MINERAL PROCESSING WEB APPLICATIONS: STATISTICS & MACHINE LEARNING

By using these tools and documents on this website, you agree to the disclaimer and End User License Agreement section.

Confidence Interval:
Calculation of the confidence interval (CI) of a parameter's mean value X with a known standard deviation and sample size, at a given confidence level.
This applies to a parameter with repeat measurements. Example: calculation of the 90 percent confidence interval of the average gold recovery from 4 repeat flotation tests.
Watch a Youtube video example.

Probability calculations:
Calculation of the probability for a parameter's value X (e.g., recovery) to be below/above a given value or within a range.
This applies to a parameter with repeat measurements (e.g., recovery from repeated flotation tests, assay from repeated measurements, etc...) .
Watch a Youtube video example.

Send large or blocked file:
Some IT policies or email clients such as Microsoft Outlook may block large or executable file attachments.
In other cases, some companies do not allow employees to use external USB storages but they can download files from online to local drives.
Use this tool to upload your file and you will receive an email with the instructions for the recipients to download it.

Plant trial data cleaner:
Due to various reasons such as shutdowns, changes in feed, operating conditions or processing equipments, other concurrent trials and data recording; plant trial data require extensive cleaning prior to conducting statistical analyses.
Watch a Youtube video example.

Cumulative sums:
Due to daily plant variations, time series plots do not often show the time at which a change has occurred.
Instead, cumulative sums are used to identify changes in operating parameters with time.
Watch a Youtube video example.

Minimum number of tests:
In order to statistically detect a difference D between two means using Student's t-tests, the number of tests or sample size for each condition needs to be above a minimum value.
Watch a Youtube video example.

t-tests:
Student's two samples t-tests assuming equal variances quickly comparing parameters'values and/or KPI for two different trial conditions (e.g., Old Reagent vs. New Reagent).
This tool also calculates the statistical significance of the differences (for 1 and 2 tails), the confidence levels and confidence intervals for the provided alpha.
Watch a Youtube video example here.

Comparison of regression lines:
Comparison of two trendlines such as recovery versus feed grade or concentrate grade versus feed grade for two different conditions (e.g., Old Reagent vs. New Reagent).
This statistical method isolates the effect of the new condition (e.g., New Reagent) from that of another important parameter (e.g., feed grade).
Watch a Youtube video example here.

Multiple regression:
Multiple linear regression analysis to model one predicted or dependent variable as a function of many predictors or independent variables: Y = f(X₁,X₂,X₃,...,x_n)
It automatically calculates the confidence intervals and removes outliers based on the standard residuals limits specifed by the user (e.g., ±3) so they don't have to be manually removed one by one.
It also accounts for categorical predictors such as "day shift / night shift" or "trial ON / trial OFF", for examples, without the need to transform them to zeroes and ones.

Example:
Recovery = f(feed grade, throughput, particles size, reagent dosages, pH, trial ON)
Watch a Youtube video example here.

Classification regression/modelling:
This classification regression uses machine learning methodologies to predict a categorical outcome from numerical and categorical input parameters or predictors.

Example:
Multiple laboratory flotation tests were completed on changing feed grade nickel ore, using different types of collectors. The results are trained/modelled to predict what is the best collector to use for a future ore at various feed grades.