Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Gurneet1928/TryMLEasy/llms.txt

Use this file to discover all available pages before exploring further.

Before passing feature values to a traditional machine learning model, TryMLEasy can optionally rescale them through a preprocessing step. This rescaling is configured on the Feature Selection page via the “Select Preprocessing Step” dropdown, and is inserted as the first stage of the scikit-learn Pipeline that runs before any decomposition or model fitting.
Preprocessing is applied only for Traditional ML models. It is not available for the neural network path — the neural network pages consume the raw train/test splits produced by the Feature Selection page directly.

​
None

Selecting None skips the preprocessing stage entirely. The raw feature values from your CSV are passed directly to the next pipeline stage (decomposition or the model itself).
  • scikit-learn class: (not applied)
  • When to use: When your data is already scaled, when you are using tree-based models that are scale-invariant (e.g., Decision Tree), or when you want to establish an unscaled baseline.

​
Standard Scaler

Standard Scaler standardizes each feature by subtracting its mean (μ) and dividing by its standard deviation (σ), producing a feature with zero mean and unit variance: 
x_scaled = (x − μ) / σ
  • scikit-learn class: sklearn.preprocessing.StandardScaler
  • When to use: When your features follow a roughly Gaussian (normal) distribution and you want them on a comparable scale. This is the most common choice for distance- or gradient-based algorithms such as SVC, KNN, and Logistic Regression.

​
MinMax Scaler

MinMax Scaler linearly transforms each feature so that all values fall within the [0, 1] range: 
x_scaled = (x − x_min) / (x_max − x_min)
  • scikit-learn class: sklearn.preprocessing.MinMaxScaler
  • When to use: When you need a strictly bounded output range, for example when an algorithm expects inputs in [0, 1]. Be aware that MinMax Scaler is sensitive to outliers — a single extreme value will compress all other values into a narrow band.

​
Robust Scaler

Robust Scaler scales features using the median and the interquartile range (IQR), making it resistant to the influence of outliers: 
x_scaled = (x − median) / IQR
  • scikit-learn class: sklearn.preprocessing.RobustScaler
  • When to use: When your dataset contains extreme values or significant outliers that would distort mean-based or range-based scaling. A good first choice for real-world messy data before trying Standard Scaler.

​
Normalization

Normalization rescales each sample (row) independently to have unit norm (L2 norm = 1 by default), rather than rescaling each feature column: 
x_normalized = x / ||x||
  • scikit-learn class: sklearn.preprocessing.Normalizer
  • When to use: When the magnitude of individual samples matters less than the direction (i.e., the relative proportions of feature values). Often used with text data represented as TF-IDF vectors or other sparse, high-dimensional data.

​
Comparison Table

Techniquescikit-learn ClassOutlier RobustUse Case
None(not applied)N/ATree models, pre-scaled data, baseline
Standard ScalerStandardScalerNoNormally distributed features; SVC, KNN, Logistic Regression
MinMax ScalerMinMaxScalerNoAlgorithms requiring bounded [0, 1] input; clean data
Robust ScalerRobustScalerYesData with outliers or skewed distributions
NormalizationNormalizerPartialText/sparse data; row-wise magnitude normalisation
Standard Scaler is the recommended default starting point for most tabular datasets. It works well with the majority of models available in TryMLEasy and pairs naturally with the PCA decomposition step. Switch to Robust Scaler if you notice extreme values in the correlation heatmap shown at the top of the Feature Selection page.

Build docs developers (and LLMs) love

Get started for freeTalk to us