biolab/orange3

Features

• Visual Data Mining Tools - Provides an interactive environment for building data analysis workflows on a visual canvas without writing code.
• Widget-Based Workflow Builders - Provides a widget-based visual canvas for building data analysis workflows without writing code.
• Classification Models - Trains supervised classifiers on labeled data to predict categorical outcomes.
• Data Preprocessing - Applies transformations such as normalization, imputation, or feature selection to prepare data for modeling.
• Decision Trees - Builds tree-structured models that split data on attribute values to predict categorical or numeric outcomes.
• Gradient Boosting - Builds additive ensembles of shallow trees that sequentially correct errors for high-performance classification.
• K-Nearest Neighbor Classifiers - Predicts outcomes by finding the most common class among the closest training examples in the feature space.
• Outlier Detection - Identifies anomalous data points using statistical and density-based outlier detection methods.
• Linear Regression - Fits linear combinations of predictors to continuous response variables for prediction.
• General Regression Models - Fit a regression model to continuous target data using algorithms like linear regression or gradient boosting.
• Logistic Regression Models - Fits a regularized linear model that estimates class probabilities using a logistic function.
• Machine Learning Training - Trains classification and regression models using a variety of algorithms including random forests and neural networks.
• Visual Model Evaluators - Ships a visual workbench with interactive model evaluation tools, performance curves, and cross-validation results.
• Model Evaluation Metrics - Assesses predictive performance of models using cross-validation and scoring metrics.
• Data Preprocessing - Provides comprehensive data preprocessing including normalization, imputation, discretization, and encoding.
• Model Performance Evaluators - Provides built-in cross-validation with accuracy and AUC metrics for evaluating classifier generalization.
• Model Predictions - Generates continuous numeric predictions from trained regression models for new data instances.
• K-Fold Cross-Validation - Ships a built-in k-fold cross-validation pipeline for evaluating model performance on training data.
• Naive Bayes Classifiers - Builds a probabilistic classifier that treats each feature as independent when estimating class likelihoods.
• Random Forest Ensembles - Builds ensembles of decision trees on bootstrapped data samples and averages their predictions.
• Regression Models - Trains regression models on labeled data to predict continuous numeric outcomes.
• Regression Scoring Evaluation - Measures prediction accuracy using root mean squared error and R-squared scores for regression models.
• Statistical Analysis - Computes descriptive statistics, performance metrics, and conducts cross-validation to assess models.
• Supervised Classification - Trains supervised models on class-labeled data using algorithms like logistic regression or random forests.
• Dimensionality Reduction - Provides PCA, t-SNE, and FreeViz projection for reducing high-dimensional data into visualizable spaces.
• Visual Data Exploration - Builds and runs interactive data analysis workflows on a visual canvas without writing code.
• Built-in Analysis Methods - Applies a comprehensive set of data mining and statistical techniques through pre-built configurable components.
• Data Analysis Frameworks - Provides NumPy-backed data tables, distance matrices, and statistical computations for programmatic data mining.
• Single-Label Prediction - Applies trained classifiers to new data instances to output the most likely class for each instance.
• Data Domain Descriptors - Attaches a domain descriptor to every data table, defining variable names, types, and roles for consistent interpretation.
• Interactive Data Visualizations - Provides an interactive canvas for constructing and exploring data visualizations through connected components.
• Data Science Toolkits - Provides a comprehensive collection of data preprocessing, modeling, and evaluation components for exploratory analysis.
• Column Type Descriptors - Provides a domain descriptor system that defines variable names, types, and roles for every data table.
• Two-Dimensional Array Table Stores - Stores tabular data in two-dimensional arrays where rows represent instances and columns represent variables.
• Interactive Tabular Analysis - Offers an interactive visual canvas for exploring and analyzing datasets without writing code.
• Multi-Format File Loading - Loads tabular data from CSV, TSV, Excel, and pickle files with automatic column type interpretation.
• Dense and Sparse Tabular Array Stores - Stores data instances in 2D arrays for attributes, classes, weights, and meta attributes in dense or sparse formats.
• Lazy SQL Query Proxies - Applies row filters that translate to SQL WHERE clauses, returning a new proxy table without fetching all data.
• Dimensionality Reduction - Implements multiple dimensionality reduction algorithms including PCA and linear discriminant projection.
• Lazy Database Table Proxies - Provides a lazy SQL query proxy that exposes database tables as data tables without loading all data into memory.
• Distance Matrix Calculations - Computes pairwise distance matrices between data instances or columns using multiple metrics.
• Interactive Plot Exploration Widgets - Displays interactive plots and widgets that update in real-time as the data analysis workflow changes.
• Principal Component Analysis - Provides principal component analysis for transforming correlated variables into uncorrelated components.
• NumPy Array Integration - Stores tabular data as two-dimensional NumPy arrays for efficient vectorized operations.
• Domain Type Definitions - Defines domain descriptors that assign names, types, and roles to every column in a data table.
• Test Set Inference Evaluators - Applies trained models to independent datasets to estimate performance on unseen data.
• One-Class SVM - Identifies unusual data points by learning a boundary that encloses the normal data distribution.
• Class Probability Estimation - Provides per-class probability estimates from trained classifiers, not just hard labels.
• Classification Metric Scorers - Computes accuracy, precision, recall, F1, AUC, and log loss to measure classifier performance.
