DATA FUSION FOR TREND IDENTIFICATION IN LARGE RETAIL BUSINESSES USING FUZZY TECHNIQUES

The work discussed in this paper is a response to major perceived needs for higher organizational standards in the retail business infrastructure. In this context effective planning of resources strongly relies on the ability to understand customer behavior and market trends based on high volumes of data made available at different levels of detail by different sources. The objective is to integrate sales information from heterogeneous sources and to combine this data using fuzzy techniques, along with traditional ones, in order to identify incipient sales trends and customer behavior patterns, tracking their evolution over time for improved store organization and planning. A set of software tools was developed and validated in order to build a model architecture capable of gathering and processing information from different sources, such as cashier’s receipts, individual customers (through store membership/advantage cards), and aggregate weekly sales indicators across stores located in a geographic area of interest, either individually, by source type, or in combinations.