Conventional fuzzy joint points clustering method calculates the transitive closure by the maxmin composition based on Euclidean distance, which makes the transitive closure anamorphic and include a lot of false data association information, and  low clustering precision and high time complexity. Aiming at these problems, we propose an improved fuzzy joint points clustering algorithm. We first use the combined kernel function to obtain the fuzzy similarity matrix, which can increase the nonlinear recognizing ability. Then we use the maxheap to store fuzzy similarity matrix, traverse the empty elements of the transitive closure matrix, and populate the empty elements by the bridging element. Experimental results on the UCI data sets and artificial data sets show that the proposed approach has a shorter time and 20% higher clustering accuracy compared with the conventional FJP. In addition, the computational efficiency on large data sets is also more excellent than the traditional FJP clustering algorithm, which verifies that the idea of the improved FJP clustering algorithm is effective and feasible.