Using Economic Calendar in Financial Time Series Forecasting

Quantitative analysis of the economic calendar allows agents of the foreign exchange market to make statistically correct decisions and realize profitable trading opportunities. In this paper, we analyze the returns of the NZD / USD exchange rate at various timeframes (M5, M15, M30, H1, H4, D1) at the time of the release of important macroeconomic indicators. The period from 2018 to 2019 was considered. Two trading strategies based on the discrepancy between actual economic indicators and forecast, and past indicators were developed and tested. The predictive power of strategies was tested using the XGBoost machine learning algorithm with an output binary classification parameter and an output linear approximation parameter. Useful metrics have been obtained that can be used in the future for optimizing high-frequency trading strategies, developing trading robots and advisers. Based on the boxplots, showing the profitability or loss-making of signals within the framework of economic news, it is possible to limit the risks of losses by placing optimal stop-loss orders. Consumer price index, employment change, GDP and interest rate related to New Zealand proved to be stably profitable at almost all timeframes.