Analytics Module Part 4: Insights Tool for Correlation Equations and Setting Alerts


Overview

The Insights tool is the heart of the Analytics module, providing you with ITwin IoTs recommended best fit linear and quadratic equations, the ability to adjust the equations by defining Confidence Intervals, and using these equations to create correlation alerts for selected sensors. 

 

Navigating to the Insights Page

 

From the correlation analysis heatmap page, click on a tile to navigate to the Insights page. 

 

 

From the Variable Importance page, click on the bar to navigate to the Insights Page.

 

 

Insights Tool 

We will now take a look at the function of the Insights tool and how to use it. Once you navigate to the Insights page via one of the methods outlined above, you will see a graph at the top of the page comparing the selected sensors with a best-fit curve and confidence intervals overlaid on the graph. These items will be discussed below. 

 

 

Best Fit Curve- The default best-fit line, which is displayed in red on the graph, is the linear equation. There are two toggle boxes on the right side of the graph allowing you to choose between a linear or quadratic best-fit equation. 

 

Below the selection boxes, you will find two very important metrics generated by the Insights Tool. Here you will find the best-fit equation of the selected equation type as well as the fit score. 

The equation provided is the Analytics tools correlation between the two selected sensors based off of the selected linear or quadratic equation type. This equation can be used to create alerts correlation alerts, which will be discussed in a following section. 

The Fit score, denoted by the R^2 value and statistically representing the coefficient of determination, provides you with an understanding of the strength of the equation. This Fit score typically ranges from 0-1.0, with a value closer to 1 indicating that the equation fits the observed values well.

Confidence Interval- The Confidence Interval is shown as the blue bands on each side of the correlation line on the graph. The Confidence Interval indicates the probability that a future datapoint will fall into the bands. The Confidence interval can be set by a confidence value (percentage) or confidence band. Each option can be selected by the radio selection button icons.

  1. Confidence Value- The Confidence value