Machine Learning and Quality Management of Quantitative Data - With an Application to Energy Finance
24 Pages Posted: 25 Jun 2019
Date Written: June 21, 2019
In this study we propose a new application of Machine Learning techniques, namely the quality management of quantitative Financial data. The datasets in this field can include, amongst others, bootstrapped forward curves or volatility surfaces. They are provided by quantitative analysts and form the basis of trading decisions. Thus, they require very thorough quality management and plausibility checks. This tedious and non-trivial task is time-consuming and can only partially be assisted by traditional monitoring jobs. To address this problem, we propose a general framework of Machine Learning algorithms that learns from verified quantitative data and evaluates the validity of new data objects. The advantage of our approach is that it does not necessarily require domain knowledge of the data or additional model assumptions or fundamental insights. We will illustrate our idea with a study on bootstrapped forward curves from power markets and Energy Finance.
Keywords: quantitative data quality, machine learning, unsupervised learning, autoencoders, financial bootstrapping, energy finance, HPFC (Hourly Price Forward Curve)
Suggested Citation: Suggested Citation