Enum NormalizationMethod
Defines different methods for normalizing data before processing in machine learning algorithms.
public enum NormalizationMethodFields
Binning = 5Groups continuous data into discrete bins or categories.
For Beginners: Binning is like grouping people by age ranges (0-10, 11-20, 21-30, etc.) instead of using their exact ages. It converts continuous numbers into categorical groups. This can help reduce noise in the data and reveal patterns that might be hidden when looking at exact values.
Decimal = 3Scales values by dividing by powers of 10 to bring all values between 0 and 1.
For Beginners: Decimal scaling moves the decimal point in your numbers until all values are between -1 and 1. For example, if your largest number is 1,000, you'd divide everything by 1,000, so that number becomes 1.0. This is useful when your data has different numbers of digits but is otherwise similar. Formula: x / 10^j where j is the smallest integer such that max(|x|) < 1
GlobalContrast = 6Normalizes data by adjusting contrast across the entire dataset.
For Beginners: Global Contrast Normalization is like adjusting the contrast on a photo to make features more visible. It's especially useful for image data, where it helps highlight important patterns while reducing the impact of variations in lighting or brightness.
Log = 4Applies a logarithmic transformation to compress the range of values.
For Beginners: Log transformation is useful when your data has a few very large values that would otherwise dominate. It compresses high values while spreading out smaller values. For example, the difference between 1 and 10 becomes the same as the difference between 10 and 100. This is often used for data that grows exponentially, like population counts or prices. Formula: log(x) (Note: typically requires handling zero or negative values specially)
LogMeanVariance = 7Applies logarithmic transformation followed by mean-variance normalization.
For Beginners: This is a two-step process: first, a log transformation compresses very large values (like turning 1,000,000 into 6), then the result is standardized around a mean of 0. This is useful for data with both exponential growth patterns and the need for standardization, such as financial or scientific measurements that vary by orders of magnitude.
LpNorm = 8Normalizes data using the Lp norm (typically L1 or L2 norm).
For Beginners: Lp Norm scaling is like making sure the total "energy" of your data equals 1. The most common version (L2 norm) is similar to making sure the length of a vector equals 1. This is useful in text analysis and when working with high-dimensional data, as it helps compare documents or data points of different lengths fairly. Formula for L2 norm: x / sqrt(sum(x^2))
MaxAbsScaler = 11Scales features to the range [-1, 1] by dividing by the maximum absolute value.
For Beginners: MaxAbsScaler is like MinMax scaling, but instead of using both the minimum and maximum values, it only uses the maximum absolute value (the largest value ignoring signs). This method preserves zeros and the sign of values, which is important for sparse data where many values are zero. For example, if your largest value is 100 and smallest is -50, everything gets divided by 100, so results fall between -1 and 1. Formula: x / max(|x|)
MeanVariance = 9Standardizes data to have a specified mean and variance (typically mean=0, variance=1).
For Beginners: Mean-Variance normalization is similar to Z-Score but gives you more control. It adjusts your data to have a specific average (usually 0) and spread (usually 1). This makes different features comparable and helps many machine learning algorithms perform better. Formula: (x - mean) / standard_deviation
MinMax = 1Scales all values to a range between 0 and 1.
For Beginners: MinMax scaling takes your data and squeezes it to fit between 0 and 1. The smallest value becomes 0, the largest becomes 1, and everything else falls in between. It's like converting heights of people in a classroom to percentages of the tallest person. Formula: (x - min) / (max - min)
None = 0No normalization is applied to the data.
For Beginners: This option means "don't change my data at all." Use this when your data is already on a similar scale or when you specifically want to preserve the original values.
QuantileTransformer = 12Transforms features to follow a uniform or normal distribution using quantiles.
For Beginners: QuantileTransformer is a powerful technique that changes your data's distribution to be either uniform (flat, where all ranges have equal numbers of values) or normal (bell-shaped). It works by ranking values and mapping them to a target distribution. This is extremely effective at handling outliers because it spreads them out across the distribution. Think of it as redistributing your data so that it matches a desired pattern, regardless of the original distribution.
RobustScaling = 10Scales features using statistics that are robust to outliers.
For Beginners: Robust Scaling is designed to handle data with outliers (extreme values that don't follow the pattern). Instead of using the minimum and maximum values (which can be skewed by outliers), it uses the median and quartiles. It's like saying "ignore the extremely tall and short people when standardizing heights." This is useful when your data contains unusual values that shouldn't influence the overall scaling. Formula: (x - median) / (Q3 - Q1) where Q1 is the 25th percentile and Q3 is the 75th percentile
ZScore = 2Standardizes values to have a mean of 0 and a standard deviation of 1.
For Beginners: Z-Score (also called standardization) centers your data around 0, with most values falling between -3 and +3. It tells you how many "standard deviations" away from average each value is. For example, a Z-Score of 2 means "this value is 2 standard deviations above average." This method works well when your data follows a bell curve pattern. Formula: (x - mean) / standard_deviation
Remarks
For Beginners: Normalization is like converting different measurements to a common scale. Imagine you have data about people's heights (in feet) and weights (in pounds) - these numbers are on very different scales. Normalization transforms all your data to similar ranges (like 0-1) so that one feature doesn't overwhelm others just because it uses bigger numbers. This helps machine learning algorithms work better and faster.