The present disclosure relates to the field of network communication, and specifically discloses an adaptive equalization method, including: obtaining a first filtered signal according to a first filter coefficient; deciding the first filtered signal based on an original constellation map to obtain a first decision signal, and deciding the first filtered signal based on a level (n−1) constellation map to obtain a level (n−1) pseudo decision signal; if average energy of the level (n−1) error signal is less than a level (n−1) threshold, switching the level (n−1) constellation map to a level n constellation map; obtaining a second filter coefficient according to the update magnitude; obtaining a second filtered signal according to the second filter coefficient; and deciding the second filtered signal based on the original constellation map to obtain a second decision signal. Embodiments of the present disclosure also disclose an adaptive equalizer.