BIG DATA RECOMMENDATION PROBLEMS IN E-COMMERCE SOLUTIONS FOR SMALL BUSINESS

Abstract

The dynamic development of e-commerce has increased the demand for efficient algorithms and systems based on statistical analysis. The simplest of them use the web traffic statistics, other use sales parameters. Because of the amazing simplicity, transparency and enhanced features, much popularity was gained by the Google Analytics tool. None of the methods, however, without the appropriate algorithms that automate operations, is suitable for use in real time. Intelligent recommendation systems, such as the mechanism of Collaborative Filtering, significantly contribute to an increase in sales but are generally characterized by poor scalability. Of course with proper computer infrastructure and specialist knowledge, it is possible to gather big volumes of data and analyze them. All sophisticated solutions, however, are rather reserved for large companies, whose activity is based on the Internet. In this article, Big Data recommendation problems are described. Advantages and disadvantages of several used in practice algorithms are considered in particular emphasis on the suitability for the small e-commerce business. The main point of the article is the proposition of the simple in implementation recommendation algorithm and thereby achievable for small business. What is more, the online test was performed and its results presented as a good performance proof. The actual data were used thanks to the courtesy of Run4Fun.pl. In the test, the aspects of a large amount of data but also their volatility and diversity was taken into consideration.