There is a lot you can get out of machine learning techniques but it requires substantial effort and trial and error. Another path is to use something like PriceActionLab which I have found to be very useful for my trading. Some will say that these tools only curve fit the past and you certainly need to be aware of the traps you can fall into but there are many techniques that can produce interesting results.
One tool that is often overlooked is Rapidminer which makes it very easy to explore different areas of time series analysis. I use it all the time. A few years ago, I started a thread on a forex forum which explained how you could use Support Vector machines to estimate the daily range of a financial time series. Unlike R, Matlab or scikit learn for python, you can get results without having to program. The details are below including a process diagram.
I have only recently made the jump into automated trading but I am finding the combination of price action datamining and machine learning to offer some significant advantages over trying to build trading systems from traditional indicators.