It is necessary for further making use of the measurements to conduct the comprehensive comparison of wind profiler horizontal wind against radiosonde data. Based on the 5year collocated wind speed data from radiosondes and wind profilers in Beijing, the wind profiler data are carefully analyzed, including total average, diurnal and annual variations, different sky conditions. The results show that the root mean square error of 5year average 〖WTBX〗U〖WTBZ〗component was from 2.2 to 5.0 m/s, the 〖WTBX〗V〖WTBZ〗component from 2.3 to 3.6 m/s, and the 〖WTBX〗U〖WTBZ〗component and 〖WTBX〗V〖WTBZ〗component minimum deviation occurred at 850 hPa. The 〖WTBX〗U〖WTBZ〗component average error of different heights was negative, which indicates the wind profiler 〖WTBX〗U〖WTBZ〗component wind speed is smaller than that of balloon sounding, and the absolute value of error increased with height. The 〖WTBX〗V〖WTBZ〗component average errors were negative only at 700 hPa. The similarity of 〖WTBX〗U〖WTBZ〗component at 500 hPa was the best and that of 〖WTBX〗V〖WTBZ〗 component wind at 850 hPa was the best. On average, the wind profiler wind was closer to balloon sounding at 850 hPa and 700 hPa. The errors in the 〖WTBX〗U〖WTBZ〗 and 〖WTBX〗V〖WTBZ〗 components at 14:00 were relatively smaller than those at 08:00 and 20:00, which was related to the strong atmospheric turbulence at noon. The errors were relatively small when the amount of clouds was large, and large when the amount of clouds was small (especially in clear sky). which indicates the wind profiler is sensitive to humidity. The 〖WTBX〗U〖WTBZ〗 and 〖WTBX〗V〖WTBZ〗 components have little difference and good similarity with balloon sounding in summer or summer half year, especially in June and July, and large difference and bad similarity in winter or winter half year, especially in December and January. This should be paid enough attention in forecast operation use and scientific researches.