In low-energy networks, energy consumption is a significant concern. The adjustment of transmission power can save considerable energy at nodes during communication. The commonly used power control schemes maintain the transmission power based on the received signal strength indicator (RSSI) that depends on the interference in the environment. It is necessary to consider interference for retaining the lowest transmission power since low-energy network signals are vulnerable to interference changes. The earlier investigations suggested only linear models for power prediction in low-power networks. Hence, this paper investigates a classification-based transmission power prediction approach with the presence of interference. The approachworks for linear and non-linear models based on RSSI, link quality indicator, neighbour node distance, and receiver power to maintain reliable communication with low energy consumption. The experiments were conducted in natural environments with common interference causes such as the human body, concrete walls, trees, and metallic bodies. The performance of the approach is analyzed with different prediction algorithms such as regression and classification. The investigation results demonstrate that it is possible to build a classification- based power prediction for linear and non-linear models by considering different spatial effects with 99% accuracy.