Machining accuracy is critical for the quality and performance of a mechanical product, and the reliability of a multi-axis NC machine tool reflects the ability to reach and maintain the required machining accuracy. The objective of this study is to propose a general methodology that will simultaneously consider geometric errors and thermal-induced errors to allocate the geometric accuracy of components, for improving machining accuracy reliability under certain design requirements. The multi-body system (MBS) theory was applied to develop a comprehensive volumetric error model, showing the coupling relationship between the individual errors of the components of this machine tool and their volumetric accuracy. Additionally, a thermal error model was established based on the neural fuzzy control theory and was compared to the common thermal error modeling method called BP neural network. Based on the traditional cost model and the reliability analysis model, a geometric error-cost model and a geometric error-reliability model were established, taking the weighted function principle into consideration. Then, an allocation approach of the geometric errors, for optimizing total cost (manufacture and QLF) and reliability, subject to the geometrical and operational constraints of the machine tool, was proposed and formulated into a mathematical model, in order to perform the optimization process of accuracy allocation by using the advanced NSGA-II algorithm. A case study was also performed in a five-axis machine tool, and the traditional NSGA algorithm was used for comparison. The optimization results for the five-axis machining center showed that the proposed approach is effective and able to perform the optimization of geometric accuracy and improve the machining accuracy and the reliability of the machine tool.