Using text analytics to assess state-of-the-art machine learning in manufacturing

Literature review is crucial for any study because researchers can discover what knowledge, problems and methodologies exist related to the topic. However, it is time consuming to summarize and find helpful articles to conduct the research because of the large amount of articles and search results may not always be good enough. A method is proposed to score and rank articles, so that potentially influential articles can be found in a shorter time. Using scores of articles and author-provided keywords (APKs), a weighted network can be built to visualize methods and problems. Given a problem, the weighted APKs network highlights suitable benchmarking methods, and vice versa. We implemented this methodology to analyze applications of machine learning in manufacturing. Popular techniques and problems and a list of recommended articles are provided as results of experimentation. The results also show that neural networks and support vector machine are most commonly used, and there is not article used random forest.