Saturday ML Analyzer

A machine-learning powered tool that analyzes test execution heatmaps to identify "Cold Spots"—areas of your application that are accessible but have low or zero test coverage.

Features

• K-Means Clustering: Uses ml-kmeans to group scattered interaction events into "Hotspots".
• Cold Spot Detection: Mathematically calculates which interactable elements are furthest from any known interaction cluster.
• Coverage Scoring: Provides a quantitative "Coverage Score" based on interactions vs. opportunities.

Installation

pnpm add @orieken/saturday-ml-analyzer

Usage

CLI

Run the analyzer against a directory containing heatmap JSON data (generated by @orieken/saturday-playwright-heatmap).

node dist/index.js <path-to-heatmap-data>

Example Output:

Test: User Login (passed)
  Coverage Score: 85.0%
  Hotspots: 3 (Username, Password, Submit)
  Cold Spots: 2
    - a "Forgot Password" 
    - a "Sign Up"

API

You can also use the analyzer programmatically:

import { HeatmapAnalyzer } from '@orieken/saturday-ml-analyzer';

const analyzer = new HeatmapAnalyzer();
const result = analyzer.analyze(testData);

console.log(result.coldSpots);

How It Works

Analysis Workflow

graph LR
    Input[Heatmap JSON] -->|Read| Loader
    Loader -->|Extract| Inter[Interactions (x,y)]
    Loader -->|Extract| Elements[Interactables]
    
    Inter -->|K-Means| Clustering[Identify Hotspots]
    
    Clustering & Elements -->|Distance Calc| Comparator
    
    Comparator -->|Output| ColdSpots[Cold Spots]
    Comparator -->|Output| Score[Coverage %]

Clustering Logic

The analyzer uses a proximity threshold to determine if an element was "tested".

graph TD
    Start --> Check{Distance to Nearest Hotspot < 50px?}
    Check -->|Yes| Tested[Mark as Covered]
    Check -->|No| Cold[Mark as Cold Spot]
    Cold --> Alert[Add to Risk Report]