Exchange traded fund day-trade investment strategy formulation based on knowledge discovery

There are numerous studies utilizing conventional APRIORI algorithms to extract information from data of mass volume. Unfortunately, the conventional algorithms are designed to handle the information in the same transaction or on the same transaction day, and thus are not suitable for predicting the trend of a market. This paper utilizes a modified APRIORI algorithm called Multi-Dimension Non-Continuous (MDNC) to eliminate the limitations imposed by traditional pattern matching of continuous data before mining the associated rules in the cross-day discrete trading data for formulating exchange traded fund (ETF) day-trade investment strategy. This paper further capitalizes on characteristics including low cost and tax and trading flexibility associated with ETF to develop the day-trade strategy with high probability of positive investment return. Our model verification suggests that the approach proposed by the present paper outperforms Random Walk investment strategy, in terms of the investment return and risk level notwithstanding the overall economy.