[Dissertation] Inventory Stock Prediction and Deep Anomaly Detection
Masters dissertation by Khutso Kgabo Sepuru, Faculty of Engineering, Built Environment and Information Technology University of Pretoria, Pretoria
Members
Khutso Kgabo Sepuru, MITC Big Data Science
Supervisor(s)
Prof. Vukosi Marivate
Abstract
Firms that have to source raw materials and use the raw materials to assemble a final product have to ensure that this raw material is available at all times along the production process. In reality this is often not the case. Material supply is often subject to two forces, that is, oversupply or under supply due to various reasons ranging from incorrect planning or inconsistent material consumption. Thus, a need for a tool to detect anomalies in these processes is required. This dissertation provides details of the methods proposed to be used to aid in curtailing the aforementioned problems through the use of deep learning prediction and anomaly detection algorithms. Multiple naive models that serves as baseline models are implemented and compared against multiple deep learning algorithms which proved to perform better with a low RMSE score.