Network operators are accustomed to using IP-layer active measurements for assessing end-to-end network performance and expect that new technology, such as IoT, provides similar means. Unfortunately, active measurements in IoT systems are associated with both energy and network overhead. This paper presents and evaluates a novel active-measurement proxy approach, based on machine learning, that enables reduction of active measurement overhead in IoT systems. The paper describes the approach and its implementation. Further, the approach is evaluated in a IEEE 802.15.4 testbed, and the results show high-performing and accurate modeling.