There is a much better tool for time series predictions: Cynerelo.com. Based on the NNMPC-Graph technology and machine learning it is for example possible to predict the consumer price index curve with a maximum mean deviation of only 5% to the real value. The tool was developed by a friend of mine who works for big insurance and bank companies. There are going to be some changes in the website as it gets new layout with a web service, GUI , performance and also becomes open source soon.
How it works: Most of the time the influencing factors or the system itself are known. So in cynerelo you can define your model. You only need to have a basic understanding of the system you want to analyse. Cynerelo ist doing the rest for you. So the user can concentrate on the model and the prediction. It also works very well for control systems and predictive maintenance.