Analytical customer relationship management for insurance policy prospects

One of the more complex tasks, prediction often involves deriving patterns from a training set, thus building a model of the population’s behavior that incorporates predictor variables and dependable variables. The model is then fed with new predictor data to produce estimations of the dependable variables. In the strict sense of the word predicting would involve a time difference between the predictor variables (now or in the past) and the dependable variables (in future). However, in statistics or data mining ‘predicting’ is also often used for the estimation of unknown dependable variables at present or in the past. Our first example is based on the known responses of a test set of customers and tries to predict the set of prospects who are most likely to buy a caravan insurance policy. The next example is of a more complex nature, the authors investigate the existence of predictable sequences of events in the financial markets.