A prediction methodology for the change of the values of financial products

A systematic algorithm based on data smoothing and the Bayes' theorem is proposed to predict the increase or decrease of a financial time series, which can be used in trading financial products when decisions need to be made between long and short positions. The algorithm compares the observed product values with those in the history to find a similar pattern with the maximum likelihood, based on which future up-down movement of the value is predicted. Empirical studies with S&P 500 Index and stocks of several companies show that the proposed methodology improves the rate of the correct predictions by about 30% or more, relative to naive prior probability or moving average convergence divergence predictions.