Pattern Recognition

Overview

The Pattern Recognition step detects behavioral patterns and anomalies in the data using both algorithmic and AI-based methods.

File: steps/pattern.step.ts

What It Does

1. Algorithmic detection for common patterns
2. Volatility analysis using coefficient of variation
3. Cross-series pattern detection (inverse correlations)
4. Agent enhancement for deeper pattern analysis

Detected Pattern Types

Pattern	Detection Logic	Threshold
Upward Trend	(last - first) / mean > threshold	15%
Downward Trend	(first - last) / mean > threshold	15%
High Volatility	stdDev / mean > threshold	0.25
Inverse Correlation	correlation(A, B) < threshold	-0.6

Algorithmic Detection

// Trend detection
const trendThreshold = 0.15;
const change = (last - first) / mean;
if (change > trendThreshold) {
  patterns.push({ name: 'upward_trend', series: id, detail: `${id} shows upward trend` });
}

// Volatility detection
const coeffOfVariation = stdDev / mean;
if (coeffOfVariation > 0.25) {
  patterns.push({ name: 'high_volatility', series: id, detail: `${id} shows high volatility` });
}

Input

{
  series: Series[],         // Time-series from preflight
  temporal?: TemporalOutput // Stats from temporal step
}

Output

{
  pattern: {
    matches: string[],      // Pattern descriptions
    confidence: number,     // 0-1 confidence score
    source: 'algorithmic' | 'agent' | 'mock',
    agentUsed: boolean,
    detectedPatterns: Array<{
      name: string,         // Pattern type identifier
      series: string,       // Which series exhibited pattern
      detail: string        // Human-readable description
    }>
  }
}

Agent Enhancement

The pattern agent provides deeper analysis:

const agent = await getPatternAgent(mastra);
const response = await callAgent(agent, {
  series: inputData.series,
  temporal: inputData.temporal,
  objective: inputData.objective
});

Agent can detect:

• Complex multi-series patterns
• Contextual anomalies
• Behavioral cycles (weekly, monthly)
• Onset/offset patterns

Example Output

{
  "pattern": {
    "matches": [
      "Stress shows upward trend (+45%)",
      "Energy exhibits high volatility",
      "Inverse pattern: stress↑ correlates with mood↓"
    ],
    "confidence": 0.85,
    "source": "agent",
    "agentUsed": true,
    "detectedPatterns": [
      {
        "name": "upward_trend",
        "series": "stress",
        "detail": "Stress increased 45% over the analysis period"
      },
      {
        "name": "high_volatility", 
        "series": "energy",
        "detail": "Energy levels fluctuate significantly day-to-day"
      },
      {
        "name": "inverse_correlation",
        "series": "stress,mood",
        "detail": "When stress increases, mood tends to decrease"
      }
    ]
  }
}

Skip Conditions

This step is skipped if:

• stepPlan exists and doesn't include 'pattern'
• No series data is available