Scalable Generalized Multitarget Linear Regression With Output Dependence Estimation

Nowadays the phenomenon of Big Data is overwhelming our capacity to extract relevant knowledge through classical machine learning techniques. Multitarget regression has arisen in several interesting industrial and environmental application domains, such as ecological modeling and energy forecasting. However, standard multi-target regressors are not designed to perform well with such amounts of data. This paper proposes a scalable implementation for a multi-target linear regression algorithm with output dependence estimation for Big Data analytics in Apache Spark. Our experiments on large-scale datasets show an accurate analysis compared to standard implementation and order of training time reduction as the available number of working nodes in the processing cluster increases.