For more information, visit Labelmaster.
Archives for December 2016
By: Vamsi Madabhushi and Swaroop Gooty
As 2016 comes to a close, it is time to reflect upon our achievements and identify opportunities for next year. However, it is also a time for us to ponder upon the future – which is filled with a great promise of change (even in Supply Chain Operations!).
A change so transformative that, soon, machines will learn and execute many tasks that we do today to meet our promise to the customer.
What can machines possibly learn and how can they help supply chain organizations? – In this post allow us to untangle the mystique behind the applicability of machine learning systems to Supply Chain Execution-Operations.
What is machine learning (ML)?
Machine learning is a branch of artificial intelligence in which algorithm based models are developed. The models learn relationships in data and can be used to predict or investigate relationships of a dataset.
For large dimensional data sets it is often difficult for a human analyst to develop models whereas ML based models excel in their ability to use vast amounts of data.
For model building, the algorithms require data which is a past evidence of a certain pattern i.e. training data. ML models have been successfully applied in computer vision, search engines, bioinformatics, and financial analysis etc.
Suggestions or predictions from machine learning algorithms are inferred based on the data used to train a model. The learning happens by providing training data to the algorithm. This type of learning is known as supervised learning.
For example: If we provide historic data with trailer pick-up and drop-off times, location, routes, driver information, trailer characteristics etc. – the algorithm will be able to learn from the data and predict transit time of a new load going to a particular destination.
How does machine learning work?
The picture above shows different steps involved in developing a ML model. In our view, the most critical step is to define the business problem.
Problem Statement: It is important to understand the kind of problems that machine learning can solve. Machine learning is suitable for scenarios where a number of factors could influence the model outcome/prediction and where rule based solutions are not viable.
Analyze, Validate & Prepare Training data: Using the data that is already available in the enterprise the model can be trained. Determining the right set of classification labels, independent variables and correct training data is the basis for developing a good machine learning model. In certain scenarios where the data is not available in a particular format, it may need some transformation.
Develop ML Model: The next step is to choose an appropriate algorithm that will learn from the training data and identify patterns. The model is iteratively tuned to achieve an acceptable level of precision.
Model Validation: A test data set that is kept separately from the training data i.e. data set that the model has not seen previously can be used to validate the model. The model is evaluated on its ability to recall patterns and on its ability to predict precisely.
Deployment & Periodic Evaluation: Model deployment requires some level of integration with the enterprise systems. Once a model is developed and deployed, it requires periodic evaluation. During evaluation, new patterns are identified and the model is trained on those. This fine tuning of the model makes it more robust and increases the precision of the model output.
Why use machine learning in Supply Chain Execution (SCE)?
Adopting advanced technologies can be a strategic lever to achieve operational goals. Using data commonly available in the enterprise coupled with machine learning models can turn out to be a game changer.
SCE activities such as order orchestration, fulfillment, DC space planning, and transportation management are, today, supported by packages or custom applications.
Yes – they get the job done. For example, a warehouse management application can be configured to achieve efficiency in order picking and shipping. However, it cannot tell you whether there is too much idle time of forklifts in specific areas or whether the door utilization this week is lower than expected.
The next wave of systems need to be intelligent and prescriptive in ways that help managers look at their processes differently.
One might ask – I already have real-time visibility of current operations via dashboards and alerts. What else do I need? – The answer is simple – Machine learning based Supply Chain Execution systems.
Most likely, systems that you currently maintain to run your operations are configured to run on yesterday’s business rules. These systems are not dynamic enough to support the ever changing operational environment. Not only is it difficult to change the rules constantly but also it is very time consuming to analyze large amounts of data to identify changing patterns.
Machine learning models precisely address the above challenges.
Where do you start?
Start with identifying problem statements - Like we mentioned above, the most critical step is to identify problems that can be solved through ML models. In this post, we will leave you with a few ideas where ML can be applied to improve your distribution center operations.
Let us know your thoughts.
Vamsi Madabhushi is a Consulting Manager specializing in Distribution Center Operations and Technology. An expert in JDA Warehouse Management System he has developed SCE solutions for leading US retailers. Vamsi is enthusiastic about using data analytics to build efficient and greener supply chains.
Vamsi holds an MBA from Johns Hopkins University.
Swaroop Gooty is a Senior Supply Chain consultant with diverse experience across areas of supply chain execution. He is passionate about using technology & data to improve supply chain operations.
Swaroop holds an MBA from Indian Institute of Management Calcutta, India.
The opinions expressed herein are those solely of the participants, and do not necessarily represent the views of Agile Business Media, LLC., its properties or its employees.