With rapid advances in computing, data science, and big data, there has been a corresponding massive expansion of datasets that are acquired, recorded, and stored and thus available for individuals and organizations to turn into actionable information. Traditional statistical methods, data mining, simulation techniques (e.g., system dynamics and agent-based modeling), and machine learning (ML) are all being used to process these datasets and produce information for decision-making. Categories of ML include supervised learning, unsupervised learning, reinforcement learning, natural language processing, computer vision, and deep learning. Transportation agencies that wish to leverage ML tools and techniques to extract information for decision-making and system operation need to understand and implement these tools as part of their business processes.
In NCHRP Project 23-16, “Implementing and Leveraging Machine Learning at State Departments of Transportation,” Old Dominion University was asked to develop research products on how to select and implement appropriate ML techniques, by identifying promising applications; by assessing costs, benefits, risks, and limitations; and by developing a roadmap for an agency ML program that includes implementation best practices. The research team (1) performed an in-depth literature review and survey of state DOTs on the state of the art and the state of the practice on ML; (2) conducted case studies with five state DOTs to document existing and potential near-term ML applications; (3) compiled illustrative available codes and tools for major use cases of ML; (4) developed a guide on how to select and implement appropriate ML techniques; and (5) developed a briefing document, along with a set of presentation slides highlighting the opportunities offered by ML.
NCHRP Research Report 1122: Implementing Machine Learning at State Departments of Transportation: A Guide presents a 10-step roadmap to building agency machine learning (ML) capabilities to (1) educate state departments of transportation (DOTs) on ML applications in transportation; (2) help assess costs, benefits, risks, and limitations of different ML approaches; and (3) assist in building a data-driven organization that can effectively utilize ML. To develop this guide, the research team conducted a review on the state of the art and state of the practice of ML, developed case studies for a variety of data environments, and compiled code examples for major use cases of ML. This publication will be of interest to state DOTs and other stakeholders by providing a foundational understanding of ML and its applications and helping to leverage ML for a safer, more efficient, and sustainable transportation future.
In addition to the research report, the following deliverables are available:
