Recent studies reflect a growing interest in applying neural networks to answer stock behavior. Most of these studies rely heavily on fundamental analysis factors to determine future stock prices. In fact, there exists another approach, called technical analysis, which attempts to predict the stock trend by using data surrounding past prices and volumes. This paper investigates whether using these indicators as inputs to a neural network will provide more accurate predictions of future stock trends and whether they will yield higher trading profits than the traditional technical indicators. Feed-forward, probabilistic, and learning vector quantization neural networks are then examined to predict the short-term trend signals of several major stocks in different industries. The overall results indicate that the proportion of correct predictions and the profitability of trading exercises guided by these neural networks are consistently higher than those guided by the buy-and-hold strategy and the individual technical indicators.