Educate Personnel on Machine Learning Fundamentals and Set Realistic Expectations Early On.
Case Studies from Five State DOTs Provide Insights on Applying Machine Learning to Various Problems and Challenges DOTs May Face.
Nebraska, United States
California
Iowa, United States
Missouri, United States
Delaware, United States
Implementing Machine Learning at State Departments of Transportation: A Guide
Summary Information
Machine Learning (ML), an emerging field over the past two decades, can “analyze vast datasets, discover patterns, make predictions, and continuously improve through experience represents a paradigm shift compared to traditional rule-based methods that agencies have deployed for decades, and therefore, offering the opportunity to improve transportation system performance and agency operation.” To better help agencies become aware of ML’s potential benefits and challenges, this study aimed to create a guide for state DOTs and other transportation agencies to establish their agency’s ML readiness and capabilities. Through interviews, the study documented five case studies based on the ML experiences of state DOTs from Nebraska, California, Iowa, Missouri, and Delaware.
The study gathered information from five case studies on ML applications through interviews with state DOTs. Some key lessons learned are presented below.
- Educate Personnel on Machine Learning Fundamentals and Set Realistic Expectations Early On. Providing foundational knowledge about ML and managing expectations helps prevent dissatisfaction among operators, leadership, and other personnel stemming from misunderstandings about ML model operations.
- Start with low-cost, low-risk applications. Before making big investments in ML first start with a low-budget project that has clear foundational benefits and demonstrate the benefits of ML in these cost-effective pilots with well-defined scopes.
- Use relevant metrics to communicate the results. ML methods have success metrics that are new to DOT staff. Make sure to communicate the performance of the ML methods using the commonly accepted metrics for stakeholders to immediately grasp the benefits.
- Assess the data storage capacity, type, and cybersecurity considerations. ML methods require vast amount of data for training its models. These data need to be stored, maintained, and kept safe. Agencies must therefore deploy data management plans that address capacity, type, and cybersecurity considerations when dealing with large quantities of data.
- Scaling-up an application requires careful planning and integration between data sources. An ML application could be scaled into a larger deployment by location, time, user base, and/or scope. However, it is important to ensure that data availability, quantity, and quality meet the needs of the ML application.
