Sample Truncation Strategies for Outlier Removal in Geochemical Data: The MCD Robust Distance Approach Versus t-SNE Ensemble Clustering

Leung, Raymond; Balamurali, Mehala; Melkumyan, Arman

doi:10.1007/s11004-019-09839-z

Cited by 15 publications

(7 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The results, which contain the concentrations of 33 elements, were analyzed using Excel 2013 and SPSS 26 software. After separating the data of different ore zones (the oxide and sulfide ores of Black Hill and East Ridge ore zones), the outliers were omitted and replaced by adjacent data, the data then being normalized (Bá rdossy and Fodor, 2004;Kwak and Kim, 2017;Leung et al, 2021). The correlation coefficients and bivariate regression between the elements were calculated.…”

Section: Methodsmentioning

confidence: 99%

3D Modeling and Determination of Factors Responsible for Zinc‐Lead Mineralization in the Mehdiabad Deposit, Central Iran, based on Statistical Analysis of Geochemical Data

Bonyadi

2022

Acta Geologica Sinica (Eng)

View full text Add to dashboard Cite

Although the Mehdiabad zinc‐lead deposit is one of the most well‐known deposits in the central Iran structural zone, the genesis of the deposit remains controversial. The host rock of the ore is a dolomitic limestone of the Lower Cretaceous Taft Formation. In the two main orebodies of the deposit, which includes the Black Hill and East Ridge ore zones, the oxide and sulfide ores are observed at the surface and at depth, respectively. The elements Zn, Fe, Mn and Mg are more abundant in the East Ridge ore zone (in both sulfide and oxide ores), with Ba, Pb, Ag and Cu being more abundant in the Black Hill oxide ore. Based on the distribution of elements and their correlation with each other in these ore zones, the elements are divided into three general groups, that of terrigenous elements, chemically‐deposited elements and ore‐forming (hydrothermally deposited) elements, a division that is supported by the results of factor analyses. The spatial distribution of elements is jointly affected by contact with host rocks, the boundary of oxide‐sulfide ores and fault zones. The main factors governing the distribution of elements are the mechanical transfer of detrital sediments, chemical sedimentation, transfer by hydrothermal fluids, oxidation and surface dissolution, all of which affected the spatial distribution of elements. The ore‐forming elements are mostly affected by hydrothermal fluids and oxidation. This study not only provides additional information about the genesis of the Mehdiabad deposit, but also could assist in the exploitation of ore and further exploration purposes. The results of this study can aid in the exploration and exploitation of the Mehdiabad deposit and similar deposits in the region.

show abstract

Section: Methodsmentioning

confidence: 99%

3D Modeling and Determination of Factors Responsible for Zinc‐Lead Mineralization in the Mehdiabad Deposit, Central Iran, based on Statistical Analysis of Geochemical Data

Bonyadi

2022

Acta Geologica Sinica (Eng)

View full text Add to dashboard Cite

show abstract

“…As the merit of temperature modelling results is largely determined by the data set, and the data collection process itself suffers from problems such as inaccurate data and many confounding factors, there is often some anomalous data in the data collected in real time, which leads to increased modelling difficulties and reduced model building accuracy [11].In particular, the actual temperature acquisition images show large transient temperature variations, where individual temperature values are clearly anomalous and therefore need to be rejected prior to modelling.The pauta criterion (3σ criterion) is used to remove outliers for data cleaning in this paper.σ represents the standard deviation and μ represents the mean. It is generally accepted that the data take values almost exclusively within the (μ-3σ,μ+3σ) interval, and that those outside this range can be considered outliers, with a probability of being outside less than 0.3%.If the sample is X={x1,x2,…xn}, then the standard deviation is calculated according to Eq.…”

Section: Exception Data Pre-processingmentioning

confidence: 99%

Prediction and Modelling of Plasma Jet Zone Temperature for DC Arc Graphite Purification

