A Loss of Flow Accident (LOFA) is an accident that causes cooling to slow down due to pump failure or stopping during operation. A fast or slow change in two-phase flow, when overlooked, can lead to an accident like LOFA, and thus, understanding its nature is essential for nuclear reactor safety. In this paper, we demonstrate that using one of the machine learning techniques called Support Vector Machine, one can find the most important factors in two-phase flow change. Using one of the commercial thermal hydraulics analysis code, MARS (A multi-dimensional thermal-hydraulic system code), simulation results were obtained for several scenarios where the mass flow rate decreased sharply. The transient flow change phenomenon near a single PWR rod, which is the simplest case of the reactor, is modeled. The outlet temperature of the coolant which is the final output factor of the transient flow change and the peak temperature of the cladding rod are very important factors for safety analysis. We also show that the outlet temperature profile of the coolant can be used to predict the unknown mass flux and the peak temperature of the cladding rod using the Multi-class Support vector machine algorithm. These results suggest that machine learning techniques may be used to analyze the complex systems of accidents that may occur in the nuclear system.