Abstarct: Market basket analysis is a generic term for methodologies that study the composition of a basket of products (i.e. a shopping basket) purchased by a household during a single shopping trip. The idea is that market baskets reflect interdependencies between products or purchases made in different product categories, and that these interdependencies can be useful to support retail marketing decisions. Due to the recent competition in retailer industry, retailers are striving to improve their operations in order to increase the efficiency and profitability. Therefore, this condition forced retail companies to consider more about fundamental issues such as marketing stimuli, products to be displayed, and space to display assorted products. Retailers collect terabytes of data every day such as transactional data, customer demographics and product sales baxsed on parameters such as seasons and festivals. This data alone cannot enable good decision making for a retailer. It is necessary to discover and understand the underlying patterns involved in the organization’s operations from these data. Hence, there is a need present for accurate, timely information to react to changing market conditions, identify new customer segments, improve inventory management, and optimize overall store performance. Recently, a number of advances in data mining and statistics (association rules) offer new opportunities to analyses such data. In the first phase of this thesis, different algorithms are used in order to analysis and discover the relationship and patterns of product’s adjacency. In the next phase, we take advantage of Multi Criteria Decision Making methods (MCDM) like TOPSIS and DEMATEL to rank the patterns. In the third phase of the thesis, by the use of the baskets generated by the algorithm in the first phase and their rank generated by the MCDM methods in the second phase, an approach for shelf allocation is presented. Finally, for the purpose of evaluation, an experimental study using real data was conducted with definition of “Confidence” and “Support” phrases as benchmark. The result shows 0.994 for “confidence” and 0.974 for “support” .The algorithm is implemented in C++ language with the help of MATLAB software.