Morgan Warren Member Name
Project Geochemist
There’s been a lot of discussions lately about big data across various industries as improvements in technology, computing power, and cloud-based services allow industries to collect and access more data and analyze it more efficiently at lower cost. But the changes in data collection techniques affect more than just the IT sector, with manufacturing, for example, seeing huge benefits. This is largely due to the advent of Industry 4.0, the Internet of Things (IoT), better robotics and sensing technology, improved monitoring capabilities, and artificial intelligence being implemented across factory floors.
Naturally, all this data can be extremely valuable, but it also generates a separate challenge: how does one analyze it? This is where Machine Learning comes into play. The power of Machine Learning is to discover the hidden insights and learnings from enormous amounts of data, which is otherwise not practically possible by humans.
What is Machine Learning and its benefits?
Machine Learning (ML) is the science of applying programmatic and statistical computer technology to analyze large datasets, and as a result, uncover new insights. It is a specialized sub-sector of Artificial Intelligence (AI), which is a larger set of predictive modeling tools that enable a computer program to learn by generalizing from examples.
Any business can gain competitive advantage by effectively accessing their data and discovering invaluable hidden insights. Evolution of hardware and software technology has made it one of the most important technologies to invest in in the 21st century.
Having said that, recognizing the use of ML in different domains is an important first step before investing in it. Just collecting large amounts of data and blindly processing them does not provide any value to an organization. Consulting with subject matter experts, using their domain expertise and data analytics skillsets, is an essential step to evaluate the risks and opportunities of ML investments at an early stage.
Benefits of successful ML use include, but are not limited to:
- As larger data sets become more and more routine, ML tools and techniques can effectively handle these large data sets and identify valuable insights that a human would miss due to the sheer volume of information.
- Enables real-time decision making from constantly evolving big data.
- Allows Continuous Quality Improvement (CQI) in large, dynamic and complex data ecosystems.
- Eliminates manual tasks by using predictive models to deliver automation.
- ML techniques have proven invaluable in optimizing many aspects of business and industry practices, resulting in significant savings in time and cost.
Is it a new technology?
Despite much confusion about the terminology, ML is not a new technology. Artificial neural networks – computer systems modelled after the human brain – were first introduced in the 1950s. One notable example of the idea is the world-famous Turing test, introduced by Alan Turing in 1950. In order to pass the test, a computer must pass for human so effectively as to fool actual humans.
Finance, telecom, and military users were among the early adopters of ML, where it was used for purposes ranging from the recognition of handwritten checks and prediction of financial fraud, to the optimization of telecommunication systems and large-scale military logistics planning. More recent adopters of ML have been industries such as sports analytics, healthcare and pharmaceutical powerhouses, and of course, Microsoft, Google, and Amazon who are using ML technology for their services as well as providing ML infrastructures to customers.
ML for Geoscience Practices
Geoscientists are quickly discovering the benefits of ML in their practices. Among the data set examples where ML is being used in projects include, but are not limited to:
- Optimizing well placement. Using data samples, ML can help to predict the best pumping rate and the best location for a well to ensure the majority of the contaminant plume gets captured efficiently.
- Optimizing land use in agriculture. Analyzing soil sample laboratory results to predict of which land areas might contain optimal levels of major nutrients like nitrogen (N), phosphorous (P), and potassium (K) for agriculture.
- Predicting material properties with hyperspectral imaging analysis. Hyperspectral imaging and data sampling of the mechanical and chemical properties of rocks can be used together to derive relationships between the spectral values in the images and the sampled data to make predictions of material properties such as Rock Quality Designation (RQD) and leachability.
- Optimizing lithological sequencing in mining. The use of ML in mining to predict sulfur content in differing lithologies based on an assay database can help better define lithological units and reduce the number of laboratory tests needed.
The majority of geoscience sample data used for these types of projects falls under two main categories:
Wide – relatively few samples, but many variables
Deep – relatively few variables, but many samples
Using approximately 80% of data available, it’s possible to create a predictive computational model using surrogate or proxy data. So long as relationships are found within this data sample, it can be used to predict future outcomes. Where more data is required, it is often possible to use data from other areas such as chemical datasets, assay databases, and geological logs.
What does the future hold for ML?
Digital transformation has gained significant momentum across multiple industry segments in recent years as cloud computing power and its applications have quickly grown. In addition to evolution of hardware and software technologies, increasing awareness among industry leaders has significantly contributed to the increased pace of digital transformation across organizations. This indicates a great future in increasing the rate of ML application in different industries.
Here at Golder, we’ve observed a significant increase in clients requesting help with ML analysis to increase efficiency and find insights to tackle their challenging problems. By applying our technical expertise and experience to their data, clients have been able to use predictive analysis to make intelligent, more accurate decisions in a timely manner. The ultimate result is substantial, ongoing cost savings for many clients.
As ML grows in popularity, it’s clear that data scientists are still only scratching the surface. Looking into the future, we can expect more people to gain a better understanding of ML, and to learn how to apply these tools. We should anticipate a much wider adoption of ML learning techniques to analyze structured data on spreadsheets, as well as unstructured data images captured from drones, spectral scanning, satellites and so forth. As computer vision grows, new algorithms will be created to analyze imagery. Companies are certain to continue innovating and investing in ML to obtain better results as the technology evolves.
