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.

TryMLEasy exposes six classification models from scikit-learn, selectable from the Traditional ML Model page once you have chosen “Classification” as the supervised model type. Every model is instantiated with scikit-learn’s default hyperparameters and is trained as part of a Pipeline that can optionally prepend a scaler and/or a decomposition step configured on the Feature Selection page. After training, TryMLEasy reports Accuracy, F1 Score, Precision, and Recall on the held-out test split, and renders a Predicted vs. True scatter plot.
All models use scikit-learn default hyperparameters — no manual tuning (e.g., C, max_depth, n_neighbors) is available in the current version of TryMLEasy.

Logistic Regression

  • scikit-learn class: sklearn.linear_model.LogisticRegression
  • Description: A linear model that estimates the probability of class membership using the logistic (sigmoid) function. Supports binary and multiclass classification.
  • Strengths: Fast to train, highly interpretable coefficient weights, rarely overfits on moderate-sized datasets, produces calibrated probabilities.
  • Scaling recommended: Yes — Logistic Regression converges faster and more reliably when features are on the same scale.
  • Typical use case: Fast, explainable baseline for any classification problem, especially when a linear decision boundary is plausible.

Decision Tree Classifier

  • scikit-learn class: sklearn.tree.DecisionTreeClassifier
  • Description: Recursively partitions the feature space using axis-aligned splits to build a tree of if/else rules that map inputs to class labels.
  • Strengths: Fully interpretable tree structure; no feature scaling required; handles mixed numeric and categorical-encoded features.
  • Scaling recommended: No — tree-based splits are threshold comparisons and are invariant to monotonic transformations of feature values.
  • Typical use case: Exploratory analysis; situations where human-readable decision rules are required. Be aware that without depth constraints the default tree may overfit.

Gaussian Naive Bayes

  • scikit-learn class: sklearn.naive_bayes.GaussianNB
  • Description: A probabilistic classifier that applies Bayes’ theorem and assumes that features within each class follow a Gaussian (normal) distribution, and that features are conditionally independent given the class label.
  • Strengths: Extremely fast to train and predict; works well with small datasets; robust to irrelevant features in practice despite the independence assumption.
  • Scaling recommended: No — the model estimates per-feature Gaussian parameters independently; scaling does not affect the result.
  • Typical use case: Small datasets; situations where a fast probabilistic baseline is needed; text or sensor classification where Gaussian feature distribution is reasonable.

Support Vector Classification

  • scikit-learn class: sklearn.svm.SVC
  • Description: Finds the maximum-margin hyperplane that separates classes in the feature space. Uses the RBF kernel by default, which can model non-linear decision boundaries.
  • Strengths: Effective in high-dimensional spaces; memory-efficient (uses only a subset of training points — the support vectors); strong theoretical generalisation guarantees.
  • Scaling recommended: Yes, strongly — the RBF kernel computes distances in feature space, so features with large numeric ranges will dominate the kernel unless scaled. Pair with Standard Scaler or Robust Scaler.
  • Typical use case: Medium-sized datasets with complex, non-linear class boundaries; problems where high accuracy on a small dataset matters more than training speed.

K-Nearest Neighbors

  • scikit-learn class: sklearn.neighbors.KNeighborsClassifier
  • Description: A non-parametric, instance-based classifier. For each test sample, it identifies the k nearest training samples (default k = 5) by Euclidean distance and assigns the majority class label.
  • Strengths: No explicit training phase; naturally handles multi-class problems; simple and intuitive.
  • Scaling recommended: Yes, strongly — KNN relies entirely on distance calculations. Features with large value ranges will dominate the distance metric if unscaled.
  • Typical use case: Small-to-medium datasets where a simple, non-parametric baseline is useful. Prediction is slow on large datasets because distances to all training points must be computed.

Stochastic Gradient Descent Classifier

  • scikit-learn class: sklearn.linear_model.SGDClassifier
  • Description: A linear classifier (by default fitting a linear SVM objective) trained using Stochastic Gradient Descent. Processes one sample at a time, making it memory-efficient and fast on large datasets.
  • Strengths: Scales to very large datasets and sparse feature spaces; supports multiple loss functions; fast incremental updates.
  • Scaling recommended: Yes — gradient-based optimisation converges more reliably when features share a similar scale.
  • Typical use case: Large datasets where other linear classifiers (Logistic Regression, SVC) are too slow; online or incremental learning scenarios.

Summary Table

Modelscikit-learn ClassScaling RecommendedStrengths
Logistic RegressionLogisticRegressionYesFast, interpretable, calibrated probabilities
Decision Tree ClassifierDecisionTreeClassifierNoHuman-readable rules, scale-invariant
Gaussian Naive BayesGaussianNBNoVery fast, works on small datasets
Support Vector ClassificationSVCYes (strongly)High-dimensional data, non-linear boundaries
K-Nearest NeighborsKNeighborsClassifierYes (strongly)Simple, non-parametric, no training phase
SGD ClassifierSGDClassifierYesLarge datasets, sparse features
Start with Logistic Regression. It is fast, interpretable, and provides a reliable performance baseline. If the linear model underperforms, step up to SVC (with Standard Scaler) for non-linear boundaries, or try KNN for a non-parametric comparison. Use the correlation heatmap on the Feature Selection page to understand feature relationships before committing to a model.

Build docs developers (and LLMs) love