Temporal data object accessed by transactions of real-time database systems have been applied in diverse domains. However, almost all of the current transaction scheduling algorithms so far only take logic consistency of object into account, while ignoring temporal consistency. This paper presents the concept of temporary deadline, and proves the feasibility of low-cost predictive algorithm in the form of theorem in order to abort or postpone those transactions that cannot satisfy temporal consistency as soon as possible, therefore, computation resources are saved to accommodate for other transactions. A new transaction scheduling algorithm with temporal constraints, called TDDBPA(Temporary Data Deadline-Based Predictive Algorithm)is proposed based on strategies mentioned above in this paper. The simulation results show that performance of our algorithm significantly outperforms that of well-known algorithms so far.