Though global positioning system is a well-accepted technology for outdoor positioning, it is ineffective for indoor environment. Therefore, the search for effective and cheap solutions still continues. In this work, it is aimed to enhance the performance of indoor positioning system by a hybrid approach integrating WiFi and magnetic field sensor data. The positioning accuracy is improved by taking advantages of these sensor types. Besides, significant improvements in terms of computation time are achieved thanks to 'ReliefF' feature selection and 'k-means' clustering algorithms employed within the work. The results of the tests, which are obtained using Extreme Learning Machine models constituted for each region acquired after clustering, approves the effectiveness of the proposed method.