The team intends to further look into the development of this software between March and May of 2018.

Expected Results

The team intends to create a low-cost solution for insurance companies to calculate risk versus cost using machine learning algorithms.

Cost

The cost for this project will most likely be minimal, seeing as the only equipment needed will be a computer and a stable internet connection. The group intends on making full use of free and/or open source software tools for machine learning.

Problem Statement

• Actuaries are too expensive even though they do many probability calculations that computers can do now.
• Actuaries are not efficient enough to meet the demands of growing company and data analytics teams.
• Actuaries need to be updated on the changing technology and adapt to new and different kinds of solutions.

Source Cause Approach

Revision Method

The revision method could not be implemented. Hundreds of firms provide actuarial services for insurance policy makers however, prices are not disclosed and are on a case by case basis. There are no low cost publicly used actuarial service software systems on the market at the moment.

Present State-Desired State Strategy

Present State: Actuaries are paid a high salary to perform mathematical operations and formulate predictions

Desired State: Decrease costs of hiring and paying actuaries

Kepner-Tregoe Analysis

Situation Analysis

Timing: Many insurance companies still employ people to do probability calculations that a computer could perform following a structural model that is not up to date with the current technology.

Trend: Some companies have begun or want to move away from their current model and use systems that would be more efficient and cost effective.

Impact: This system would have a low to medium impact due to the fact that actuarial science is beginning to shift towards computer science anyway. Currently, there are no low-cost machine learning solutions and only standalone statistical models that need to be continuously updated. A machine learning model would provide a large leap ahead in the field.

Problem Analysis

What?

A system that allows us to provide Actuarial Services to companies instantly through a software platform. The system would use data to review the proposed policy and propose potential risks and costs of the policy tailored to individuals.

When?

Our solution will become increasingly popular as more companies focus on massive data collection and data analytics.

Who?

The solution is geared towards insurance companies as a way to increase productivity and reduce cost.

Where?

This is a universal issue for insurance companies. All insurance companies require actuarial services however, we will be focusing on the United States.

Why?

The solution will shift the focus of actuarial science from calculating the mathematical equations to programming the mathematical equations so that greater amounts of data can be processed in a shorter amount of time.

How?

The solution will be created using various machine learning tools and algorithms available to the group. Most machine learning programs are written in Python or Java, so the group will look into which language would be most suited to reach a solution.

Goals

• To create a software that is able to perform mathematical operations
• To create a software that is able to continuously learn from new data
• To be able to market this software to insurance companies
• Low cost
• Easy to use

Specifications

• Minimal bugs
• Iterative, recursive software
• Minimal storage space
• Minimal runtime

Tasks

• Gather data for life expectancy based on age, living area, diseases between others
• Design statistical models like ones that are currently being used
• Design an API to take policy inputs and output probability predictions based on the statistical model
• Design an application that allows users to input policy plans
• Set up a website to advertise / sell the product
• Marketing for the product

Patents:

• Existing Patents
  • Patent US20110071855A1- Statistical modeling technique used to measure changes in risk metrics
  • Patent US20110106572A1- Statistical modeling technique used to measure changes in risk metrics
  • Patent US20030191672A1- System for appraising life insurance and annuities
  • Patent 7329346 - Insurance rating calculation software component architecture (Type: Grant)
  • Patent 20070112605 - Software Support Insurance (SSI) Policy
• None of the patents above include the use of machine learning software, which the team intends on including in the software prototype.

System Model

Process Model