A Time Series is an ordered sequence of data points. Typically it's measured at successive times spaced at uniform time intervals. Examples of time series are the annual flow volume of the Nile River at Aswan or the daily value of a stock market index. Time Series of the German Stock Index (DAX) from January 1999 to December 2004 The research area called Time Series Analysis comprises methods for analyzing time series data in order to extract meaningful statistics, rules and patterns. Later on these rules and patterns might be used to build forecasting models that are able to predict future developments. In case we want to predict future trend directions (e.g. up/down) we have to solve a Classification problem. If we try to forecast future time series data points (e.g. the Dow Jones will be at 12000 point at end of the next month) the relevant data mining technique is called Regression. In dDM we use different machine learning algorithms to discover and extract valuable patterns in order to built forecasting models. The necessary algorithms are integrated in the open source data mining framework RapidMiner that is used for the most dDM task. To solve classification problems we use the following machine learning algorithms: Decision Trees, k-nearest Neighbours, Support Vector Machines, Neural Networks. Furthermore, Linear Regression, Support Vector Machines, LeastMedSquare Regression and Logistic Base Regression are used to build Regression models. |