MLE is a solid tool for learning parameters of a data mining model. It is a methodlogy which tries to do two things. First, it is a reasonably well-principled way to work out what computation you should be doing when you want to learn some kinds of model from data. Second, it is often fairly computationally tractable. In any case, the important thing is that in order to understand things like polynomial regression, neural nets, mixture models, hidden Markov models and many other things it’s going to really help if you’re happy with MLE.