The rapid expansion of the Internet of Things (IoT), in parallel with the substantial reduction in the cost of connected devices, have enabled the deployment of sensors in large scale. This massive deployment allows for more comprehensive coverage of the studied area when monitoring a physical quantity over time. As sensors become more densely packed, they often provide similar data due to their proximity. In this paper, we look to identify such similarities among sensors based on their returned data, in order to build groups of similar sensors. Groups of similar sensors can have several advantages, such as detecting sensor failures and performing reduction of sensors transmissions. Our primary focus is on a generic scenario that has received limited attention in the existing literature: sensors are deployed at different moments and exist in the environment for a limited duration, transmitting noisy and irregular data over time, without synchronization among them. To address this, we define a data-driven similarity metric, which is then used for clustering similar sensors. According to the set of messages sent by a sensor, we apply an interpolation method that enable us to evaluate similarity between two sensors as the mean magnitude difference between the interpolations over their common definition interval. The duration of this common definition interval characterizes the trustworthiness of the distance. Hence, in the hierarchical clustering method we propose, we introduce a linkage method that assigns higher weights to distances calculated over longer comparison durations. Through simulations, we demonstrate the superiority of our method compared to the state-of-the-art Dynamic Time Warping (DTW) distance and a hierarchical clustering with complete linkage inspired by related works. Our results establish a mean improvement of 23% of our approach in terms of V-Measure. Moreover, we provide comprehensive experiments assessing the robustness of our solution under various sensor measurement noise levels and employing different stopping criterion strategies.