Artificial intelligence (AI) and Robotic Process Automation (RPA) are popular buzzwords today, yet the practical applications for and approaches to these concepts are often less understood and (in many cases) may still be on the horizon. However, there is a technology-driven differentiator within this overarching category that has been in use across a number of industries for some time. Machine learning is the means by which a program analyzes a number of data sources and, by using algorithms, “learns” from that analysis to predict likely scenarios or future end-state results. In this way, machine learning can turn large volumes of passive data into active business information.
Machine learning has been used in consumer-facing businesses for quite some time to understand concepts such as consumer actions, seasonal trend information and likely demand for new products. Amazon uses machine learning to drive its well-known “other people who bought X also purchased Y” capability. This type of item-to-item collaborative filtering forecasts likely purchases based on correlations of similar purchases across multiple users with comparable parameters. As another example, Netflix uses machine learning to identify movies, shows and entertainment “you might like” based on individual (and collective) analysis of your past and real-time viewing selections and subsequent ratings. Netflix leverages Restricted Boltzman Machines (RBM) as well as a form of matrix factorization to conduct its analyses and make recommendations.
Machine Learning in Action
Just as these well-known B2C companies leverage machine learning to drive efficiency, B2B executives can take advantage of machine learning capabilities to create value that has the potential to surpass traditional labor arbitrage via outsourcing or out-tasking. A number of machine learning use cases and examples span business industries, which can be considered a roadmap for other companies to use to take advantage of these value levers themselves.
Logistics: Many warehousing and materials planners use a type of machine learning known as Multi-Echelon Inventory Optimization (MEIO). MEIO automatically adjusts inventory positions by optimizing the balance of inventory at the right locations and establishing optimal buffer locations and quantities. Implementing an MEIO driven toolset can drive significant warehousing and materials planning value. And commercial MEIO capabilities from well-known providers such as Oracle, whose data cloud community offers packaged analytics that drive MEIO, make it relatively easy to implement.
Manufacturing: Industrial manufacturing businesses have been using machine learning to sift through the SKU-Locations to identify “clusters” with similar seasonality profiles, improving service levels and increasing inventory turns.
Telecom/Tech: Telecommunications market leaders have utilized machine learning to model and predict likely fraud indicators, allowing real fraudulent behavior to be recognized in its infancy and shut down accordingly. Tech enterprises have also leveraged machine learning deployments to accelerate data center optimization opportunities that drive lower costs and greater efficiencies.
Human Resources: Businesses are also using machine learning to better understand and engage their workforces. For example, by analyzing attributes of successful employees, employers can recommend internal positions that may be a strong fit or identify specific learning and development opportunities for their staff to achieve individual career goals. Also, businesses with large temporary or seasonal workforces are using machine learning to better understand location-specific demand patterns to optimize staffing while reducing overtime and, more importantly, aligning staffing levels with demand to better meet the customer needs.
Machine Learning Solution Considerations
Beyond leveraging third-party machine learning capabilities as a service, there are also opportunities for businesses to develop machine learning capabilities on their own. With knowledge of programming languages such as R or Python already in place, the machine learning tools and approaches commercially available today allow companies to leverage graphical or scripted environments to create their own machine learning algorithms.
Working from within a graphical environment, tools such as Weka provide a machine learning “workbench” that is designed for sandbox/experimental testing, allowing algorithms to be “plugged” together. Alternatively, BigML is a web-based upload that provides easy to use interfaces with APIs, and Orange offers a platform with over 100 widgets and a Python API and library for integrating apps. And, in cases where a more scripted environment approach is desired, SciKit Learn provides a scripting environment and library with machine learning algorithms and data preprocessing. Additionally, Waffles is a collection of command-line tools that are useful in prepping and visualizing data, running algorithms and summarizing results. Regardless of the approach, there are a variety of tools available to help make entry into the world of machine learning an easier process than might be imagined.
Companies that are interested in staying current as well as looking at the future are considering machine learning applications and opportunities to help them achieve their goals. Understanding how machine learning operates, the insights it can provide and the steps to follow to apply it (either via third-party providers or via a homegrown approach) will allow companies to begin to take advantage of the technological capabilities available today and tomorrow.