A Bayesian data-mining approach is introduced for dynamic spectrum sensing in spectrum-heterogeneous cognitive radio networks. The goal is to find the spectrum opportunities in temporal, spectral, and spatial domains in the presence of unreliable sensing data and unavailable location information. In spectrum-heterogeneous networks, the availability of spectrum varies over the space and different sensors experience different spectrum opportunities. Thus, the measurements from sensors cannot be simply aggregated to detect the spectrum opportunities. Moreover, unreliable data will negatively impact the decision-making process. The task of inferring the spectrum status becomes even more challenging when the sensors are not equipped with location-finding technologies. In this paper, we propose a probabilistic model to cluster the sensors solely based on their observations, not requiring any prior knowledge of the network topology, location of the sensors, or the number of clusters. After receiving the sensing data, the base station updates the probability distributions of cluster membership, channel availability, and device reliability. All the update rules are derived mathematically by the variational inference. Then, the distributions are employed to find spectrum opportunities via multi-label graph cuts method. Experimental results demonstrate the effectiveness of the proposed approach with respect to existing algorithms.