• Hierarchical Clustering - Groups data points into a tree of nested clusters based on pairwise distances for multi-level exploration.
• Regression Trees - Builds decision trees that predict numeric values by partitioning data into homogeneous regions.
• Distance-Based Clustering - Groups similar data instances together using distance-based or hierarchical algorithms.
• XGBoost Integrations - Fits a gradient boosting classifier using optimized tree construction for high performance.
• F1 Metric Scorers - Computes the harmonic mean of precision and recall to balance both metrics in a single score.
• Feature Scale Normalization - Ships a widget that scales continuous attributes to a standard range or distribution for consistent analysis.
• Categorical Feature Transformations - Provides a widget to transform categorical attributes into numeric indicator variables using encoding strategies.
• Top Feature Selectors - Provides a widget to filter datasets by retaining only the highest-scoring attributes based on statistical ranking.
• Elliptic Envelope Detectors - Flags unusual data points by fitting an elliptical boundary around the main data distribution.
• Isolation Forest Detectors - Identifies anomalies by measuring how easily each data point can be isolated with random splits.
• Local Outlier Factor Detectors - Identifies anomalies by comparing the density around each point to the density around its neighbors.
• Neural Network Classification - Ships a multi-layer perceptron classifier trained via backpropagation for non-linear decision boundaries.
• Training Evaluation - Tests models on the same data they were trained on to measure in-sample performance.
• Prediction Augmenters - Appends model predictions, probabilities, and fold indices as meta attributes to the original test data.
• Majority Class Baselines - Provides a majority class baseline classifier for comparison against more complex models.
• Stratified Folders - Preserves class proportions across folds during cross-validation to ensure balanced training and testing sets.
• Redundant Feature Removers - Provides a widget to drop constant attributes and unused categorical values from datasets.
• Regression Neural Networks - Trains multi-layer perceptrons to minimize squared error for continuous value prediction.
• Polynomial Regression - Constructs polynomial features of a specified degree and learns a prediction model on transformed data.
• Precision-Recall Visualizations - Ships precision, recall, F-measure, and support calculations for each class in classification results.
• Log Loss Scorers - Measures the uncertainty of probabilistic predictions by penalizing confident incorrect predictions.
• Fast - Builds random forest ensembles using optimized tree construction for faster training on large datasets.
• ROC AUC Scorers - Evaluates a classifier's ability to distinguish between classes by computing the area under the ROC curve.
• Leave-One-Out Cross-Validators - Provides leave-one-out cross-validation to evaluate model performance by training on all but one sample.
• Linear Discriminant Analysis - Implements linear discriminant analysis as a supervised dimensionality reduction technique.
• Support Vector Machines - Fits support vector classifiers that map data into higher-dimensional spaces using kernel functions.
• Conditional Data Filters - Provides a widget to select data instances based on a conjunction or disjunction of multiple filter conditions.
• Multi-Value Column Filters - Ships a widget that returns data instances matching a given value in a specified column.
• Continuous Value Filters - Ships a widget that selects instances based on comparison operators applied to a continuous column.
• FreeViz Projections - Ships a unique force-directed projection method for visualizing class separations in labeled data.
• Cross-Tabulations - Provides a widget to build cross-tabulation counts between one or more column variables and a discrete row variable.
• Data Quality Monitors - Ships a widget that checks for missing values, missing class values, and computes checksums over data arrays.
• Conditional Row Filters - Ships a widget that selects rows meeting criteria such as defined values, known classes, or matching variable values.
• Precision-Recall Curve Generators - Computes classification accuracy, F1, precision, recall, and other statistics across all probability thresholds.
• Missing Value Imputation - Provides a widget to detect and process missing entries using imputation or removal strategies.
• Optional Value Filtering - Ships a widget that selects data instances with no undefined values, optionally restricting the check to specific columns.
• Derived Variable Computations - Ships a widget that computes a variable's value from other variables using a stored transformation function.
• Recall Difficulty Evaluators - Calculates the ratio of true positive predictions to all actual positive instances for classifiers.
• Model Performance Curve Plotters - Generates curves that show how model performance changes across different thresholds or operating points.
• Numerical Binning - Ships a widget that converts continuous numeric attributes into categorical bins using partitioning strategies.
• Mean Absolute Error Calculators - Calculates the average absolute difference between predicted and actual values for regression models.
• Statistical Moment Calculation - Ships a widget that calculates min, max, mean, variance, and missing value counts for specified variables.
• Mean Squared Error Scorers - Computes the average squared difference between predicted and actual values for regression models.
• Prediction Calibrations - Adjusts predicted probabilities from classifiers to better reflect true likelihoods using regression-based calibration.
• Random Sampling - Provides a widget to filter rows or draw random subsets from a dataset for focused analysis.
• Exploit Prediction Scoring - Computes the ratio of true positive predictions to all positive predictions for classifier evaluation.
• Classification Accuracy Scorers - Computes classification accuracy as a standard metric for evaluating predictive model performance.
• Feature Relevance Scorers - Ships statistical scoring methods to rank attribute usefulness for feature selection in data preprocessing.
• Scientific Research Tools - Component-based software for interactive data mining and visualization.

Star-Verlauf