There are various modelling versions available for predictive engines. We outline the important and wide accepted ones. It is mainly divided into regression techniques and machine learning techniques.
Regression models are the backbone of predictive engines. The main goal is to model and represent the correlation(s) between at least two different variables that are in consideration. The variety of regression techniques is wide. However, not all the techniques are applied at once. Depending on the situation the number can change. Linear regression model, discrete choice model, profit regression, logistic regression, time series models, multinomial logistic regression, multivariate adaptive regression, survival or duration analyses are among the most used ones.
Machine learning techniques are is a processes in which that various statistical methods for regression and classification is are applied to find/define a wide variety of data. Thus, specific practises of machine learning such as deep learning, reinforced learning and forecasting are important functions. The purpose of those specific practises is focused on reinforcing the learning throughout the historical collected data and maximising the cumulative result that supports the precision of the models.