To reduce the bias drift of fiber optic gyroscopes due to the temperature effect, a temperature compensation model of the fiber optic gyroscope was established using an ensemble learning algorithm based on a least squares polynomial model and back propagation (BP) neural network model optimized by a genetic algorithm (GA-BP). A temperature compensation experiment of the fiber optic gyroscope was conducted after online compensation. Experimental results show that the model reduces the bias drift of the fiber optic gyroscope by more than 85% in an environment with a temperature change of -40~+60℃, and the average bias output of the compensated starting section is closer to the zero position.