Trend

---

Description

The Trend Analysis processor monitors numerical values and detects significant increases or decreases within a configurable time window. It:

• Detects percentage-based value changes
• Supports both increase and decrease detection
• Uses configurable time windows
• Preserves original event data
• Works with any numerical field

---

Required Input

The processor requires an input event stream with at least one numerical field to monitor for trends.

---

Configuration

Value to Observe

Select the numerical field that should be monitored for trends.

Operation Type

Choose the type of trend to detect:

• Increase: Detects when values increase by the specified percentage
• Decrease: Detects when values decrease by the specified percentage

Percentage Change

Specify the percentage threshold for trend detection:

• For increase: Values must increase by this percentage
• For decrease: Values must decrease by this percentage
• Range: 0-500%
• Step size: 1%

Time Window Length (Seconds)

Specify the duration in seconds to monitor for the trend.

Output

The processor outputs the original event when a significant trend is detected within the specified time window.

Example

Input Event

{
  "device_id": "device1",
  "measurement": "temperature",
  "value": 25.5,
  "timestamp": 1586380105115
}

Configuration

• Value to Observe: value
• Operation Type: Increase
• Percentage Change: 10
• Time Window Length: 300

Output Event

{
  "device_id": "device1",
  "measurement": "temperature",
  "value": 28.0,
  "timestamp": 1586380205115
}

Use Cases

1. Anomaly Detection

   • Detect sudden temperature changes
   • Monitor pressure variations
   • Track resource usage spikes
   • Identify unusual patterns

2. Performance Monitoring

   • Track system metrics
   • Monitor resource utilization
   • Analyze performance trends
   • Detect degradation

3. Quality Assurance

   • Monitor process parameters
   • Track product quality metrics
   • Detect process deviations
   • Ensure consistent output

Notes

• The processor detects percentage-based changes
• The time window is specified in seconds
• The output includes the original event data
• The processor works with any numerical field
• Results are emitted when the trend is detected