Ultrawideband’s (UWB) unstable ranging results from the multipath effect and interference from non-line-of-sight (NLOS) indoors, which reduces positioning accuracy. Therefore, we have developed a distance-adaptive weighted positioning algorithm based on random forest (RF) optimization. Firstly, the ranging data are identified and evaluated for any missing values and subsequently processed to determine the initial positions of the tag using different three-base station combinations. Secondly, outlier coordinate points are eliminated using the interquartile range (IQR) method. Thirdly, the initial coordinates of the remaining tags are assigned according to corresponding weights and the measured distance of each base station. A weighted calculation is then performed to obtain the estimated coordinates for tags. Finally, the UWB ranging data is employed as the input to the RF regression model, thereby rectifying the inaccuracies in the estimated coordinates. This process is used to determine the final tag coordinates. Compared with previous methods, the developed method enhances precision in localization and provides more stable positioning.