This master thesis deals with production-related improvements by data mining. The problem is posed by EPCOS OHG, which produces PTC (Positive Temperature Coefficient) components in a multi-stage production process. Sintering is the bottleneck of production and influences the electrical properties (resistance value) of the finished component. In order to obtain the desired resistance, repeated adjustments of the sintering settings are necessary. These setting times lead to a poor utilization of the production capacity and an increase in the lead time. For this reason, a prediction model for the resistance value is established based on the data provided by the company. The data comes from two different production sites which include different production stages and release measurements. The process standard CRISP-DM (Cross Industry Standard Process for Data Mining) provides the framework for this thesis. To analyse data methods like calculating statistical parameters, outlier analysis and analysis of missing data, grouping and dimensionality reduction are applied to the data. For the prediction of the resistance value, methods of classification are used, including decision trees, discriminant analysis, support vector machines, k-nearest neighbor and ensemble methods. The results of the practical part are four prepared and analysed data sets, one each with and without outliers per location. The models are applied to independent test data and their results are compared. The unsatisfactory output is the result of small ratios between the number of attribute values and the number of attributes. Therefore, the implementation within the company won’t be promising. A working model results in 13 % free capacity during the period in question and shortens the lead time. In order to utilise this in the future while using data mining, possibilities to improve the data and the data processing are presented at the end of the thesis.