Towards the development of an intelligent inventory management system

This paper is concerned with the development of an intelligent inventory management system which aims at bridging the substantial gap between the theory and the practice of inventory management. The proposed system attempts to achieve this by providing automatic demand and lead time pattern identification and model selection facilities. The process of demand pattern identification together with the statistical tests used is discussed. The models incorporated cover deterministic demand models including: constant, quasi-constant, trended and seasonal demand as well as stochastic demand models. This paper includes an empirical evaluation of the system on real data from the manufacturing and airline industries which shows that this system can lead to significant savings in inventory cost. the user has provided all the parameters necessary for selecting models. This makes the system just like a classification of inventory models by using computers, or at best, like the commercial inventory management packages. Obviously, if such systems cannot select the appropriate inventory model automatically, then they are also unable to switch to a new model for an item when its demand pattern has changed. This will have severe drawbacks when managing inventories with several thousand items, which is typical in both manufacturing and service industries. Third, the problem of interfacing such inventory expert systems with existing inventory information systems has been largely neglected despite the fact that 90 per cent of companies (Fleming, 1992) use computers for inventory management purposes. Finally, these published systems seem to be academic exercises. There are no examples given by any of the authors to show the correctness and efficiency of their systems on real data. To overcome the drawbacks of the published systems and in order to achieve an efficient knowledge-based inventory system one should answer the following fundamental questions. First, does the system require responses from the user to questions in order to select a suitable model? Second, how are the parameters used to choose suitable models estimated? Third, which inventory models should be included in the knowledge base? Finally, how should the system communicate with the user and access other management information systems? The responses to these questions cannot be made unless a detailed study is made of inventory modelling, quantitative forecasting, existing knowledgebased inventory systems, and the tools available to develop a KBS. The main objective of this research is to develop such a system which has an appropriate knowledge scope and focuses on the interrogation of the historical data rather than on asking the user to describe the system under analysis. In addition, the tools used to develop the system must be compatible with the most popular software to allow the proposed system to communicate with other information systems. This study has attempted to address the above questions in order to develop a system which can offer a new approach to solving the inventory management problem. Outline structure of the proposed inventory expert system In general, an expert system is designed to use knowledge and inference procedures to solve problems that are difficult enough to require significant human expertise for their solutions. It is distinguished from other types of computer-based information systems by employing knowledge of the techniques, information, heuristics, models, and problem-solving processes that human experts use to solve such problems. There are three main phases in developing an expert system: knowledge acquisition, knowledge representation, and knowledge implementation. The methodology of the research presented in this paper is the practical development of the software called knowledge-based inventory management system, and its evaluation using real data (Liang, 1997). The integration approach, particularly adapting and incorporating a pattern identifying component and rule base component into a unified system to integrate the data collection, parameter estimation, model selection and order decision functions is the central idea behind the system developed in this project. This makes the system more applicable because it greatly reduces its reliance on the user. Building such an efficient knowledge-based inventory system requires a coherent strategy of combining the computer technology with quantitative methods. The structure of the system is outlined in Figure 1. The user interface of the system developed in this study includes a top-level menu (Figure 2), the dialogue boxes, alert text, confirmation text, and help information facilities. The appearance of the user interface of this system is highly graphical and the menus, commands, and dialogue boxes are visually the same as other Windows applications. The data manager manipulates the historical demand data and other useful information. The operations of the data manager are classified into two categories. The first category of commands performs the general operations on data files such as creating a data file, deleting a data file, and renaming a data file. The second category of commands manipulates the records of the data files by carrying out the following actions: 1 adding a new item; 2 modifying an existing item; 3 deleting an existing item; and 4 displaying an item. [ 355 ] Khairy A.H. Kobbacy and Yansong Liang Towards the development of an intelligent inventory management system Integrated Manufacturing Systems 10/6 [1999] 354±366