About

Solve Geosolutions is a data science consultancy based in Melbourne. Founded in 2016, we specialise in statistical analysis and machine learning-based solutions for exploration and mining.

As an industry we are seeing a rapid increase in the volume and diversity of geoscientific data. We need a new generation of tools to harness the complexity of this data and extract everything we can from our dataset. 

At Solve, our aim is to develop and implement data science-based solutions that improve how we mine and explore large and complex multivariate datasets.

 In addition to designing end-to-end solutions, we also want to empower our clients to think about their data in new and innovative ways through collaborative engagement and training.

What we do

Solve uses a wide variety of mathematical/statistical tools, workflows and algorithms to obtain geological insights. This process generally begins with exploratory data analysis to understand the structure in the data. Once the problem is clearly understood, there are a number of solutions that Solve can provide, which are centred around the classification and prediction of geoscientific datasets.

Explore large multivariate datasets

By using our toolkit of mathematics and statistics workflow, we can optimise our extraction of information from complex geoscience datasets. 

To understand and visualise high dimensional data, and analyse outliers and domains within multivariate datasets, we frequently use the following methods:

  • Self Organising Maps (SOM);
  • t-Distributed Stochastic Neighbour Embedding (t-SNE);
  • Principal Components Analysis (PCA).

Build quantitative relationships between datasets

Significant value can be added to new and existing datasets by using regression analysis to understand the relationships between different datasets. This can allow prediction between datasets.

  • Prediction of expensive variables from cheaper variables for potential cost savings;
  • Prediction of variables with long lead times from variables collected quickly to aid time-sensitive decision making;
  • Extrapolation of spatially constrained datasets to cover unsampled regions.

Use machine learning for 2D and 3D prospectivity analysis

Using sophisticated algorithms including Neural Networks and Ensemble learners, we are able to search for complex geological signatures across many layers of geoscience data.

  • These workflows can be applied in 2D or 3D across very large regions for regional exploration or deposit scales in data-rich environments;
  • Workflows allow for the simultaneous interrogation of many layers of data including remote sensing, geophysics, mineralogy, geochemistry, imagery and geomorphology;
  • These workflows generally output 2D or 3D probability maps as well as quantified information about what variables are important for the definition of the geological target.

Classify geological data

In exploration and mining, we constantly need to categorise rocks into groups and domains that share properties.

  • Groups can be constructed using Unsupervised learning methods (such as clustering) and based only on the data irrespective of a model;
  • Supervised classification methods can be trained to categorise data based on an experts model and the available data.

In the mine environment, these methods are commonly applied to classification models for:

  • Metallurgical classes;
  • Mineral domains;
  • Geological logging.

Our Toolbox

Gone are the days of geologists relying on a small set of expensive and task-specific commercial software packages to carry out their day-to-day jobs. Enter Solve: we prefer using open source software in our projects, and believe in the power and flexibility of collaborative and transparent solutions. Some of the software that we commonly use and provide training for includes:

R

R (and the associated R Studio) is a programming language primarily focused on statistical techniques, which can be used to automate complex processes and create high quality visualisations for any geological problem. Similar to QGIS, R has an extensive online community meaning that answers to problems are readily available.

Orange

Orange is an open source, widget-based software package for data visualisation, machine learning, data mining and data analysis. Orange takes away the requirement of learning code, which makes it a powerful tool for teaching machine learning concepts and workflows.

QGIS (previously Quantum GIS)

QGIS is an open source, cross platform GIS package that can read, edit and process a vast array of data types and allow for easy integration between different software packages. Its large development community mean that QGIS improves almost on a weekly basis and specific software problems can be addressed and integrated rapidly.

Amazon Web Services

With the decreasing price of cloud computing services, the need to have substantial personal computing power is rapidly diminishing. The AWS EC2 platform provides an easily scalable solution for even the largest computational problem.

Inkscape

Inkscape is an extremely flexible graphics package that allows for the creation of unique, publication-quality graphic solutions for any situation, and features a vast library of add-ons and extensions.

Who we are

Our consultants are from diverse backgrounds and specialise in both geoscience and mathematics, which means our data science workflows are grounded in a strong understanding of geology.

Brenton Crawford - Director

After studying structural geology and geophysics at Monash University, Brenton has consulted as a structural geophysicist for PGN Geoscience as well as working in a variety of geological and geophysical roles—predominantly in exploration. Most recently, Brenton has worked as a geophysicist and data scientist for MMG Exploration on a range of commodities. Brenton has a passion for the integration of regional geophysical and remote sensing datasets using machine learning and prospectivity analysis.

Liam Webb - Director

Liam is a geoscientist based in Melbourne and specialises in geophysics, 3D geological modelling (Leapfrog) and the integration of multivariate datasets in 3D. Originally from Tasmania, Liam studied at the University of Tasmania where he majored and completed his honours in geophysics, focusing on machine learning methods in blast hole ore categorisation. From greenfields exploration through to project mine geology, Liam has worked across a variety of commodities and terranes. Outside of geology, Liam is an avid sports fan and enjoys playing, watching and using machine learning to win fantasy sports.

Tom Carmichael - Director

The technical lead on most of Solve’s projects, Tom is an experienced mathematical geologist with a strong background in statistics, computer science and geology. Tom develops, integrates and implements fit-for-purpose machine learning solutions that enable clients to expand their knowledge base. Outside work hours, Tom can be found at sports grounds around Melbourne either playing (badly) or supporting (far too enthusiastically) any sport that involves people chasing a ball.