Unearthed Industry Lead - Crowdsourcing, Holly Bridgwater's presentation from the Copper to the World conference on 18 June 2019 about outcomes of the OZ Minerals Explorer Challenge.
Fertilization: Sperm and the egg—collectively called the gametes—fuse togethe...
New Frontiers - Mining the data for our next copper discovery
1. New frontiers
Mining the data for our
next copper discovery
Copper the the World, June 18th 2019
Holly Bridgwater, Industry Lead
holly@unearthed.solutions
2. We know we need to increase our
discovery rate.
How do we increase our confidence in
the targets we generate?
How do we make the right bet?
6. • The Mount Woods Project is an
area ~ 5000km2 near the
Prominent Hill Mine in South
Australia
• Participants in the competition had
access to the OZ Minerals private
exploration database of >5TB
• The challenge was to use the data
to predict the location of economic
mineralisation of any kind, not just
another Prominent Hill
Crowdsourcing Exploration at Mount Woods
7. What was in the data?
• 621 datasets
• 678 drillholes with 115,000 assay
results
• 2.7TB of geophysics data in 62
datasets
• 60 prospect datasets
• Magnetics, gravity, seismic,
radiometrics, IP, EM and MT
• Petrology
• Geological maps and reports
Crowdsourcing Exploration at Mount Woods
8. The Explorer Challenge
• >5TB of data instantly accessible online
around the world
• >1000 participants
• >10,000 data downloads
• >60 countries
• Geologists, data scientists, startups, students,
consultants, universities, research
organisations
9. The Crowd and Exploration
Open and accessible data creates conditions
for:
• Multiple results to be developed in parallel
• Independent approaches – no group think,
no bias!
• Consensus = Confidence – statistically
relevant consensus due to independence
and diversity
• Speed, new knowledge, new workflows
and << cost are an additional bonus.
11. Explorer Challenge Outcomes
• Multiple approaches never before imagined
internally
• Applications of robust leading edge machine
learning, data science and geological
techniques
• New ways of extracting data, fusing and
analysing multiple data layers
• >400 targets – robust new targets generated,
confidence increased in known targets
12. Data Science and Exploration:
What did we learn?
• Multidisciplinary teams rule!
• DS enables state, national and global
datasets to be trained on and pulled into
local problems, a great way to reduce
bias
• Explainable/interpretable machine
learning is key for it to be added to the
geologists toolkit
• Geoscience data is not friendly for data
scientists
13. The Startup Ecosystem in Exploration Data Science
https://www.linkedin.com/pulse/startups-leveraging-machine-learning-
improve-holly-bridgwater/
14. Key takeaways
• CONSENSUS = CONFIDENCE,
geology is complex: search for
consensus not the best
• Open, accessible data creates an
environment where you can achieve
consensus.
• Data scientists + geologists = success!
• Internal – 1yr, 2-3 models max
• Crowd – 3 months, 40 models