We provide expertise in deep learning architectures and machine learning algorithms

Data Science in Mining


Knowledge Discovery involves unveiling meaningful, actionable, and consistent patterns and connections within vast datasets. By using tailored methods suited to the dataset's nature, we uncover significant hidden attributes. APMT offers expertise in knowledge discovery through:

Causal inference modeling on continuous and categorical variables
Unsupervised learning for classification and clustering
Multivariate data analysis and dimensionality reduction


Statistical learning involves deciphering the statistical correlations among several variables to establish precise predictive models. This understanding can manifest through unsupervised, (semi)supervised, supervised, and reinforced learning techniques. At APMT, our aim is to offer models that are both intuitively comprehensible and exhibit top-tier accuracy, anchored in:

Regressions: linear, logistic, step-wise, and lasso regression
Machine learning: k-nearest neighbors, random forest, support vector machine, hidden Markov models, and Markov chain Monte Carlo
Deep learning: feedforward, convolutional, recurrent, and generative-adversarial neural networks


Machine Learning (ML), with its specialized subset Deep Learning (DL), offers an expansive suite of techniques for tasks like classification, regression, and forecasting. Given their sensitivity to data context and workflow representation, a performance evaluation is crucial before their real-world deployment to ensure informed decision-making. At APMT, we derive the best predictive models through:

Training a wide range of ML/DL in parallel over different combinations of hyperparameters
Measuring their performance and picking up the best model (tunned hyperparameters) per method
Comparing all best models (one per method) and selecting the optimal ML/DL architecture for industrial applications


In the mining sector, there are instances where a singular predictive model doesn't fully address our objectives. Ensemble methods frequently outperform by integrating multiple ML/DL frameworks into a coherent workflow, culminating in a singular dependable predictor. At APMT, we merge mining operational insights with predictive frameworks to deliver ideal ensemble workflows and steadfast predictions.