Abstract: An apparatus and method for predicting a friction coefficient of a road surface are disclosed. The apparatus includes a microphone that obtains a sound signal around a tire mounted on a vehicle, and a controller that estimates a state and a type of the road surface and a degree of tire wear corresponding to the sound signal around the tire based on a first deep learning model, estimates a peak adjustment factor and an initial inclination adjustment factor of a friction coefficient curve corresponding to the state and the type of the road surface and the degree of tire wear based on a second deep learning model, and predicts the friction coefficient based on the peak adjustment factor and the initial inclination adjustment factor of the friction coefficient curve.