Wang,

Zhang,

et al. 2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

The random forest algorithm was used to analyze the degree of influence of four variables, namely, voltage, current, main gas flow rate, and powder feed, on the temperature of the plasma jet zone of the DC arc graphite purification. The BP algorithm was then used to build a network prediction model. After the training of the network, the validity of the model was verified by comparing the measured temperature values with the predicted results. From the results, it can be seen that the relative error of the prediction results is between -0.3 %and +0.3%, which proves that the built BP neural network temperature model has a satisfactory prediction effect. In addition, this paper also derives specific mathematical expressions based on the threshold and weight values of this temperature prediction model, which can be used to fit the temperature variation curve in the plasma jet region.

show abstract

“…For learning, 48,000+ assay samples were collected from exploration holes from a Pilbara iron-ore deposit located in Western Australia. Each sample consists of (c, g) where g ∈ Z denotes the geozone label, and c ∈ R 10 measures the chemical composition in terms of Fe, SiO 2 , Al 2 O 3 , P, LOI (loss on ignition), TiO 2 , MgO, Mn, CaO and S. The geochemical characteristics of this data is described in [30]. Samples were randomly split 60:40 to produce training and validation sets.…”

Section: Machine Learning Techniques For Likelihood Estimationmentioning

confidence: 99%

“…Following the approach in [30], the chemical data was ilr-transformed. Geozone classification performance associated with p(g | ilr(c)) is reported in Table 6.…”

Section: On the Use Of Isometric Log-ratio Transformmentioning

confidence: 99%

See 1 more Smart Citation

Surface Warping Incorporating Machine Learning Assisted Domain Likelihood Estimation: A New Paradigm in Mine Geology Modelling and Automation

Leung,

Balamurali,

Lowe

2021

Preprint

Self Cite

View full text Add to dashboard Cite

This work demonstrates an application of machine learning to domain likelihood estimation within an orebody modelling system. In this application area, demarcation of boundaries is a key determinant of success in 3D geologic modelling. The placement of these boundaries is critical and when they are specified by mesh surfaces, a wide variety of topological features and geological structures can be represented, including stratigraphic boundaries that delineate different domains. In surface mining, assay measurements taken from production drillings often provide a rich source of information that allows initially inaccurate surfaces (for example, mineralisation boundaries) created using sparse exploration data to be revised and subsequently improved. Recently, a Bayesian warping technique has been proposed to reshape modelled surfaces based on geochemical and spatial constraints imposed by newly acquired blasthole data. This paper focuses on incorporating machine learning in this warping framework to make the likelihood computation generalisable. The technique works by adjusting the position of vertices on the surface to maximise the integrity of modelled geological boundaries with respect to sparse geochemical observations. Its foundation is laid by a Bayesian derivation in which the geological domain likelihood given the chemistry, p(g | c), plays a similar role to p(y(c) | g) under certain conditions. This observation allows a manually calibrated process centred around the latter to be automated since machine learning techniques may be used to estimate the former in a data-driven way. Machine learning (ML) performance is evaluated for gradient boosting, neural network, random forest and other classifiers in a binary and multi-class context using precision and recall rates. Once the machine learning likelihood estimators are integrated in the surface warping framework, surface shaping performance is evaluated using unseen data by examining the categorical distribution of test samples located above and below the warped surface. Extended analysis provides further insight on the utility of isometric log-ratio transform and incorporating spatial information as part of the feature vectors with respect to domain classification and surface warping. Large-scale validation experiments are performed to assess the overall efficacy of ML assisted surface warping as a fully integrated component within an ore grade estimation system where the posterior mean is obtained via Gaussian Process (GP) inference with a Matérn 3/2 kernel.

show abstract

Sample Truncation Strategies for Outlier Removal in Geochemical Data: The MCD Robust Distance Approach Versus t-SNE Ensemble Clustering

Cited by 15 publications

References 17 publications

3D Modeling and Determination of Factors Responsible for Zinc‐Lead Mineralization in the Mehdiabad Deposit, Central Iran, based on Statistical Analysis of Geochemical Data

3D Modeling and Determination of Factors Responsible for Zinc‐Lead Mineralization in the Mehdiabad Deposit, Central Iran, based on Statistical Analysis of Geochemical Data

Prediction and Modelling of Plasma Jet Zone Temperature for DC Arc Graphite Purification

Surface Warping Incorporating Machine Learning Assisted Domain Likelihood Estimation: A New Paradigm in Mine Geology Modelling and Automation

Contact Info

Product

Resources

About