Solving a Storage Optimization Problem Using Data Mining Techniques

OData support
Supervisor:
Nagy István
Department of Telecommunications and Media Informatics

Properly managing warehouse operations is an important factor from perspective of success of an organization. One hand if the product managers make inaccurate estimation on necessary amount of products during the planning of the supply process, then the costs of storing will increase unnecessarily. On the other hand if they underestimate the future demand, then they will not be able to fulfill the customers’ order. These factors may cause decreasing profit

The data mining would give an effective tool for the product managers, which can help them dealing with optimisation of supply chain processes and reducing the cost of storing.

In my study I had to optimize a supply chain process of a car dealer organization. I have rephrased this task to time series prediction. Accessible data have contained information about vehicles: specific type, equipment and date of sold.

Studying the behavior of sales in the past, we can build a predictive model for the concrete problem, which can help to forecast the necessary quantity of products in the future.

I have also created some baseline algorithms. These methods predict the future behavior of sales with simple techniques. The main aim was to make models which have better performance than the baselines.

I have used some different data preparation and modeling methods, such as decision trees, neural networks, linear and logistic regressions. With these techniques I was able to make models which can predict the future sales in three different time interval (for one, two and four week period). The comparison of the models and baselines have showed that best models have approached better performance than the baselines, so I have accomplished the business goal.

Downloads

Please sign in to download the files of this thesis.