Hongyuan Wang, PhD & MAS, director of proactive analysis, data and analytics, employers

Hongyuan Wang, PhD & MAS, director of proactive analysis, data and analytics, employers

Proactive Analysis (PA) is the process by which data is converted into business decisions and then converted into a result. PA can be achieved using traditional statistical predictive models or advanced machine learning models that are actively used in all major industries. PA is a complex process that starts with the original business idea and progresses through data preparation, model creation, predictive model development and validation, model implementation, and model results monitoring. A very good installed model that cannot be applied correctly or that does not materialize for several reasons is not a successful PA application. This article focuses on PA applications and some of the challenges of using them in the insurance industry.

PA can help insurance companies reduce loss-to-expense ratios and grow organically through a variety of applications. For example, by predicting the reimbursement frequency and severity of an individual policy, targeting the right policy, defining the potential potential lifetime value in practice, and smoothing operational processes.

Business requirements and understanding of information

The PA application always comes from business needs. The key is to understand what business is required and the historical data associated with it. For example, the sales team wants to know how the retention ratio can affect both growth and earnings, which are two important goals for an insurance company. The insurance team wants to know how the company should price the risks of each reform. Based on such questions, the analysis team must examine and understand the requirements and available information, both internal and external, to establish the PA. Determining the final reliable and consistent data to use for a PA is always a challenge because there are so many data sources and noisy data out there. At the same time, the analytical tools and model methods to be used must be specified. They are critical and directly affect the development and implementation of the model.

Data preparation and development of a proactive model

There are several analysis tools that can be used to prepare data and develop a proactive model. For example, SAS, R or Python. The tool used depends on the company’s IT infrastructure and the coding and modeling of analysts. But one thing is for sure, the tool must be able to connect to all internal databases, be able to import easily accessible external data, be able to handle large data efficiently, and operate smoothly and quickly during model development. Some external information can be very complex, can have different shapes or structures, and requires the use of other specific tools to obtain useful information.

Once the data preparation is done, analysts can continue to use the tool to convert the data and begin model development. Other modeling tools can automate model adjustment and variable selection from many different modeling methods at very high computational speeds, such as the black box modeling tool. One of the challenges is to determine which model methods (e.g., GLM vs. GBM) should be used. Business requirements, regulations, or IT infrastructure can prevent you from using a wide variety of more advanced model methods, but if very high quality and relevant information is available to the PA, the different methods may not make a big difference.

Implementation of the model

It is critical to create an implementation system or API that automatically collects data from different data sources, performs a model scoring process, and integrates with other business systems to make final business decisions based on model points. It can be a real-time process or it can be scheduled for a daily, weekly or monthly run. The biggest challenge is to create a sufficiently large implementation system to integrate future changes to the IT infrastructure (Data Warehouse changes, operating system changes, or cloud environment changes, etc.).

Conclusion

A successful PA application must meet at every stage of the process. Building a consistent and high-quality PA process requires a project manager to have a complete understanding of business, data, modeling, and tools. It is a collaboration of different teams, which can include business, data and analytics, and information technology.

.

LEAVE A REPLY

Please enter your comment!
Please enter your name here