Foreign exchange market is the largest trading market in financial market. As electronics and network are popular and developed, investors who want to take part in foreign exchange trade don’t have to trade in banks or foreign exchange institutions. Foreign exchange market changes all the time, amount of increase and decrease is greatly larger than stock market so it is hard to predict foreign exchange development. This study designs an automatic exchange system to trade in MT4 platform and lead to back propagation neural network for optimization. The system program can improve order correctness to verify the designed system can provide decision making assistance for foreign exchange. It is found from the research that trading volume and time series are related. It is possible to try to invest cash in real market and do some adjustment according to other currency features and strategy of this study.