Data mining is a powerful and a new field having various techniques. It converts the raw data into useful information in various research fields. Clustering is used to collect similar data in a group. It is a process of putting similar data into groups. A popular technique for clustering is K-means in which data are partitioned into K clusters. In this method, the number of clusters is pre de?ned and the technique is highly dependent on the initial identi?cation of elements that represent the clusters well. But we cannot changethe number of cluster at mid of execution of algorithm. But in k-mean, important factor is that how many clusters we should take, it may be less and it may be more. There is not any mechanism to estimate the number of clusters in k-mean. It totally depends upon user, how many he takes. But for large amount of data user can't decide how much data have similar. For example, if maximum data have common similarities, so why we take more cluster. For this it may be minimum number of s for better evaluation and better performance. similarly if we have a amount of dissimilar data so we should take more cluster in k-mean. For this we are using a priori to generate association rules and with the help of association rule we put the values in my proposed equation and calculate the probability factor to give us the estimated number of cluster sfor k-mean.

Data mining clustering a priori k-means association rules probability factor.