The future of machine learning and predictive analytics

Abstract

The history of artificial intelligence shows us that there has been a gradual and evolutionary development within the special aspects of computational sciences that underlie the technologies prevalent in machine learning. Most of the technologies consist of methods defined by so-called computational intelligence, including neural networks, evolutionary algorithms and fuzzy systems. Data mining topics have also become more significant, due to the rapid growth in the quantities of data now available (Big Data), combined with having to face the same challenges encountered in the IoT (Internet of Things). The question now is: How can computers be made to do what needs to be done without anyone prescribing how it should be done? Nowadays, a whole range of providers offers frameworks for machine learning. Some of them allow us to use machine learning tools in the cloud. This option is mainly provided by the big players like Microsoft Azure ML, Amazon Machine Learning, IBM Bluemix and Google Prediction API, to name but a few. Machine learning algorithms extract high-level, complex abstractions as data representations by means of a hierarchical learning process. Complex abstractions are learnt at a given level based on relatively simple abstractions formulated in the preceding level in the hierarchy. Deep learning is a subarea of machine learning, and could even be described as a further development of this. While traditional machine learning al- gorithms rely on solid model groups for recognition and classification, deep learning algorithms develop and create their own new model levels within the neural networks independently. New models do not need to be repeatedly developed and implemented manually for each new set of data based on different structures, as would be the case for classic machine learning algorithms. The advantage of deep learning is the analysis and learning of massive amounts of unsupervised data, making it a valuable tool for Big Data Analytics where raw data is largely unlabeled and uncategorized.