Indoor localization has received tremendous attention in the last two decades due to location-aware services being highly demanded. Wireless networks have been suggested to solve this problem in many research works, and efficient algorithms have been developed with precise location and high accuracy. Nevertheless, those approaches often have high computational and high energy consumption. Hence, in temporary environments, such as emergency situations, where a fast deployment of an indoor localization system is required, those methods are not appropriate. In this manuscript, a methodology for fast building of an indoor localization system is proposed. For that purpose, a reduction of the data dimensionality is achieved by applying data fusion and feature transformation, which allow us to reduce the computational cost of the classifier training phase. In order to validate the methodology, three different datasets were used: two of them are public datasets based mainly on Received Signal Strength (RSS) from different Wi-Fi access point, and the third is a set of RSS values gathered from the LED lamps in a Visible Light Communication (VLC) network. The simulation results show that the proposed methodology considerably amends the overall computational performance and provides an acceptable location estimation error.