It this survey the author attempts to present data processing problems arising in high-energy and nuclear physics as an extension of the data mining approach known mostly due to its applications in business, social and medical sciences. The author, known by his rich experience with data handling for many physical experiments, summarizes peculiarities of data streams produced by modern experimental setups, expounds sequential stages of their processing and mathematical methods used to fulfil those stages. Special attention is devoted to applications of artificial neural networks, boosted decision trees and so-called growing neural gas. A few examples from the practice at the Joint Institute for Nuclear Research are given to elucidate these applications.