Correlation Heatmap

Overview

The Correlation Heatmap is a multivariate analysis tool that helps identify unexpected correlations among parameters, identifying root causes and affecting output fields. It can be used on Cycle and Part models for high-granularity and lower-granularity comparisons, making it a valuable first step in data analysis.

Before You Begin

Ensure you have:

• Access to the Application tab in Sight Machine
• Permission to view selected assets or part types
• At least 3–4 numeric parameters for correlation
• Minimum 30 observations in your selected date range

📝 Note: The more complete your dataset, the more reliable the correlations.

How to Create a Correlation Heatmap

1. Open the Correlation Heatmap

• Navigate to Application
• Select Correlation
• Select Correlation Heatmap

2. Select Your Data Model

Choose between:

• Cycles – analyze process parameters within the same asset or asset type
• Parts – analyze part-level characteristics across multiple machines

💡 Tip:

•  Use Cycles when comparing readings from similar or identical machines.
•  Use Parts when comparing finished product features across lines.

3. Choose Assets or Part Types

• Select Asset (Cycles) or Part Type (Parts)
• Choose one or multiple similar assets

📝 Note: Selecting multiple similar assets increases sample size, improving correlation reliability.

4. Set the Time Range

Choose from:

• Relative Ranges: Last 7/30/90 days
• Absolute Range: Specific start/end dates

⚠️ Warning: If the date range includes major process changes or downtime, correlations may be distorted.

5. Select Parameters

• Select 3–10 numeric parameters
• Categorical fields are automatically filtered out

💡 Tip: Start with 5–7 parameters for clearer visual results.

6. Configure Carry-Forwards

Choose how to handle forward-filled values:

• Keep All (default)
• First
• Last

7. Generate the Heatmap

• Select Update
• The heatmap calculates all correlations
• Results display in the matrix

8. Interpret the Heatmap

• Dark blue → strong positive correlation
• Dark red → strong negative correlation
• White/gray → weak or no correlation

💡 Tip:

 Start by reviewing the darkest cells to identify the strongest parameter relationships.

Watch the full workflow here 👉

Tips, Notes, and Warnings

💡 Tip:

 Use correlation findings to guide deeper analysis with Curve Fit or Time-Series Correlation.

📝 Note:

 Parameters with very low variability often show weak correlations simply because they don’t change much.

⚠️ Warning:

 Correlation ≠ causation. Always validate findings with process knowledge.

Feature Benefits

• Simultaneous Multi-Parameter Exploration: Enables the exploration of relationships among many parameters at the same time, accelerating the discovery of complex or unexpected correlations within the dataset.
• Visual Root Cause Discovery: Helps efficiently pinpoint potential root causes by visually identifying how changes in input fields may influence output fields, supporting early-stage diagnostic analysis.
• Pearson-R Coefficient Matrix: Computes the Pearson-R correlation coefficient for up to 10 selected parameters, quantifying the strength and direction of every pairwise relationship in a clear matrix format.
• Multivariate Model Flexibility: Supports multivariate analysis on Cycle and Part data models, allowing users to tailor the scope of the analysis based on their data structure needs.
• High-Granularity Cycle Analysis: Allows investigation of parameters from the same machine type with high granularity when using the Cycle model, ensuring detailed operational insights.
• Cross-Machine Comparison Capability: Utilizes the Part model to facilitate comparison of fields across different machines, which is essential for understanding interactions in end-to-end production processes.
• Integrated Summary Statistics: Provides an immediate table of summary statistics (Count, Standard Deviation, Min, Max) for each selected parameter, offering essential contextual data alongside the correlation matrix.
• Foundation for Detailed Follow-Up: Acts as an ideal exploratory starting point that logically leads to deeper investigation using other tools, such as the Curve Fit Analysis, for validating identified pairs of interest.

Summary

The Correlation Heatmap tool displays a heatmap of the correlation coefficient between all possible pairs of parameters that are associated with the selected machine type. It provides you with a visualization of the nature of correlation using color (positive in blue or negative in red), and the magnitude of the relationship using intensity or hue.

The value of correlation is represented by the Pearson-R correlation coefficient displayed in each cell.

This is an excellent way for you to discover which parameters are correlated to each other.