When constructing a portfolio of stocks, an investor should view the returns of the stocks under consideration as a multivariate problem. This paper uses multivariate statistical techniques, multidimensional scaling and principle component analysis, to explain qualitatively and quantitatively the covariance of monthly returns. These techniques yield results that suggest that industry and market factors are necessary to explain the co-movement of returns. Five optimal portfolio selection algorithms are then implemented in the following section. The portfolios selected by the two algorithms that take both of these factors into considerations are found to perform better in subsequent years.