Moving Average

---

Description

The Moving Average processor smooths numerical data streams by calculating either the mean or median of the last n values. This processor is essential for:

• Reducing noise in sensor data
• Smoothing out fluctuations
• Identifying trends
• Improving data quality

---

Required Input

A numerical field is required in the input stream.

---

Configuration

Numerical Field

• Select the numerical field to be smoothed
• The field must contain numeric values

N Value

• Specifies the window size (number of previous values to consider)
• Larger values create smoother output but increase latency
• Smaller values are more responsive but may show more noise

Method

Choose between two smoothing methods:

• Mean: Calculates the arithmetic average of the last n values
• Median: Uses the middle value of the last n values (better for handling outliers)

Output

The processor appends a new field named "filterResult" containing the smoothed value.

Example

Input Events

{
  "temperature": 25.5,
  "timestamp": 1586380104915
}
{
  "temperature": 26.0,
  "timestamp": 1586380105015
}
{
  "temperature": 25.8,
  "timestamp": 1586380105115
}

Configuration

• Numerical Field: temperature
• N Value: 3
• Method: Mean

Output Events

{
  "temperature": 25.5,
  "timestamp": 1586380104915,
  "filterResult": 25.5
}
{
  "temperature": 26.0,
  "timestamp": 1586380105015,
  "filterResult": 25.75
}
{
  "temperature": 25.8,
  "timestamp": 1586380105115,
  "filterResult": 25.77
}

Use Cases

1. Sensor Data Processing

   • Smooth temperature readings
   • Filter noise from measurements
   • Stabilize sensor outputs
   • Improve data quality

2. Trend Analysis

   • Identify long-term patterns
   • Reduce short-term fluctuations
   • Highlight significant changes
   • Monitor system behavior

Notes

• The processor maintains a sliding window of the last n values
• Mean method is more sensitive to outliers
• Median method is more robust against outliers
• Window size affects smoothing intensity
• Original values are preserved in the output
• First n-1 events will have partial windows