Market basket analysis of heterogeneous data sources for recommendation system improvement

Abstract

Information systems that process a large amount of data become an integral part of our lives. Development of online markets and market technologies lead to the need for retailers to analyze customers’ behaviour. The result of the effective analysis may improve both supplier’s profitability, quality of service and customer satisfaction that attracts increased interest for research. One of retailing data analytics applications is the construction of recommendation system. Increase the quality of the recommendation system is possible when analyzing a larger amount of data, which can be obtained from external heterogeneous sources. Examples of sources for data integration can be data from online and offline markets inside one company or data from partner companies. Within one market area, a range of offered products may be similar, while the characteristics or associative rules formed for them may differ. Therefore, for the correct integration of external data sources into the existing recommendation system, it is required to analyze the structure and content of additional data sources to use only beneficial parts of that data. In this work, we propose a study on the integration of heterogeneous data sources from a grocery supermarket based on Market Basket Analysis methods.