Table of Contents

Enum OutlierDetectionMethod

Namespace
AiDotNet.Enums
Assembly
AiDotNet.dll

Defines different methods for detecting outliers in datasets.

public enum OutlierDetectionMethod

Fields

Combined = 2

Uses both Z-Score and IQR methods together to identify outliers.

For Beginners: The Combined method uses both Z-Score and IQR approaches together for more reliable outlier detection. A data point is flagged as an outlier only if both methods identify it as unusual.

This is like getting a second opinion - if only one doctor thinks something is wrong but another doesn't, you might want more information before taking action. But if both doctors agree there's an issue, you can be more confident.

The Combined approach reduces the chance of incorrectly flagging normal data points as outliers, making it more conservative but potentially more accurate, especially when you're not sure which method is best for your specific data.

IQR = 1

Detects outliers using the Interquartile Range method, which is resistant to extreme values.

For Beginners: The IQR (Interquartile Range) method is like focusing on the middle 50% of your data and then identifying values that fall too far outside this range.

Here's how it works:

  1. Arrange all your data from smallest to largest
  2. Find Q1 (the value 25% of the way through your sorted data)
  3. Find Q3 (the value 75% of the way through your sorted data)
  4. Calculate IQR = Q3 - Q1
  5. Any value below Q1 - 1.5×IQR or above Q3 + 1.5×IQR is considered an outlier

The advantage of IQR is that it's not affected by extreme values, making it more robust than Z-Score for skewed data or when you're not sure if your data follows a bell curve.

ZScore = 0

Detects outliers based on how many standard deviations a value is from the mean.

For Beginners: The Z-Score method measures how far away a data point is from the average (mean) in terms of standard deviations. Standard deviation is just a measure of how spread out the data is.

Imagine a classroom where the average height is 5'6" with a standard deviation of 2 inches. Using Z-Score:

  • Someone 5'10" tall has a Z-Score of +2 (2 standard deviations above average)
  • Someone 5'2" tall has a Z-Score of -2 (2 standard deviations below average)
  • Someone 6'4" tall has a Z-Score of +5 (5 standard deviations above average) - likely an outlier!

Typically, values with Z-Scores beyond +/-3 are considered potential outliers. This method works best when your data follows a bell curve (normal distribution).

Remarks

For Beginners: Outliers are data points that differ significantly from other observations in your dataset. Think of them as unusual values that stand out from the pattern - like if most people in a classroom are between 5'0" and 6'0" tall, someone who is 7'5" would be an outlier. Detecting outliers is important because they can skew your analysis or cause your AI model to learn incorrect patterns. These methods help you identify which data points might be outliers so you can decide whether to investigate them further or remove them from your analysis.

Edit this page