Industrial Machine Learning Ushers in New Era of Analytics
As the volume of available data grows, companies need more powerful tools to gain insights at enterprise scale.
by Jerry Overton
Many large enterprises realize that their data can give them rapid, useful insights for implementing changes that will benefit their business. But as the volume of data grows, it becomes more difficult for these companies to extract meaning from it.
The solution is industrial machine learning (IML), which can consistently produce data-driven insights at enterprise scale. IML can ingest data, build algorithms, deploy them into production and generate continuous insights into ongoing business problems.
A modern twist
IML is a modern take on a very old idea: the scientific method. Data scientists start with a hypothesis and collect data that could be useful in evaluating the hypothesis. They then generate a model and use it to explain the data. They evaluate the model’s credibility based on how well it explains the observed data, and continue to evaluate it based on how well it explains new data collected later. When it comes to discovering insights, this method works consistently well.
The modern twist uses digital infrastructure to apply this method in an enterprise, collecting data through a continuous pipeline and using business algorithms running in production as the models. Experiments are done in very short sprints that force data scientists to focus on discovering insights in small, meaningful chunks.
IML for healthcare
In healthcare, insights about hospital procedures can help improve patient care and hospital outcomes. We used a data strategy based on IML to supplement hospital administrative data with rich information from the healthcare provider, including electronic patient records and other routinely collected data.
We looked for features that were most important in predicting length of stay for patients undergoing hip or knee replacements. We found key leading indicators, such as the patient’s age, their core healthcare providers and secondary diagnosis. From these, we built a regression model, which allowed us to predict how long a patient would stay in the hospital.
Those predictions became the basis of operational dashboards that alerted hospitals to future costs and helped identify patients who might experience problems in recovery.
IML in energy
In mining, smart equipment management can save a lot of money and improve operations. Thousands of factors affect the performance of complex machines, but by using a data strategy based on IML, we can monitor operations and predict machine problems before they happen.
In a study, we found that the main causes of unscheduled machine maintenance were time spent waiting for other processes, crew meetings and training. The mining crew tends to spend downtime maintaining the equipment; it’s helpful to create a new maintenance category to track and give credit for these opportunistic maintenance events. Equipment damage happens after the crew spends a lot of time waiting due to mine blasting delays: The blasts create dust and, over time, the dust accumulates in the machinery. We found that some failures could be prevented just by alerting operations when a machine logs too many hours waiting due to blasting delays.
While these two examples illustrate the thinking and process of IML, the technique can be applied to businesses in all industries as a way to improve data-pipeline management, optimize applications and networks, enhance security and incident management, and streamline other tasks.
Potential projects include:
Retail and e-commerce
- • Optimizing supply chain
- • Managing inventory
- • Fueling recommendation engines and real-time, smart coupon delivery
- • Enabling 360-degree views of customers
- • Enhancing clickstream and social media analysis
- • Optimizing ad delivery
Communications and media
- • Enabling smart metering and billing
- • Optimizing antennas
- • Monitoring equipment
- • Enhancing fraud detection
- • Monitoring risk
- • Improving mobile transactions
- • Enabling intelligent alerts and notifications
- • Enhancing fraud detection
- • Improving underwriting and reducing risk
- • Enabling dynamic pricing models
There is no shortage of data: Organizations must now produce reliable, data-driven business insights at enterprise scale, or find themselves at a serious disadvantage. This is the beginning of a new phase of big data, one that has little to do with data capture and storage — and everything to do with producing understandable and useful insights.
JERRY OVERTON is a CSC Distinguished Engineer and head of advanced analytics research in CSC’s ResearchNetwork.
BEN BRIDGEWATER, senior principal industry strategist in healthcare, and REBECCA POE, business manager for CSC’s healthcare and life sciences research network team, contributed to this article.