Finding of hidden and previously unknown information in large collection of data is the process of data mining. Mining association rules is a very important model in data mining. Using association rules different type of regularities and patterns can be identified. The main approach of association rules is the market basket analysis which exposes relationships between the items customers are regularly buying. In most of the previous approaches of finding association rules a single minimum support threshold value is used for all the items or itemsets. But all the items in an itemset do not behave in the same way where some appear very frequently and some appear very rarely. Therefore the support requirements should vary with different items. Here we proposed new algorithm and was tested using different data sets to prove the advantages. The analysis showed that the proposed algorithm is easy and efficient and it saves time by focusing only on necessary associations comparing to existing algorithms.