Publikation

Estimating Model Analysis for Multivariate Regression Problem for The Mars Express Power Challenge Open Data

Praveenkumar Jayanna

Mastersthesis Technische Universität Kaiserslautern 8/2017.

Abstrakt

Predictive analytics has been an ongoing trend in business analytics. The classification and regression problems are part of the predictive analytics. In this paper, Mars Express Power (MEP) challenge is taken as a case study for multivariate regression problem. The MEP challenge is a competition organized by ESA. The challenge is to predict a Martian year thermal power consumption of Mars Express. For which, the collected data which is composed of context and observed power data of previous three Martian years has to be examined. The main goal of this research paper is to propose a best-fitting regression model for the MEP challenge. To achieve this goal, the paper introduces a data analysis model, namely MEPCM model. This model contains regression model analysis, that addresses the different multivariate regression algorithms. The resultant ensemble of LSTM, XGBoost, and Random Forest algorithms results in the best predictive model to address the power consumption problem.

Projekte

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence