Mining frequent patterns of stock data using hybrid clustering

Patterns and classification of stock or inventory data is very important for decision making and business support. In this paper we proposed an algorithm for mining patterns of huge stock data to predict factors affecting the sale of products. Identification of sales patterns from inventory data indicate the market trends which can further be used for forecasting, decision making and strategic planning. The objective is to get better decision making for improving sales, services and quality as to identify the reasons for dead stock, slow moving and fast moving stock. We have two phases in which first phase includes initial clustering which is performed on the database with the help of a clustering algorithm. In the second phase we use most frequent pattern, MFP algorithm to find the frequencies of property values of the items. The existing system uses k-means clustering algorithm along with MFP for mining patterns. In order to improve the execution time the proposed system uses efficient methods for clustering which includes Partitioning Around Medoids, PAM and Balanced Iterative Reducing and Clustering using Hierarchies BIRCH along with MFP. The most efficient iterative clustering approach called as PAM is used for initial clustering and is then combined with frequent pattern mining algorithm. In order to meet the memory requirements, an incremental clustering algorithm BIRCH is also used for mining frequent patterns. So, the evaluation of these clustering algorithms along with MFP is made with respect to the execution times. The results are compared and shown graphically.