Indoor fingerprint localization algorithm is easily affected by many factors, then causes the fuzziness of fingerprint data set. In response to the Adaptive Neuro-Fuzzy Inference System (ANFIS) applied in indoor positioning, the increasing number of input parameters in the fuzzy system causes the curse of dimensionality and computational complexity of ANFIS. This paper proposes an indoor localization algorithm based on the Hierarchical Adaptive Neuro-Fuzzy Inference System Tree (HANFIS-Tree). The algorithm divides the overall ANFIS into an interconnected HANFIS-Tree structure, reducing the number of fuzzy rules, thereby improving the computational efficiency and interpretability of the system. Additionally, an improved feature weighting algorithm is used to select the difference in signal strength (RSSI) with high correlation to the coordinate position as input parameters, enhancing the positioning accuracy of the system. Moreover, a subtraction clustering algorithm is introduced to initialize the input parameters, improving the convergence speed of the system. Experimental results demonstrate that compared to ANFIS, HANFIS-Tree achieves an improvement of 3.5 cm in positioning accuracy on the x-axis, and approximately doubles the computational efficiency on the y-axis in the indoor localization algorithm.