A wireless sensor network consists of many energyautonomous micro-sensors distributed throughout an area of interest. Each node has a limited energy supply and generates information that needs to be communicated to a sink node. To reduce costs, the data sent via intermediate sensors to a sink, are often aggregated. The existing energy-efficient approaches to in-network aggregation in sensor networks can be classified into two categories, the centralized and distributed approaches, each having its unique strengths and weaknesses. In this paper, we introduce PMDA (Prediction based Mobile Data Aggregation) scheme which uses a novel data aggregation scheme to utilize the knowledge of the mobile node and the infrastructure (static node tree) in gathering the data from the mobile node. This knowledge (geo-location and transmission range of the mobile node) is useful for gathering the data of the mobile node. Hence, the sensor nodes can construct a near-optimal aggregation tree by itself, using the knowledge of the mobile node, which is a similar process to forming the centralized aggregation tree. We show that the PMDA is a near-optimal data aggregation scheme with mobility support, achieving energy and delay efficiency. This data aggregation scheme is proven to outperform the other general data aggregation schemes by our experimental results.