There are some reasons to work on this topic. Medicare doesn’t compensate that much because of the strict compensation policy. Due to the policy constraint, medical report for compensation is required. The size of tumor cannot really show the severity which is contrary to the policies nowadays. And the most expensive treatment, target therapy content, will not be paid. That’s why the cancer insurance products come out – try to complement the gap that Medicate doesn’t cover. However, the general cancer insurance actually can’t include all the details since the treatments of the cancers depend on the type of the cancer. Therefore, patients’ pain points still remain and we should make some efforts on it.

Our business goal is to provide the cancer crude rate forecasts for better insurance products design which helps patients to gain compensation they deserved. Specific cancer may occur in certain gender with greater possibilities. Therefore, we decide to forecast among different genders and choose Top 3 cancer crude rate forecasts in each gender as prototypes for Insurance Companies. Finally, we set our forecasting goal for providing Insurance Companies the forecasts of TOP 3 cancer crude rate each gender for 4 years.

The yearly cancer crude rate dataset that we use comes from the government website. After trying various methods for each series, we include different external information series. However, we didn’t use those series to optimize our forecast models at last because the models’ performance didn’t get well. We choose one best model among Naive, Holt, Regression, Neural Network, and ARIMA. Finally, we generate cancer crude rate forecasts, use empirical rolling-forward method to plot histogram and find out the forecast intervals. We then compare the best model to the ensemble method.

The recommendations for forecasts implementation are stated as four main points.

First, the forecasting models we built are the prototypes. The ensemble model is for generating forecasts for all cancers and we set the best one model as premium package for Insurance Companies. Compare to ensemble model, the performance and the forecasting accuracy of choosing the best one model on each cancer would be higher. Second, Insurance Companies can design better cancer portfolio insurance. By doing so, customers may receive enough compensation they deserved. Third, as long as people know that these Insurance Companies provide the new products which are meet their needs, Insurance Companies can make profits and establish good reputation in the long run because of solving the customer's pain points. Last but not least, Insurance Companies can take some marketing strategies and CSR (Corporate Social Responsibility) actions. For instance, they can promote products to customers with low risk of getting cancer by telling them what factors will cause cancer. The project we proposed can surely benefit both Insurance Companies and customers in the future.