Evaluation of probability of market strategy profitability using PCA method

Abstract

Principal component analysis (PCA) is a commonly used and very powerful method for simplification of multivariate data. In the presented contribution, it is, however, utilized in a novel way. The methodology described in the paper provides tables of probabilities of market strategy profitability. Values of the first two principal components are utilized in evaluation of conditional probability. The resulting probability tables can help to detect buying signals for profitable trade opportunities or can provide a warning against a high financial loss. A time period of two weeks before an actual buying signal is analyzed. After one week of holding, the share is sold and the resultant relative price differences determine the group of profitability (high profit, profit, loss, high loss). For each group, tables of occurrences and conditional probabilities are evaluated using the PCA method. The whole methodology is tested on a case study that takes in account one year long dataset (1st Aug 2014 – 14th Aug 2015) of L-3 Communications Holding shares.