A Framework for Predictive Data Mining in the Telecommunications Sector

Nowadays, the Internet comprises of huge amount of electronic information concerning different companies’ financial performance. This amount greatly exceeds our capacity to analyze it, the problem being that we often lack tools to quickly and accurately process these data. DM techniques are interesting mechanisms that can be applied to rapidly changing industries, in order to get an overview of the situation. One such market is the international telecommunications industry. In this paper we construct a framework using DM techniques that enables us to make class predictions about telecommunication companies’ financial performance. Our methodology allows us to analyze the movements of the largest telecommunications companies, to see how companies perform financially compared to their competitors, what they are good at, who are the major competitors in this industry, etc. The dataset contains 88 companies from five different regions: Asia, Canada, Continental Europe, Northern Europe, and USA, and consists of seven financial ratios per company per year. The data used to calculate the ratios were collected from companies’ annual reports (between 1995 and 1999), using the Internet as the primary medium. We used data from 2000 and 2001 to test our classification models. We have obtained good maps (SOMs) in terms of easeof-readability and the average quantization error and clearly identified the six financial performance clusters. The results of our class prediction models also correspond very well with the SOM model.