Data Science Methodologies – A Benchmarking Study

There are several Data Science methodologies that entities and organizations have daily contact with however real-time decision support is seen as a decisive factor for success in making a decision. Due to the complexity, quantity, and diversity of data currently existing, a set of Data Science methodologies has emerged that help in the implementation of solutions. This article arises, fundamentally, with the purpose of answering the following question: What is the most complete and comprehensive data science methodology for any Data Science project? In carrying out this article, twenty-four methodologies were found and analyzed in detail. This study was based on a comparative benchmarking of methodologies, consisting of three phases of analysis, a first that evaluates and compares the phases of all the methodologies collected, a second that analyzes, compares and evaluates the cost, usability, maintenance, scalability, precision, speed, flexibility, reliability, explainability, interpretability, cyclicity and the support of OLAP technology by each methodology, and a third phase where the previous evaluations are compiled and the methodologies with the best results are returned. Quotes. After the three analyses, the methodologies that stood out the most were AgileData.io and IBM – Base Methodology for Data Science, however both obtained a quotation of 63.03%, which demonstrates a low percentage compared to the requirements.