Regression-Based Uplink Interference Modeling Using Area-Limited Information

The proliferation of connected devices in wireless systems result in severe intercell interference (ICI). To mitigate ICI, an accurate interference model of link-granularity is highly beneficial. The data generated by wireless devices during networks operation can contribute to mining the interference relationship if properly utilized. Existing frameworks of data-driven interference modeling require data collected from the entire network. Such data are unobtainable in a large network due to the communication constraints of a single network node. This paper proposes an uplink interference modeling algorithm that takes as input area-limited wireless big data instead of global data. With interference modeling formulated as a supervised learning task, the proposed algorithm can precisely model the interference relationship between an interested user equipment (UE) and any of its primary interfering users. Simulation shows that under high interference scenario the proposed algorithm achieves the same signal to interference plus noise ratio (SINR) prediction accuracy with significant reduction in training complexity compared to baseline algorithms.