According to the shortcoming of low convergence precision of the high dimensional function of Drosophila algorithm, an improved algorithm based on fan-shaped search fruit fly optimization algorithm (FS-FOA) is put forward. This algorithm improves the search path of drosophila group on the basis of the original fruit fly FOA algorithm, gives individual of fruit flies profit-seeking, makes more individuals of the fruit flies move in the direction of bigger taste concentration, makes fruit fly group search direction more selective and increases convergence speed and convergence precision for the fruit fly algorithm to deal with high dimensional function. A FS-K clustering algorithm is forwarded based on the combination of FS-FOA algorithm and k-means clustering. Through the comparison experiments of the original k-means clustering and the original FOA algorithm, five classical test functions are introduced to test the optimizing results of the original FOA algorithm and FS-FOA algorithm, and the results show that FS-FOA algorithm has higher convergence precision. Five UCI public data sets are introduced to test the improved FS-K clustering algorithm, the original k-means algorithm, SOM clustering algorithm and FCM clustering algorithm, and the results show that FS-K clustering algorithm has better clustering effect