The main aim of the thesis was to optimise the metallothermic production of ferromolybdenum at the Treibacher Industrie AG. At the beginning a thermodynamic characterisation of the process took place. Subsequent it had been determined that the batches were not perfectly mixed and therefore investigations were carried out to check the influence of a further intensive batch mixing. The first step was experimental work in laboratory scale. In order to proof experimental data of these investigations, several trials took place at the production unit. Afterwards analyses of additional mixed charges were compared to the characteristic data of the common ferromolybdenum production to quantify the amount of influence on the yield and the properties of slag. Another part of this work was to collect thermodynamic information in order to create a wide range of data sets. Afterwards a neuron network was generated predicting the optimal burden combinations and the main process parameters. With an additional parameter profit it was possible to show the trend of the financial success depending on the burden composition.