Time-series Mining Methods And Their Applications

Portfolio optimize means investor choose all kinds of stock and other asset to form a combination according to the return demanded, then optimize these combination for realizing investment goal. The reason of the investor optimize Portfolio is reducing Not-system risk. Investor can find a balance point between return and risk through optimize Portfolio, namely realizing the maximization of the return on the premise of taking on a certain risk, or minimize the risk on the premise of the fixed return.For a proper combination of assets, we aim to select asset within the classes. But a good distance measure is key to Time-Series clustering, which is high dimension, complex volatility, and high noise. So common distance measure in clustering is no available. For solving the problems above, sub-sequence measure and Markov chains transferring matrix are proposed, one of which is aim to measure the figure between series, the other is focus on the status transferring statistics. Further the distance between transferring matrix and the length of similar subsequence could be the measure of time series. The main research and innovation as follows:1. A similarity measure based on subsequence is proposed, distances between Series are obtain, which is a better measure for sequence clustering as compared to Euclid distance and so on.2. A distance measure based on Markov transferring matrix is proposed, which is focus on the status transferring statistics.3. On base of SAS Application module, a stock analysis module is developed, whose functions including Importing Data, Graphics, Volatility and Risk, Clustering, and Portfolio.4. Predicting Time Series with ANN corrected by ARMA.