You’ve only just wrapped your mind around the concept of big data. Now there’s machine learning, deep learning, machine vision, neural networks, and artificial intelligence. What do these things mean? And, more importantly, what do they have to do with the business of transportation?
Quite a lot, as it turns out. In a recent survey of 433 senior executives in transportation, logistics, and supply chain industries conducted by Forbes Insights, 65 percent say their businesses are undergoing “tectonic shifts” driven in part by rapid advances in technology — specifically telematics, data mining, artificial intelligence, and machine learning.
To help you decode these concepts and understand how they will affect your fleets, we’ve put together a series of articles, starting with this one on data mining and machine learning. Other articles in this series will cover how fleet operations can benefit from advances being made in machine vision systems, how anomaly detection can significantly improve fleet safety, and whether artificial intelligence can improve driver performance. Let’s get started.
What is Data Mining?
Data mining techniques are used to analyze large sets of data (sometimes referred to as “big data”) to reveal patterns, trends, and correlations. This exercise can often yield clues for how businesses can improve their performance. Many fleets, for example, comb through data on near collisions in an effort to predict and prevent future collisions. Others mine location and time-of-day data to find hot spots and reroute vehicles to avoid delays or collisions during certain times of the day.
How Data Mining Relates to Machine Learning
As telematic sensors proliferate in and around commercial vehicles, fleets are quickly collecting vast quantities of information, and data mining is one way to extract value from it.
But it’s not the only way to leverage data. Increasingly, fleets are turning to machine learning to get even more out of their data and gain greater competitive advantages. These algorithms can be used to make faster, more intelligent decisions that can automatically optimize and deploy resources. Imagine a system that automatically assigns loads, adjusts schedules, and routes vehicles based on real-time changes traffic, weather, customer requests, driver proximity, and other variables.
“Data mining looks to discover relevant information from a larger dataset,” said Stephen Krotosky, manager of applied machine learning at Lytx, “whereas machine learning is focused on designing algorithms to make predictions on the data. The two are intertwined, as the output of data mining is often used as the training data for machine learning algorithms.”
In other words, data mining can be used to develop smarter, more accurate algorithms that can “learn” from additional data.
Advancing Fleet Safety and Operational Efficiency
Machine learning can make short work of large amounts of data — serving up in seconds or minutes predictions and recommendations that would have taken days or weeks in the past. That’s good news for anyone who’s ever felt overwhelmed by the sheer volume of data being generated today from all corners of the industry.
But as advanced as machine learning has become, it hasn’t come close to displacing humans. Human intuition continues to play a vital role in operations and strategy. Both data mining and machine learning are best leveraged in organizations that combine them with human judgement.
At Lytx, for example, events are reviewed by professionals who know what to look for. Each event is combed by trained human eyes and tagged for potentially dozens of driving behaviors and conditions. This analysis, combined with traditional telematics data, surfaces far more insights than machine analysis can alone, Krotosky said.
Lytx clients aren’t the only ones to benefit from the combination of human and machine analysis. The company’s researchers, looking to build the next generation of video telematics technology, also stand to gain.
“The result [of this combination of human and machine analysis] is a rich dataset that can be leveraged to build models to detect and predict targeted tasks, such as cell phone use or rolling stops,” Krotosky explained. “It can also be used to evaluate fuel usage, delivery efficiency, and so on.”
The ultimate goal of Krotosky and his team of researchers isn’t to develop cleverer algorithms. It’s to create technologies that help fleets improve the safety of their drivers and advance their operational efficiencies to thrive in today’s ultra-competitive environment. To do so, they have another instrument in their toolbox: machine vision.
Learn how advances in machine vision systems have the potential to transform fleets in the next article in our series.