A computational referencing approach to stocks correlation analysis

Abstract

The activity of the stock market is dynamic and complicated, with financial figures changing every minute. Thus, it is not ever an easy task for a professional investor to discover the beneficial knowledge that will advantage his investment intuition.

Nevertheless, various studies have proved that a stock's volatility is always associated with several financial factors, even though the association is versatile and the strength of the association varies from one to another. By finding and referencing these factors, investors can either be led to profit, or misled to loss. It follows that stocks correlation presents an often positive influence to future stock movement prediction. Unlike typical technical and fundamental stocks correlation analysis, we developed in this research a hybrid method on the basis of technical analysis with attention to the theory of fundamental analysis. In other words, we deploy a mathematical model to numerically measure stocks correlation in parameterization of a selected fundamental economic factor. The approach promotes a two-tier correlation computation architecture. In the top tier, a Pearson product-moment correlation is derived to measure the direct connection of a stock to a pre-defined referencing factor. Next, the obtained results are used for referencing featured stocks correlation modeling. In contrast to traditional stocks correlation analysis, the employment of a referencing approach for computational correlation knowledge discovery will enhance the accuracy, credibility and intelligibility of the interrelationship between each pair of stocks.

This is because those pre-defined referencing factors are characterized with remarkable stability and reputation in the stock market as well as global economy.

We have applied the proposed referencing model to stocks correlation analysis of S&P 500 for the period of January 2000 to December 2011. In our case studies, we use the U.S. crude oil price and the S&P 500 indices price respectively as the referencing factor to compute correlation of a selected family of stocks, with them being compared to the results of same correlation analysis applied to random selected stocks within the scope of S&P500.

The performance of the proposed model is demonstrated clearly through both numerical and visual observations generated from the experiment. The correlation knowledge extracted by the proposed approach will give investors a different way to interpret stock volatility in order to strengthen their investment confidence. However, the selection of referencing factors is subject to criticism on the ground of subjectivity and the arbitrary nature of the selection process - unavoidable problems in this area of research. This leave us a future research direction that, by employing multiple referencing factors to ensure that the outcome developed by stocks correlation analysis is utterly impartial.