Estimating manufacturing hours in a job shop manufacturing environment: Aerospace and defense industry application

Especially for the aerospace production companies, besides making accurate cost estimations in the bidding process, it is also important to make cost estimations fast enough. This is important for the companies to maintain their competitiveness and sustainability. In such an environment, companies should choose the best of their cost estimating methodologies they are capable of regarding to the constraints of the proposal and the company. In this dissertation, some research questions arise about the ability of the company for making cost estimations accurate and fast enough. To find the answers for these research questions, a real data set belonging to an aerospace company is used in the models. Three different cost estimating approaches are built according to their level of detail. Artificial neural networks, random forest and linear regression methods are used in each approach and their estimating performances are compared to each other. Since each three cost estimating approaches are different in level of detail, time needed for making the estimation and the accuracy of each approach is also different from each other. Findings of this dissertation is aimed to help the decision maker to choose the right estimating approach and the right estimating method subject to the related constraints. Also, the impact of the data transformation of the manufacturing hours according to the learning curve on the model performance is investigated in this dissertation.