Some literature adopted a weighted-scoring approach to construct the multi-factor stock selection model. However, this approach leads to two shortcomings. First, it cannot effectively identify the connection between the weights of stock-picking concepts and portfolio performances. Second, it cannot provide the optimal combination of weights of stock-picking concepts to meet various investors’ preferences. This paper aims to employ a mixture experimental design to collect the weights of stock-picking concepts and portfolio performance data, as well as to build up performance prediction models based on the weights of stock-picking concepts with multi-variable polynomial regression analysis. Furthermore, these performance prediction models and optimization techniques are employed to discover the optimal combination of weights of stock-picking concepts. The samples consist of all stocks listed in the Taiwan stock market. The 1997-2008 period and the 2009-2015 period are employed as the modeling period and the testing period. Empirical evidences showed that (1) our methodology is robust in predicting performance accurately, and can discover significant interactions between the weights of stock-picking concepts. (2) It can discover the optimal combination of weight of stock-picking concepts which can form stock portfolios with the best possible performances to meet investors’ preferences. Thus, our methodology is able to resolve the two shortcomings of classical weighted-scoring approach.
2018 International Conference on Multidisciplinary Challenges in Business Management and Social Science Theories (MCMS 2018)