Abstract

Data science algorithms are on the way to becoming an integral part of every company, and we can already see the effects in many corporations that have invented their own data science teams and also implemented the latest data science algorithms. To be able to work with all the different challenges that are emerging, new powerful data science tools have been developed (e.g. Python, R, H2O, Weka, Tensorflow, Spark, Flink, BigML or KNIME). One balance that companies that want to use these new tools have to face is the cost of implementation vs. the enhanced development that they give in return. Nowadays, most of the advanced algorithms are open source and available on multiple platforms and programming languages, which helps to minimize the development cost challenges that each company has to overcome.

Still, one of the main dangers lurking inside these development teams is that they do not know what the state of the art of advanced algorithms is and which problem they can address. To help mitigate this problem, a review of algorithms has been implemented in this paper. This review gives us a perspective on which algorithms are being developed and which problem areas they can address. With the development of more powerful data science algorithms, we are also enabling the possibility of tackling more complex and interesting problems. However, one characteristic of the review is that there are missing algorithms from the many that are currently being produced and frequently selected by the community as the best performers in many benchmark datasets.