Predicting the Financial Growth of Small and Medium-Sized Enterprises using Web Mining

Abstract

Small and medium enterprises (SMEs) play an important role in the economy of many countries. When the overall world economy is considered, SMEs represent 95% of all businesses in the world, accounting for 66% of the total employment. Existing studies show that the current business environment is characterized as being highly turbulent and strongly influenced by modern information and communication technologies, thus forcing SMEs to experience more and more severe challenges in maintaining their existence and expanding their business. To support SMEs at improving their competitiveness, researchers turned their focus on applying data mining techniques to build growth prediction models. However, current prediction models only include few data types such as financial or operational data and thus cannot explain the whole and complex context of SME growth. Moreover, data used to construct these models is primarily obtained via questionnaires, which is very laborious and time-consuming, or is provided by financial institutes, thus not publicly available and highly sensitive to privacy issues. Recently, web mining has emerged as a new approach towards obtaining valuable insights in the business world. Web mining enables an automated and large scale collection and analysis of potentially valuable data from the web, a popular and interactive medium with immense amount of data freely available for users to access. While web mining methods have been frequently studied to anticipate growth of sales volume for e-commerce businesses, it remains unclear how web mining can be applied to leverage SMEs growth prediction. In investigating this question, the present thesis analyses the use of web mining for SMEs growth prediction. In a case study, we demonstrate the use of publicly available web data for growth prediction in the gastronomy industry. First, a comprehensive overview of factors influencing the growth of restaurants is provided through a systematic literature review. In total, 49 factors influencing the growth of restaurants are identified, serving as a knowledge base to develop a growth model for restaurants. Next, the usability of various web data sources is manually inspected with respect to the identified growth factors. Web mining techniques are applied for large-scale collection and preprocessing of unstructured web data. Finally, based on data from 403 Swiss restaurants, we build and compare different binary classification models using supervised machine learning algorithms. More specifically, the developed models classify a restaurant either in a non-growing or growing restaurant. The algorithms for predictive modeling include logistic regressions, random forests and artificial neural networks. The present thesis makes a significant contribution to the body of literature at the intersection of SMEs growth research, web mining, and applied machine learning. To summarize, our findings suggest that web mining is a feasible approach to leverage growth prediction modelling for SMEs. By means of web mining, valuable business insights can be extracted from the web, which then can be further used for predictive modelling by applying machine learning techniques. Moreover, to the best of our knowledge, our case study is the first to apply web mining combined with supervised machine learning techniques to model the growth of restaurants based on publicly accessible web data.