State-of-the-Art Machine Learning Frameworks for Training or Inference on Business Process Dataset

Machine learning (ML) is being used to solve complex problems. With the recent emergence of machine and deep learning (DL) architectures, the number of available frameworks to choose from has also increased. The different frameworks have different strengths and weaknesses. Therefore, it is crucial to ensure the highest accuracy of ML/DL models by choosing the right framework for the problem and the data. There are two main categories of frameworks. Those which are feature rich and have been optimized for training, and those which are fast, lightweight, and have been optimized for inference. Training refers to the process of teaching a model to learn from the data it sees. Inference refers to the process of using a trained machine-learning algorithm to make a prediction. Business process data enables the monitoring of events related to the system and is an important source of information for future decisions. The public dataset from a business process that is ready to be processed by machine learning algorithms is used for research. This paper compares and evaluates the features and benefits of various frameworks for training or inference. In this paper, the important metrics are discussed when considering and choosing an ML/DL framework with regard to its limitations.