With the globalization of trade, logistics reliability is critical as transit time irregularities lead to more delays further down the supply chain and drive up overall costs. The Client, leveraging the power of the AI & Analytics Engine, is able to provide its customers with near real-time predictions on the probability of departure delays via their platform.
PI.EXCHANGE worked with a cloud-based real-time logistics visibility and execution platform (the Client) to power an additional intelligence-based feature for their customers. The Client provides complete visibility throughout the entire supply chain on one platform, enabling their customer to make better logistics decisions.
The objective: The Client wants to enhance their platform with an automated near real-time prediction capability. This would enable the Client’s customers to select the appropriate carrier and vessel and coordinate their supply chain efficiently and cost-effectively.
- 90% of the world's goods are transported by container
- Only 50% of ships arrive within 24hrs of schedule
- Schedule unreliability increases cost across supply-chain
(Predicting Shipping Time with Machine Learning, MIT, Krempl, Jonquais, 2019)
The Client's customers are companies that require logistics and supply chain intelligence. The Client is looking to solve a known pain point for their customer: “Is this vessel usually on time” and “Is this carrier largely reliable”.
PI.EXCHANGE was tasked with providing an additional predictive overlay via the AI & Analytics Engine.
Using PI.EXCHANGE’s AI & Analytics Engine, the Client could quickly prepare historical data on carriers and deploy the recommended optimal model that would predict the probability of being late within 4 categories; on time, 1 day, 2 days or 3 or more days late.
The AI and Analytics Engine uses Machine Learning (ML) to develop a model capable of improving the predictability of shipping departure times. This was accomplished with two years worth of historical data. The data was comprised of the attributes associated with each trip between a loading port and a discharge port.
During the data preparation phase, the Engine provided recommendations, guiding the user on actions to prepare the data for modelling.
A calculated column was added for the difference between the actual time of departure and the expected time of departure. The calculated column results were then categorised into categories: early or on time; 1 day late; 2 days late; and 3 or more days late.
Once the target column was identified, the Engine’s model recommender provided recommendations on the optimal model, given the prepared dataset.
The model can be deployed with one click to seamlessly integrate with the Client’s Business Intelligence Platform, providing near real-time predictive insight of departure delays to its users.
- 80% accuracy of predicting the probability of delay across all categories.
- 10 minutes total time taken to clean and prepare data, select the optimal model, train and deploy the model to the endpoint.
Using historical data, the Engine was able to build a model that reliably predicted the probability of a carriers departure delay.
For a related Case Study on using AI/ML in the Shipping Industry, check out this article!
These rapidly delivered and on-demand insights would allow the customers to make better decisions on vessels and carrier selection to mitigate negative downstream operational effects on the supply chain.
The above confusion matrix provides a measure of the model's performance
“We are impressed by PI.EXCHANGE’s capability as a self-management machine learning platform with such a great integration and real-time data processing capability.”
- Senior Product Manager
The value of the PI.EXCHANGE’s AI & Analytics Engine extends beyond immediate customer benefits. The Engine as a tool, empowers internal teams to support a sustainable AI/ML capability, providing an additional competitive edge in a highly competitive and rapidly evolving industry.
Using the AI & Analytics Engine, the Client is able to build a repeatable collaborative ML development workflow for their data science team. The value of which extends beyond the use-case at hand, and supports a continued competitive advantage through a robust and sustainable AI capability. Read on for more AI/ML industry solutions.