To reduce the number of volume transmission of similar packets from multiple nodes, the data is aggregated and transmitted as single packet. All the current works construct data aggregation trees (DAT) under the deterministic network model (DNM). Since wireless sensor network consists of lossy links, therefore, it is possible to construct a DAT based on the probabilistic network model. In current literatures, the load balance factor is not considered in the construction of data aggregation trees (DATs). In this paper, we are going to construct a load-balanced data aggregation tree-based on the probabilistic network model (PNM). Three problems namely the load-balanced maximal independent set problem (LBMISsn), the connected maximal independent set problem (CMISsn) and the load-balanced data aggregation tree (LBDATsn) construction problem are investigated. First we construct a load-balanced maximal independent set (LBMISsn) using approximation algorithm and then the LBMIS nodes are connected by choosing a minimal set of nodes, i.e., CMISsn and finally each node is assigned a parent in CMISsn, then by giving direction for each link forms an LBDAT. Map-reducing approach is used to aggregate data. Data aggregation is carried out in two phases: sMap and reduce. sMap protocol is used to sense and compute, while reduce protocol performs forwarding and aggregating data. The energy consumed by each node is saved and enhanced the network lifetime. It helps to avoid redundancy in transmission of data and performs well in terms of communication overhead and therefore increases the throughput. © 2017 Pushpa Publishing House, Allahabad, India.