Online Portfolio Selection Using Spectral Pattern Matching

Nowadays, due to the rise of turnover and pace of trading in financial markets, accelerating of analysis and making decision is unavoidable. Humans are unable to analyze big data quickly without behavioral biases. Hence, financial markets tend to apply algorithmic trading in which some techniques like data mining and machine learning are notable. Online Portfolio Selection (OLPS) is one of the most modern techniques in algorithmic trading. OLPS allocates capital to a number of stocks and updates portfolio at the beginning of each period by some techniques. Actually, individual has no role in portfolio selection and the algorithm determines the way of investing in each period. In this article, an algorithm which follows pattern matching principle has been introduced. In pattern matching principle, the portfolio is selected based on identical historical patterns and in this article these patterns are found by spectral clustering in data mining. At the end of article, there is a numerical example which uses the most 20 active stocks in New York Stock Exchange (NYSE) data and its results has been compared with other algorithms in this topic.