With the widespread adoption of crowdsourcing models, volunteer crowdsourcing has demonstrated significant advantages in fields such as data collection and social welfare. However, traditional task allocation methods based on external factors such as cost or location often overlook volunteers’ subjective preferences in crowdsourcing. Although subjective preferences play a crucial role in volunteer crowdsourcing, few studies in this area consider them in a systematic way. While general crowdsourcing research incorporates subjective preference modeling, it typically assumes complete preference data, which is often unrealistic in real-world settings. To address this issue, this paper proposes a two-stage task allocation framework based on partial pairwise preferences. First, the Double Transitive Anchor-KNN (DTA-KNN) algorithm is proposed, which addresses the problem of partial pairwise preferences by designing a double-layer transitive closure structure and constructing a neighborhood of volunteers with similar preferences. Second, the Flexible-Priority General Dictatorship (FPGD) task allocation mechanism is developed, which introduces dynamic rotation of volunteer order during the allocation process and builds a priority index to support decision-making in cases of competition. This approach enhances the consistency of task assignments with volunteer preferences and improves the overall fairness and robustness of the system. Experimental results show that the proposed approach significantly outperforms existing methods in terms of preference completion accuracy, allocation fairness, and resistance to manipulation, offering a more practical solution for task allocation in volunteer-based crowdsourcing.