Eliminating the influence of environment temperature is critical for high-accuracy carbon nanotube polymer nanocomposites sensors. In this work, the temperature effects on the nanocomposite are studied by both experiment investigation and simulation calculation. Nearly constant resistance values were found at a CNT loading around 3 wt%. By considering the temperature effect and CNT structural distortion in the developed percolation network model, simulation results agree well with experimental data. On this basis, results show that the thermally assisted tunneling on CNT junctions and thermal expansion of polymer matrix are the two core mechanisms, relaxed CNT junctions in CNT/polymer nanocomposite depressed the temperature effect, and the influence of environment temperature could be significantly reduced by adjusting CNT loadings and choosing a matching polymer matrix. All of these findings will benefit for the design of high-accuracy sensors.