Intermittent demand planning is one of the most challenging tasks in supply chain management. Planning inventory at the point of sale often involves dealing with discontinuous historical data. This discontinuous demand refers to demand that is sporadic and does not follow a consistent pattern. Various factors, such as seasonal changes, market disruptions, or unique customer behavior, can contribute to this irregularity. Planning for discontinuous forecasting involves managing irregular or unpredictable demand patterns, which can be challenging for businesses. In this paper, we propose a machine learning approach to address this problem. This approach was successfully implemented approximately a year ago, resulting in up to a 50% improvement in demand estimation in certain scenarios. This significant enhancement has greatly minimized the frequency of unfulfilled customer requests and improving overall customer satisfaction. Our model consists of a two-phase algorithm. First, we apply random forest as a regression method. Second, we obtain the expected discontinuous series by extrapolating the results of the random forest model. In mathematics, extrapolation is used to estimate the value of a quantity outside the interval in which that quantity is known.