Projects / Programmes
Development and implementation of decision support system in the ordering process of the purchasing logistics
| Code |
Science |
Field |
Subfield |
| 5.04.00 |
Social sciences |
Administrative and organisational sciences |
|
| Code |
Science |
Field |
| S189 |
Social sciences |
Organizational science |
decision support systems, modelling, simulation, fuzzy logic, inventory control
Researchers (1)
| no. |
Code |
Name and surname |
Research area |
Role |
Period |
No. of publicationsNo. of publications |
| 1. |
23452 |
PhD Davorin Kofjač |
Computer science and informatics |
Head |
2008 - 2010 |
0 |
Organisations (1)
Abstract
The proposed project is dealing with development, implementation and validation of algorithm for inventory control cost optimization in the order placement process regarding stochastic demand in uncertain environment. The project leader has developed an adaptive inventory control algorithm with fuzzy logic in his doctoral dissertation. Algorithm is capable of adapting itself to the dynamic environment and is, on average, achieving up to 30% lower costs than the classical inventory control algorithms. The algorithm was preliminarly validated on historic demand data, but for its further development, there is a need to transfer the methodology into the business process of project beneficiaries, which will enable algorithm's validation and adequate completion. Hence, the decision support system in the purchasing logistic could be developed that could rationalize the ordering process. The system will be implemented on a selected set of materials, production lines and vendors with selected users, who will take part in the system's development and implementation. The quality of the system will be measured by lowering inventory control costs regarding actual business data as well as users' satisfaction. The project will be considered successful if the users will evaluate the simulation model as an useful tool in their decision making process and if it will become a part of the beneficiaries' information system, which is the ultimate goal of the proposed research.
Significance for science
Within the project we developed the concept of anticipatory inventory control with the simulation model. We conducted research on a real case in order to limit the impact of unpredictable environment that results in stochastic lead times and stochastic demand, and to and reduce inventory control costs without stock-outs occurring and inventory capacity being exceeded. We presented an innovative algorithm for inventory control (FZY), based on fuzzy logic and applied a complex cost function that we have not yet observed in the relevant literature. Developed algorithm was compared with conventional algorithms on historical data of a representative sample of materials. Results of preliminary studies indicated that the FZY algorithm surpassed other algorithms and reached a significant costs reduction up to 40%. Within this study we supplemented FZY algorithm expert knowledge database and further validated the algorithm with the actual data, where the algorithm, as expected, achieved significantly lower cost reduction (up to 10%) as the actual stochastic production plan, according to previous analysis, differs from the actual consumption of up to 50%, both in quantity and in time distribution.
Bellman's algorithm of dynamic programming was expanded with a complex cost function and used as a inventory control algorithm. We validated the algorithm on historical data of representative materials and compared the results with previously mentioned algorithms. The study showed that the algorithm is not suitable for optimizing the procurement process on a selected data set, since the results show a 16% increase of costs with regard of other algorithms, as well as an increase of stock-outs.
We carried out a study of the impact of unreliable supplier, which supplies ordered shipments in variable time component, as well as in quantitative variable component. We carried out a Monte Carlo simulation of inventory control model, where we considered two scenarios: a) the quantity ordered is delivered in one shipment, and b) the quantity ordered is delivered in a maximum of three shipments and the quantity is being distributed by the exponential distribution. Orders are supplied according to variable lead times defined by the uniform distribution of the selected interval for a particular material item. The study was carried out on representative sample of materials on historical data. The study revealed that an unreliable supplier significantly increases the cost of supplies, especially if high transportation costs are involved. Costs increased up to 37% if orders are supplied in several consignments rather than in one.
We conducted a study to determine whether we can obtain useful information from stochastic production plan by applying data mining techniques in order to determine whether the appropriate inventory control algorithm can be used on a selected time interval according to the given cost function and constraints. This is an innovative approach to inventory control, where we wanted to narrow the search space and speed up the simulation of adaptive fuzzy inventory control algorithm. In this preliminary study several machine learning methods were used, such as: Naive Bayes, kNN - k nearest neighbors, classification trees, etc. Methods are taught on a production plan characteristics such as average consumption, average lead times, variability coefficient, 1st - 10th period of a Fourier analysis of a production plan sample, etc. The study showed an average performance of machine learning methods, since the classification accuracy ranges between 33 and 58%. kNN classifier yielded the best results (60% classification accuracy). Classifiers are accurate (90% or more) if they classify negative examples, while the classification of positive examples is not satisfactory (only up to 53.6%). We can conclude that these production plan characteristics are not sufficient and the classification of such a changing demand is not satisfactory.
Significance for the country
With the project, we are exposing the collaboration between science and economy on a common project, since Slovenia needs more collaboration on the mentioned level, which is also ascertained in the National research and development programme resolution for the 2006 - 2010 period (ReNRRP). The research performed by academic institutions is crucial for development of Slovenian economy. One of the visions of development policy of our country is the creation and transfer of internationally available knowledge into public benefit and economic exploitation, which is unambiguously supported by this project, since the results of research conducted so far are internationally published and recognized, which is shown with achieved research award, publication in the renown journal and scientific monography, and a presentation at the international conference.
The implementation and completion of developed methods for inventory control cost optimization and integration into the company's existing business information system represents the transfer of innovative knowledge into the business of leading Slovenian companies, which will contribute to a better competitive position. With implementation of such a decision support system in the ordering process, we provide users a chance of better and faster adaptation to new situations - learning by model. The consequence of learning by model will be lower operational costs. This project also coincides with information society technology research that develops human resources. The developed decision support system is intended for human resources development of knowledge and skills, since they are learning by applying simulation models in their process, thus improving their logistic processes and making the companies' business more effective. At the same time, faster human resource introduction is possible, thus making the human resource policy more efficient, since any delay in the process leads towards worse competitive position and consequently towards lower income, i.e. profit.
Furthermore, the decision support system is designed in a way, that allows its dissemination in the companies that belong to other economic branches. The core of the system is the expert data base, that can be adapted to any company or economic branch, thus contributing to development of other economic branches, which are facing similar logistic challenges.
By this project, we bring new knowledge into the operational research area, hence increasing economic effectiveness, competitive position and productivity to reach a higher added value, international recognition and technological and development progress of Slovenian companies.
The project also coincides with the directives that support faster development of base economic areas regarding EU priorities, such as information and communication technologies, and complex systems and innovative technologies. The decision support system, which we have developed during the project, belongs to the information and innovative technologies that are intended for support and understanding of complex logistic systems, which connect international markets and part of which are also Slovenian companies.
Most important scientific results
Annual report
2008,
final report,
complete report on dLib.si
Most important socioeconomically and culturally relevant results
Annual report
2008,
final report,
complete report on dLib.si
