Large data transfers have become more critical with the increasing volume of data in scientific computing. For scientific user facilities dedicated to support such large data transfers, accurately predicting network performance is essential for workflow scheduling and resource allocation. In addition, detecting anomalous situations is also a crucial function for reliability. This research explores a data-driven, machine learning (ML) approach to develop these functionalities using a large set of network traffic monitoring data from the tstat system. We are particularly interested in network performance prediction and anomaly detection at this time, may expand to additional targets as time permits.