A K-Nearest Neighbors（KNN） based fiber nonlinearity compensation technique for Polarization Division Multiplexing-Quadrature Phase Shift Keying（PDM-QPSK） unrepeartered Raman amplification system is studied. The nonlinearity compensation of optical fiber is realized by classifying the test data by calculating the distance from the test data to all the training data. All the simulation results show that the KNN-based fiber nonlinearity compensation technique can alleviate the fiber nonlinearity in 112 Gbit/s PDM-QPSK 350 km dispersion compensated unrepeartered Raman amplification system. Especially

the input optical power can increase by 1 dBm at the bit error rate of 10

-3

after KNN-based nonlinearity compensation.