Advances in wirelesscommunication and microelectronic devices technologies have achieved the success production of low-power micro-sensors, enabling the large scale deployment of wireless sensor networks (WSNs). In order to track objects energy efficiently in OTSNs, many methods have been proposed but the performance is still not satisfactory. In fact, improving the performance of flooding is necessary for energy efficiently object tracking in WSNs. In this paper, we will propose a data mining method for discovering the partition structure, which can be applied to the improvement of flooding in OTSNs. Through empirical evaluation on various simulation conditions, our proposed method is shown to deliver excellent performance in terms of accuracy and execution efficiency.