Construction of Optimal Stock Market Portfolios Using Outlier Detection Algorithm

Abstract

It is not easy for investors to trade in stock markets as building stock portfolios requires financial knowledge and consumes much time. Thus, this study aims to construct optimal stock market portfolios for investors using a LOF-based methodology. We used an outlier detection algorithm called Local Outlier Factor (LOF) to identify outperforming stocks from a stock pool. We then constructed two portfolios using these outperforming stocks, namely, tangency and equal-weighted portfolios and compared their performance against the benchmark portfolios, namely, the market portfolio and the cash market. It was followed by using Mean-Variance Portfolio Optimisation (MVPO) to measure the performance of the portfolios and determined whether they were efficient. To identify the most efficient portfolio, we used the Sharpe ratio. In general, the results showed that both tangency and equal-weighted portfolios gave better returns than the benchmark portfolios. To conclude, the LOF-based methodology helps to build and identify profitable stock portfolios for investors.