BigML

Machine learning made beutifully simple for everyone — BigML provide a user-friendly platform for machine learning that allows corporations to easily integrate artificial intelligence in their business and processes.

How we collaborate?

Cintra US and BigML collaborate since 2018 to create accurate traffic predictions on our assets in US. The most important collaboration between Ferrovial and BigMl is named RTPF (REAL TIME PROPENSITY FACTOR), this system combine sensor Network and vehicle Intelligence Data and Dynamically manage toll road pricing to maximize capture rates.

Modernize Managed Lane Analytics with Dynamic Pricing 

Cintra leverages the innovative Machine Learning based Real Time Propensity Factor solution powered by BigML, which offers highly accurate predictions to benefit toll road operators that are looking to: 

• Respond to brief traffic pattern changes
• Increase traffic capture and toll revenue
• Act automatically without human monitoring or intervention

Business Problem

Traditional toll road management systems are based on a toll rate curve that is set nightly, and remains static throughout the day. BigML’s RTPF solution employs a custom Machine Learning workflow to detect anomalous traffic patterns dynamically and suggest toll rate adjustments in real time. To enable this, each entry point (i.e., gantry) is modeled separately to predict capture rates in optimal fashion.

Solution & Benefits

BigML RPTF has been implemented at the LBJ TEXpress and its 8 mile network of managed toll lanes in Dallas, Texas. LBJ TEXpress is part of larger TEXpress network of 100 miles of roadways in the Dallas-Fort Worth metro area. 500,000 vehicles travel on LBJ TEXpress toll lanes on a daily basis generating over a million events to be recorded and analyzed. The toll rate for managed lanes increases with congestion on general purpose lanes.

Key Features

• RTPF users can view the transformed gantry data and start new Machine Learning workflow executions after spot checking the underlying data. Key metrics include capture rates, average speed, total volume, and occupancy.RTPF users can easily manage all existing executions in various states of completion in the same place.
• RTPF users can visualize the Machine Learning models for each gantry, direction and segment combination for utmost explainability to enable the quick spread of insights revealed among operators and relevant workgroups.
• Time series visualizations depicting pertinent variables like capture rates, occupancy and related anomalous data points per gantry are accessible on-demand.

Focus on Business Goals, not Infrastructure Concerns

Deployment: RTPF can be deployed on-premises or in any cloud service (AWS, Azure or Google Cloud, A1 Digital). RTPF installations are auto-deployable via Docker containers and multiple instances can be activated in parallel to serve different concessions, departments, business units or subsidiaries. RTPF runs on Linux servers with industry standard CPUs. GPUs are optional but recommended for large image volume applications with low latency requirements.

Auto-Scaling: The RTPF backend provides autoscaling capabilities out-of-the-box to optimize the utilization of available system resources autonomously among multiple users, tasks and servers.

Administration: Admins can manage access tokens, user accounts, user groups in a granular fashion. periodic tasks and traceability reports. System resource utilization reports let admins monitor performance based on CPU/GPU and memory usage, cache size, uptime and mean response time. Execution logs are available for traceability and more detailed diagnostics purposes.

• For more information visit BigML Website.