The Wireless Sensor Networks (WSNs) deployment introduces many issues and challenges mainly in terms of energy independence. In this context, we adopted a self-adaptation strategy composed of four steps (Monitor, Analyze, Plan and Execute) to manage Quality of Service (QoS) [1]. This paper focuses on the first step which consists in monitoring and sending QoS values such as the value of power remaining in the battery of each sensor. We notice that the transmission process is very expensive in terms of energy and reduces the battery lifetime. In this work, we propose a probabilistic approach that estimates a part of these QoS monitoring values and therefore economizes their transmission energy and extends the sensor battery life. Our approach is based on the hidden Markov chain and the fuzzy logic and composed of two steps: (i) learning which allows apprehending the WSNs behavior and (ii) predicting which estimates QoS monitoring values. A WSN application deployed in a datacenter is studied as an illustration. The carried out experiments over AZEM WSN simulator show its feasibility and efficiency.