The problem of customer acquisition and retention has been well studied. Number of soft techniques has been discussed earlier towards the development of retail marketing. However, the methods suffer to achieve higher performance and require some strategically approach. The application of data mining techniques has great impact in the development of retail marketing. Also, the volume of customer logs is higher which challenges the methods in identifying user interest in exact way and such issue can be handled with the inclusion of machine learning techniques. Towards this, a novel customer interest prediction algorithm with Multi Variant K-means clustering and pattern mining techniques is presented in this article. The method is focused to support E-Commerce systems. The method first identifies the purchase histories and enquires of various users. The logs have been clustered using Multi variant k-means clustering algorithm. Second, the method identifies the list of purchase patterns. From the patterns being generated, the method identifies the concrete interest of the user and identifies the similar interested users. Using this, set of recommendations has been generated for the user which has been used to populate advertisements, placing banners in user web page, and so on. By identifying the user interest according to their purchase pattern and by generating the recommendations based on the logs of similar interested users, the method supports the customer retention in higher ratio.