The growth of tourism demand is an important source for economic development, jobs, tax revenues, and incomes. Forecasting is needed to monitor fluctuations in tourism demand. Most tourism demand influenced by several factors, such as economic conditions, seasonal variation, or politics. Thus, the task of forecasting was quite difficult to accommodate these factors. One method of forecasting tourism demand is the integration of Empirical Mode Decomposition (EMD) with a neural network-based Feed-forward Neural Network (FNN). To accelerate learning process and improve the accuracy of forecasting, this research using Cascade Forward Backpropagation (CFN). Data tourist visits was decomposed using EMD. Then, all the results from the decomposition (IMF and residues) are used as inputs in the CFN. Outcomes of CFN merged using Adaptive Linear Neural Network (Adaline) to obtain the final forecasting value . Results of this research was accelerated the learning process and improved the accuracy of forecasting tourism demand better than EMD and FNN.