16 APAC / Issue 6 2018 , Forecasting is an integral part of supply chain logistics. The predictions start when a vessel leaves Port A for Port B. Will it arrive on a certain date? At what time? How long will it take to pass customs? When should the trucks be booked for pick up? Each question is affected by a decision made earlier in the supply chain. Each decision is based on data about the individual links in the chain. In the past, industry veterans carried this data in their heads, making predictions on the fly. Today, that’s just not possible. The ships are bigger and more frequent, the cargo exponentially increased. There are too many terminals to remember the quirks of each one, the variations that can lead to delays. There is just too much data to process. From the IoT to AI via ML The Internet of Things (IoT) is responsible for this amazing amount of data. It has tagged and networked everything possible, from container weights and dispatch dates to vessel speed and weather conditions – all in real time. This raw data is then cleaned up and processed, which is where machine learning and artificial intelligence come in. Machine Learning, AI, the IoT and 1-Stop Connections Machine learning (ML) is the engine room of Artificial Intelligence – it’s where the data gathering and analysis happens that leads to the non-human decision-making we call AI. It recognises patterns in the logistics chain that humans would not. An AI platform can predict that a container leaving a port six weeks from now is 65% more likely to go missing than normal. An alternative route can be found and recommended within minutes. Machine learning is all around us ML is at work right now on your email account. For example, the focused inbox option in Outlook Mail is powered by machine learning. It learns through explicit and implicit actions. So moving an email to the Focused Inbox is explicit, while replying to lots of emails from one person implies that person is important. ML uses data from previous decisions to make predictions about future decisions. The same techniques can be used in the port community, which is why Adam Compain founded ClearMetal, a predictive logistics company that uses AI to unlock supply chain efficiency. He describes predictive logistics as “using sophisticated technology to solve the most 1706AP06 complex operational problems in shipping and logistics”. The aim is to help the links in the supply chain – brokers, truck companies, terminals – overcome the overwhelming complexity of the system. “There’s a lot of uncertainty around customer behaviour, cancellations, and stuff like that,” Compain told tech blogger Nathaniel Mott. “There’s also a lot of uncertainty around the market: Will commodity prices change? Will there be broader changes in cost?” Predictive risk and simulation models One of ClearMetal’s tools is Predictive Risk calculation. This helps freight operators understand the likelihood of a shipment moving from “On Time” to “Late” by using simulations. Hundreds of simulation models are generated per shipment to provide the probability of a status change. Once the data is fed in to the platform, the result might be: “87% of 130 simulations predict this shipment will arrive late”. ClearMetal lists four elements needed for this prediction of global logistics in real time: 1. Understand the data: Robust data ingestion capabilities are needed to collate and cleanly structure data from multiple sources. 2. AI/ML : ML offers a fundamentally different approach to interpreting data, recognising complex patterns between parties and events, and generating accurate predictions. 3. Simulation : An engine for running what-if scenarios and analysing all events and outcomes is essential. ML, alone, is often insufficient. 4. Industry expertise : A set of industry-tailored solutions are needed to empower ocean carriers, freight forwarders, terminal operators and shippers to get the predictions they need to make optimal business decisions. Machine learning goes global Every port that relies on international trade is investing in ML. A Californian analytics software company, Maana, is working to mathematically model the process that goes into rerouting a vessel in the event of a port closure or natural disaster. Using traditional methods, it could take eight hours to evaluate the potential of an alternative port. With AI, it can be done in minutes.