To predict whether a policy was cancelled or not based on a number of different parameters.
I used a data set involving 10,000 rows of policies with over 70 columns of parameters with information like coverage premiums, policy type, deductibles, state, and year of the house construction. Combined with a two-class neural network algorithm. I split this data 70/30 with 70% feeding into a tune model hyper-parameter module which optimized my algorithm in the best way for this data. The rest of the 30% was used to scored this test model. I then evaluated it giving me an almost 100% accuracy with only one prediction being wrong out of 3,000.