An Improved Partial Ambiguity Resolution Algorithm is proposed to address the issues of misjudgment or omission in subset selection and threshold setting of traditional partial ambiguity resolution algorithms. The algorithm performs an initial screening based on elevation angle and signal-to-noise ratio, and then progressively eliminates the largest variance ambiguity according to the success rate estimation until the success rate exceeds the predetermined threshold, achieving optimal subset selection. This algorithm effectively reduces the search space of ambiguity and improves the fix rate and reliability of ambiguity fixing. Experimental results show that the improved PAR algorithm has a significantly higher fix rate compared to existing algorithms, with a maximum static positioning fix rate of 95.0%. The horizontal and vertical accuracy are improved by 30%~45% and 60%~75%, respectively. Furthermore, the algorithm can effectively reduce the initialization time of RTK positioning and accelerate the convergence speed of the solution.