Feature Engineering Methods on Multivariate Time-Series Data for Financial Data Science Competitions

Financial data are often available in the form of time series. These time series are often highly dimensional with complex relationships between them. The complexity of financial data can be demonstrated in different aspects. Firstly, training data are often limited and the number of features that researchers can create is often much greater than the number of observations. In some research, such as [1], the ratio of the number of features over the number of observations, defined as model complexity can increase up to hundreds for financial instruments with a limited amount of history. Traditional setups in machine learning are not well-equipped for these data-scarce environments.