• DATVF.ATLPHL
    1.643
    -0.074
    -4.3%
  • DATVF.CHIATL
    1.951
    0.018
    0.9%
  • DATVF.DALLAX
    0.880
    0.015
    1.7%
  • DATVF.LAXDAL
    1.501
    0.007
    0.5%
  • DATVF.SEALAX
    0.966
    -0.092
    -8.7%
  • DATVF.PHLCHI
    0.929
    -0.038
    -3.9%
  • DATVF.LAXSEA
    2.005
    0.035
    1.8%
  • DATVF.VEU
    1.508
    -0.031
    -2%
  • DATVF.VNU
    1.395
    -0.016
    -1.1%
  • DATVF.VSU
    1.191
    0.011
    0.9%
  • DATVF.VWU
    1.486
    -0.028
    -1.8%
  • ITVI.USA
    9,836.710
    -180.070
    -1.8%
  • OTRI.USA
    4.790
    0.100
    2.1%
  • OTVI.USA
    9,831.280
    -180.470
    -1.8%
  • TLT.USA
    2.410
    -0.010
    -0.4%
  • WAIT.USA
    150.000
    0.000
    0%
  • DATVF.ATLPHL
    1.643
    -0.074
    -4.3%
  • DATVF.CHIATL
    1.951
    0.018
    0.9%
  • DATVF.DALLAX
    0.880
    0.015
    1.7%
  • DATVF.LAXDAL
    1.501
    0.007
    0.5%
  • DATVF.SEALAX
    0.966
    -0.092
    -8.7%
  • DATVF.PHLCHI
    0.929
    -0.038
    -3.9%
  • DATVF.LAXSEA
    2.005
    0.035
    1.8%
  • DATVF.VEU
    1.508
    -0.031
    -2%
  • DATVF.VNU
    1.395
    -0.016
    -1.1%
  • DATVF.VSU
    1.191
    0.011
    0.9%
  • DATVF.VWU
    1.486
    -0.028
    -1.8%
  • ITVI.USA
    9,836.710
    -180.070
    -1.8%
  • OTRI.USA
    4.790
    0.100
    2.1%
  • OTVI.USA
    9,831.280
    -180.470
    -1.8%
  • TLT.USA
    2.410
    -0.010
    -0.4%
  • WAIT.USA
    150.000
    0.000
    0%
Less than TruckloadLogisticsNewsTechnologyTruckingTruckload

Machine learning in logistics: Separating fact from fiction

Several logistics and transportation software providers claim to have machine learning capabilities, but in many cases, the results don’t match the hype. ( Photo: Shutterstock )

Commentary

 Chris Ricciardi is chief operating officer of Logistical Labs. The views expressed below are solely those of the author and do not necessarily reflect those of FreightWaves.  

 

It seems like every day there is a new headline about machine learning or artificial intelligence (AI).  We’ve seen the technology behind these concepts do incredible things—from customer service chat bots to speech pattern recognition to disease diagnosis.

Several logistics and transportation software providers claim to have machine learning capabilities, but in many cases, the results don’t match the hype. Even when the right technology is involved, getting real value from machine learning takes considerable effort. Given the complexity and an industry rife with buzzwords, supply chain decision makers need a way to separate the experts from the rest of the pack.

 What Is Machine Learning…and What Isn’t?

 Because machine learning is an emerging concept, there may be confusion about what it is and what it isn’t. In simplified terms, machine learning is a way to use complex mathematics to train machines to think for themselves. Instead of instructing computers or machines on how to carry out a task, humans teach machines to imitate human thought processes and then give them access to ample data, which they use to generate better and faster solutions. They also learn from mistakes and improve over time.

Machine learning starts with two sets of data. First, training data gets fed into the machine to teach it what correlations to look for and to create a mathematical model to follow. Then, the test data you want to analyze goes in. This dataset contains the unknowns you’d like to understand better. For example, if you wanted to predict future costs, you’d first need to train the computer on how to do the actual analyzing. After that, you’d feed the all the data you collected on costs into the machine for analysis.

Savvy logistics companies today use machine learning for forecasting, real-time decision-making, optimizing fleets, preventative maintenance, and more. Yet, it’s important to clarify that machine learning is not a summary of historical data. That’s simply a report. Machine learning is also not a one-size-fits-all solution you can simply plug in, turn on, and start using. Instead, machine learning requires a cooperative effort between skilled data scientists and business leadership, who painstakingly select and validate the right data, and choose the best self-learning algorithms to meet your particular needs.

 4 Ways to Know True Machine Learning When You See It

 Machine learning is indeed complex. The good news is that you don’t need to be an expert. But you should know enough to choose a qualified provider. Here’s how to get the right insights:

1. Scrutinize their expertise. Machine learning is not easy to implement. It requires working knowledge of complex science and math, and data science subject-matter experts must be involved in solution development.

2. Ask how the solution works. Find out what data is required and how it’s used. Can the vendor clearly articulate this? How is the machine trained, and how does it ultimately “learn” patterns? What kind of subject-matter expertise is needed to set up the solution properly?

3. Request proof. Can the vendor show you results and proof of experience in your industry segment? Have they solved the problems you need to solve? How about case studies they can point to? What metrics do they use to track predictive accuracy?

4. Be wary of buzzwords. Most importantly, know the signs of hype. Vague statements, overusing buzzwords, and avoiding specifics should be red flags in your search for the right provider. If the vendor can’t translate the problem into specific hypotheses and tests, you might be better off looking elsewhere.

Knowing what to look for is the first step in finding the right machine learning solution for your business, even if you’re a novice. Solution providers that can share real-world, tangible examples are more likely to be the real deal. 

 Logistical Labs builds innovative technology for the logistics and supply chain industries. Their pricing and mode optimization platform, LoadDex, simplifies transportation pricing and carrier selection across all modes through data-driven insights and social collaboration. To learn more, visit www.logisticallabs.com.

Tags
Show More

2 Comments

Close