The invention provides a semi-supervised positioning method, which relates to the field of indoor positioning technologies. The WLAN-based semi-supervised positioning method comprises the steps of: step 1, establishing an initial mark data set; step 2, clustering the initial mark data set through fast search and look-up classification, and marking missing position information of a cluster center of the initial mark data set; step 3, training the initial mark data set by means of an extreme learning machine; step 4, configuring and expanding the initial mark data set according to data results of the step 2 and the step 3, so as to obtain an expanded mark data set; step 5, and evaluating the correctness of the expanded mark data set by adopting a transposition concept, and outputting a positioning result. The WLAN-based semi-supervised positioning method improves the positioning accuracy rate, and is short in running time.