Computational tools for the exploration of the chemical space and for molecular simulation are today necessary in designing new chemicals and materials. Modeling chemicals and predicting their properties is still challenging in the design cycle. This chapter shortly introduces the methods to create models by using mathematical and computational tools. Predictive models, usually parametric regression models, used classical statistics to understand data and to derive properties. Pattern recognition (PR)Pattern recognition (PR) was then developed to deal with complex data, which cannot be represented as simple numbers, as biological signals or images; they worked on features extracted from data. New algorithms, called classifiers, have introduced the main methods, terminology, and steps for using data to create predictive models. Machine learningMachine learning (ML) (MLMachine learning (ML))Machine learning (ML), initially aiming at making the computer able to solve tasks without explicitly programming them, has proposed new methods, considering different levels of representation and removing some limitations of the classical approaches, as relying on features and on parametric models. The main MLMachine learning (ML) methods are shortly illustrated, including the algorithmic ones and concluding with the connectionist ones, based on neural networksNeural networks (NN) and dealing also with natural language. MLMachine learning (ML) for predictive modelingPredictive modeling can afford many open challenges in materials science.