How to Measure Machine Learning Performance

Learn about Machine Learning Performance, including how to measure it, and leverage it in dashboards and visualizations with Metabase.

What is Machine Learning Performance?

Machine learning performance metrics are a set of metrics that tell you about various classification performance indicators. These metrics aim to identify how well machine learning is working on coming up with predictions, gathering information, and how specific your machine learning can be. The machine learning we use is almost never 100% accurate; achieving that level of accuracy would take up a lot of resources. The true purpose of these metrics is just to make sure your machine learning is at a usable standard.

Get Started
Graphs of Machine Learning Performance

How to calculate Machine Learning Performance

There are several different calculations you can use to assess the reliability of your machine learning. You’ll need to understand your machine learning’s true and false positives and negatives based on actual and predicted classifications. Some examples of calculations you can do to achieve this are: Recall (or sensitivity) - How many times a positive prediction was made by a model. Accuracy - How the model performs across all classes Precision - The quality of a positive prediction made by a model. Logarithmic loss - How close a prediction probability is to its true value.

Other KPIs to measure related to Machine Learning Performance

  • Issue Resolution Time
  • Production Incidents
  • Queue Time
  • Regressions Open/Closed
  • Release Burndown
  • Security Vulnerabilities
  • Service Line Health Impact
  • Soak Testing
  • Static Code Analysis

Why build a dashboard for Machine Learning Performance?

Sales per source graph

Everything in one place

Get everyone on the same page by collecting your most important metrics into a single view.

Sales per source graph

Share your perspective

Take your data wherever it needs to go by embedding it in your internal wikis, websites, and content.

eCommerce orders filter

Unlock exploration

Empower your team to measure their own progress and explore new paths to achieve their goals.

How to use Metabase to measure Machine Learning Performance

Get started

Step 1.
Skip the custom quote

That's right, no sales calls necessary—just sign up, and get running in under 5 minutes.

types of databases

Step 2.
Plugin your database

We connect to the most popular production databases and data warehouses.

dashboard example

Step 3.
Build your KPI dashboards

Invite your team and start building dashboards—no SQL required.

Get started with Metabase

  • Free, no-commitment trial
  • Easy for everyone—no SQL required
  • Up and running in 5 minutes