Mobile commerce means “the delivery of electronic commerce capabilities directly into the consumer's hand, anywhere, via wireless technology.” In this paper, we consider a system, called Mobile Commerce Predictor (MCP), for mining and prediction of mobile user behaviours under the context of mobile commerce. The main objective of this framework is to predict future m-commerce behaviour of the user on the basis of his current transaction. Predicting future always associated with risk. By improving prediction accuracy, we can minimize that risk. We present efficient strategy to improve accuracy of prediction by introducing confidence as new parameter in existing prediction strategies of MCP framework. We also present results of applying this strategy to transaction data obtained from sample database (i.e. AdventureWorks) which shows effectiveness of the strategy over existing ones.