Machine Learning Models
regression
'linear'Linear regression model. Assumes a straight-line
relationship between independent variables and target.
See possible hyperparameters at
Scikit-learn LinearRegression'ridge'Linear model with L2 regularization. Penalizes large coefficients.
See possible hyperparameters at
Scikit-learn Ridge'lasso'Linear model with L1 regularization. Encourages sparsity in the coefficients.
See possible hyperparameters at
Scikit-learn Lasso'mlp'Multi-layer perceptron regression model. Basic neural network made of dense layers.
See possible hyperparameters at
Scikit-learn MLPRegressor'svr'Support Vector Machine regression model. Fits the regression line to target values within a specified margin.
See possible hyperparameters at
Scikit-learn SVR'decision_tree'Decision Tree regression model. Piecewise constant approximation of continuous values.
See possible hyperparameters at
Scikit-learn DecisionTreeRegressor'random_forest'Random Forest regression model.
Combines multiple decision trees using bootstrap aggregation (bagging).
Improves accuracy and reduces overfitting.
See possible hyperparameters at
Scikit-learn RandomForestRegressor 'gradient_boosting'Gradient Boosting regression model. Builds decision trees sequentially,
optimizing each tree to correct the residual errors of its predecessors.
See possible hyperparameters at
Scikit-learn GradientBoostingRegressor 'ada_boost'AdaBoost (adaptive boosting) regression model. Constructs a strong regressor by
sequentially training weak decision tree regressors, weighting samples based on their absolute errors.
See possible hyperparameters at
Scikit-learn AdaBoostRegressor 'knn'K-Nearest Neighbors regression model. Predicts a target value
by averaging the outputs of the K nearest samples.
See possible hyperparameters at
Scikit-learn KNeighborsRegressor'catboost'CatBoost regression model.
Handles categorical features without previous one-hot or label encoding.
Uses ordered boosting to reduce target leakage. Optimized for speed with GPU acceleration.
See possible hyperparameters at
CatBoost CatBoostRegressor'lgbm'LightGBM regression model.
Bins continuous features into histograms. Uses a leaf-wise growth strategy.
Optimized for speed with GPU acceleration. Reduced memory usage.
See possible hyperparameters at
LightGBM LGBMRegressor
classification
'logistic'Regularized Logistic Regression classification model.
Predicts binary classes probabilities. Parameters optimization is done via maximum likelihood estimation to minimize log-loss.
See possible hyperparameters at
Scikit-learn LogisticRegression 'mlp'Multi-layer perceptron classification model. Basic neural network made of dense layers.
See possible hyperparameters at
Scikit-learn MLPClassifier'svc'Support Vector Machine classification model.
Finds an optimal hyperplane that maximizes the margin between classes in a transformed higher dimensional feature space.
See possible hyperparameters at
Scikit-learn SVC 'decision_tree'Decision Tree classification model. Partitions the feature space into regions by recursively splitting the data along feature
thresholds to minimize impurity, assigning each leaf node the most frequent class label.
See possible hyperparameters at
Scikit-learn DecisionTreeClassifier'random_forest'Random Forest classification model.
Combines multiple decision trees using bootstrap aggregation (bagging).
Improves accuracy and reduces overfitting.
See possible hyperparameters at
Scikit-learn RandomForestClassifier 'gradient_boosting'Gradient Boosting classification model. Builds decision trees sequentially,
optimizing each tree to correct the residual errors of its predecessors.
See possible hyperparameters at
Scikit-learn GradientBoostingClassifier 'ada_boost'AdaBoost (adaptive boosting) classification model. Constructs a strong classifier by
sequentially training weak decision tree classifiers, weighting samples based on their absolute errors.
See possible hyperparameters at
Scikit-learn AdaBoostClassifier 'knn'K-Nearest Neighbors classification model.
Assigns the most common class among the K nearest samples.
See possible hyperparameters at
Scikit-learn KNeighborsClassifier 'catboost'CatBoost classification model.
Handles categorical features without previous one-hot or label encoding.
Uses ordered boosting to reduce target leakage. Optimized for speed with GPU acceleration.
See possible hyperparameters at
CatBoost CatBoostClassifier'lgbm'LightGBM classification model.
Bins continuous features into histograms. Uses a leaf-wise growth strategy.
Optimized for speed with GPU acceleration. Reduced memory usage.
See possible hyperparameters at
LightGBM LGBMClassifier