Predictive modelling refers to the development of advanced statistical models to analyse, extract insights and make inferences from large data sets. Statistical tools are used to separate organized patterns from random noise. Predictive modelling is used extensively in the actuarial field and generalised linear models (GLMs) are perhaps the most popular. GLMs are only one aspect of predictive modelling though. In recent times, the amount of data available to actuaries, coupled with better computing power has opened up new opportunities to build more accurate models for predictive purposes. Other, more sophisticated forms of predictive modelling techniques like the use of neural networks, machine learning, classification and regression trees (CART) and statistical clustering can now be employed by actuaries. The areas to which these techniques can be applied are diverse. These range from price setting to improving marketing efficiency.
These techniques however should not be used in isolation. Without an expert knowledge of the underlying business, the results of these models could be misleading.