Improving the Quality of Indicator Systems by MoSi – Methodology and Evaluation

Long term business success highly depends to how fast a company reacts on changes in the market situation. Those who want to be successful need relevant, up-to-date, and accurate information. Business or economic decisions rely on indicators. One facet of data quality is the integrity of data. Most of the main business and economic indicators suffer from statistical discrepancies. These indicators are based on non-linear equation systems and are normally not crisp, but random due to measurement errors. Consequently, computation of the corresponding probability distributions is usually not trivial. Handling uncertainty within indicator systems is a major challenge for improved decision making. Different approaches exist for dealing with uncertainty, e. g., Fuzzy set theory and the probabilistic method. The shortcomings of both approaches can be reduced by the use of simulation. As the Gaussian distribution is not closed under all four arithmetic operations, there is the need for Markov Chain Monte Carlo (MCMC) simulation to determine the probability distributions. A combination of data, generated by MCMC simulation, which is based on prior knowledge about a fully specified non-linear, stochastic balance equation system with noisy measurements, is proposed for handling uncertainty within indicator systems. The Metropolis Hastings algorithm enables the use of any computable target probability function. SamPro is the algorithm that implements the MCMC simulation approach for indicator systems. The estimation of unobservable quantities of such models is improved by SamPro and data inconsistencies to the equation system are revealed. MoSi is proposed as a software tool for the modelling of indicator systems as well as their simulation. The implementation of the SamPro algorithm is consequently included in MoSi, as well. MoSi can be used efficiently in the processes of planning, decision making, and controlling.