On Internet of Things (IoT), everything can be accessed anytime, anywhere, and works without human intervention. In IoT everything collaborates to deliver services and applications to users, Ambient Assisted Living (AAL) being one of these applications. Some AAL smart homes uses infrared sensors to recognize some activities of daily living and to track people along the environment. Location tracking is vital in Ambient Assisted living and can be a useful information to improve AAL systems. A common problem in such systems is that each tracking model is based on a specific sensor's placing architecture. In order to assure that the system will work properly, the model has to be fitted by an expert. Modeling is usually costly and it relies on a specific architecture. In our previous work, the tracking model needed to be fitted manually. In order to introduce adaptability, this work proposes an approach to automatically fit the model avoiding the need of an expert to fit a different model for each kind of sensor's placing architecture. The proposed approach was evaluated using real data from a set of pyroelectric infrared sensors and a set of scenarios performed in a simulated apartment