Data Fusion methods and challenges in Machine Learning

Abstract

Data fusion is a prevalent way to deal with incomplete raw data in order to collect reliable, usable and accurate information. A machine-learning framework that automatically learns from past experiences without programming specifically compares a variety of conventional probabilistic data fusion techniques, unintentionally renovates the fusion process via high computer and predictive abilities. However the literature also lacks a comprehensive analysis of the recent developments in data fusion machine learning. It is also useful to study and summaries the state of the art in order to gain a deeper insight into how machine learning can support and optimize data fusion. In this paper, we include a detailed survey of data fusion approaches based on machine learning. First, we give a comprehensive introduction to the context of data fusion and machine learning in terms of concepts, implementations, architectures, processes and typical techniques. Through the literature review, study and comparison, we finally come up with a range of open issues and suggest future directions for research in this area.