Science is a tool for acquiring knowledge about how the world works. It involves an elaborate dance between observations and theories. A theory is a model or map of a system of events that abstracts those characteristics that the scientist decides are relevant and ignores those characteristics that the scientist decides are not relevant. The scientist develops a theory to try and explain how the system works. Based on the theory, the scientist can make predictions about cause and effect in real world events. The scientist must then devise experiments to test those theories, by recreating and then manipulating the system in specific ways and observing the effects of those manipulations. The scientist must be careful to isolate and identify all variables, so that she can be reasonably certain that the observed effects are truly the result of her manipulations and not of some extraneous source. She must also be careful in how she classifies events; no two events are exactly the same, so she must be reasonably sure that the events she is observing are similar in relevant ways to each other and to her theoretical model. If the observed results of the experiment are consistent with the predictions of the model, then the theory is considered to be a useful one. If the observed results of the experiment are not consistent with the predictions, then the theory is discarded.
In that sense, then, theories cannot ever be proved, only disproved (this is referred to as falsifiability; scientific theories are falsifiable because they can be proven wrong). No matter how much evidence I have that is consistent with a theory, someone could show up tomorrow and conduct an experiment the results of which contradict a prediction of the theory, and the theory must then be re-examined. It is an extremely common misconception among people - including far too many science teachers - that scientific theories are true or real in some absolute sense. Theories are not reality; they are only attempts to describe how events happen in such a way as to help scientists understand those events. A scientific theory is 'good' insofar as it helps us to think about the universe in useful ways, and it is useful insofar as it makes predictions that are borne out through experiment.
If we observe a number of specific events that follow the same pattern, we may induce a general theory from those specific examples and then test that theory under different conditions. The theory follows from the observations. If we come up with a general theory first and then test that theory with specific examples, we may deduce those predictions from the general theory. Some times, the results of 'unsuccessful' experiments can suggest new theories that do a better job of predicting events. Other times, the theories emerge fully formed from the foreheads of scientists. Occasionally, theories are borrowed from other disciplines and applied fresh.
At any time, an experiment can be conducted that contradicts the predictions of a theory, so for that reason no theories are ever proved. They can only be disproved. A theory that successfully predicts ten different results must eventually be discarded if it cannot predict the eleventh result. If someone introduces a new theory that predicts all eleven results, then it replaces the old theory. As such, science is not the search for truth, but a search for theories that are useful and a gradual elimination of theories that are demonstrably false.