Automated Valuation Modelling: Analysing Mortgage Behavioural Life Profile Models Using Machine Learning Techniques

The main purpose of this research is to study the predictive power of behavioural life profile models for mortgages using machine learning techniques and emerging languages from the same data sets. Based on the results, banks can determine whether the predictive power of the model can be improved regarding estimates of probability of redemption, and probability of internal transfer beyond traditional techniques. Model training will take place using algorithms based on machine learning such as: random forests, extreme gradient, boosting, light gradient boosting, Adaboost, and ExtraTrees. To perform simulations on fast learning and permit testing of hypotheses, the IBM cloud environment and the Watson proven analytical environment will be used, in order to maximize the value derived from the investment and determine the decision on the implementation and modelling strategy for business disciplines. Therefore, these factors could provide a solid basis for the sustainable development of the mortgage market, and the approach in this research is a starting point for identifying the best decisions taken by banking institutions to contribute to the sustainable development of mortgage lending.