Abstract: This paper focusses on aspects of applied data mining in the context of team handball. It presents an approach to transform the collected data of team handball matches into formats that allow the use of classification and methods to search for association rules. To be able to search for patterns at arbitrary times of matches a concept of a logical clock is introduced, which becomes an essential part of the data preparation. The applied data mining methods are described in detail using RapidMiner processes and their settings. However, the approach is independent of the used data mining tool. Based on the results of the data mining processes, the applicability of data mining techniques in the given context will be discussed. Particularly it will be shown that rule-based results have significant advantages compared to approaches using support vector machines in the given context. The results are also compared based on the logical clock which will show how patterns evolve over time in case of team handball. We will show that the overall prediction accuracy of a model is not the primary concern in the chosen application area. It is rather to discover rules which clearly help to identify the need for action. The concept of time is crucial in this context because rules are less helpful if they are detected when the game is over, and we are at the end of a slippery slope which could have been prevented long before.Abstract: This paper focusses on aspects of applied data mining in the context of team handball. It presents an approach to transform the collected data of team handball matches into formats that allow the use of classification and methods to search for association rules. To be able to search for patterns at arbitrary times of matches a concept of a logical ...Show More
Abstract: We first consider the Multiplicative Error Model (MEM) introduced in financial econometrics by Engle (2002) as a general class of time series model for positive-valued random variables, which are decomposed into the product of their conditional mean and a positive-valued error term. Considering the possibility that the error component of a MEM can be a Weibull distribution and the need for data transformation as a popular remedial measure to stabilize the variance of a data set prior to statistical modeling, this paper investigates the impact of the inverse square root transformation (ISRT) on the mean and variance of a Weibull-distributed error component of a MEM. The mean and variance of the Weibull distribution and those of the inverse square root transformed distribution are calculated for σ=6, 7,.., 99, 100 with the corresponding values of n for which the mean of the untransformed distribution is equal to one. The paper concludes that the inverse square root would yield better results when using MEM with a Weibull-distributed error component and where data transformation is deemed necessary to stabilize the variance of the data set.Abstract: We first consider the Multiplicative Error Model (MEM) introduced in financial econometrics by Engle (2002) as a general class of time series model for positive-valued random variables, which are decomposed into the product of their conditional mean and a positive-valued error term. Considering the possibility that the error component of a MEM can ...Show More