Discovery science (also known as discovery-based science) is a scientific methodology which emphasizes analysis of large volumes of experimental data with the goal of finding new patterns or correlations, leading to hypothesis formation and other scientific methodologies.
Discovery-based methodologies are often viewed in contrast to traditional scientific practice, where hypotheses are formed before close examination of experimental data. However, from a philosophical perspective where all or most of the observable "low hanging fruit" has already been plucked, examining the phenomenological world more closely than the senses alone (even augmented senses, e.g. via microscopes, telescopes, bifocals etc.) opens a new source of knowledge for hypothesis formation. This process is also known as inductive reasoning or the use of specific observations to make generalizations.
Data mining is the most common tool used in discovery science, and is applied to data from diverse fields of study such as DNA analysis, climate modeling, nuclear reaction modeling, and others.
The use of data mining in discovery science follows a general trend of increasing use of computers and computational theory in all fields of science. Further following this trend, the cutting edge of data mining employs specialized machine learning algorithms for automated hypothesis forming and automated theorem proving.