AutoML capabilities

FEDOT is capable of setting its ‘automation rate’ by omitting some of its parameters. For example, if you just create the FEDOT instance and call the fit method with the appropriate dataset on it, you will have a full automation of the learning process, see automated composing

At the same time, if you pass some of the parameters, you will have a partial automation, see manual composing.

Other than that, you can utilize more AutoML means.

Data nature

FEDOT uses specific data processing according to the source of the input data (whether it is pandas DataFrame, or numpy array, or just a path to the dataset, etc).

Note

Be careful with datetime features, as they are to be casted into a float type with milliseconds unit.

Apart from that FEDOT is capable of working with multi-modal data. It means that you can pass it different types of datasets (tables, texts, images, etc) and it will get the information from within to work with it.

See also

Detailed multi-modal data description and usage

Dimensional operations

FEDOT supports bunch of dimensionality preprocessing operations that can be be added to the pipeline as a node.

Feature transformation operations definitions

API name	Definition	Problem
rfe_lin_reg	Linear Regression Recursive Feature Elimination	Feature extraction
rfe_non_lin_reg	Decision Tree Recursive Feature Elimination	Feature extraction
rfe_lin_class	Logistic Regression Recursive Feature Elimination	Feature extraction
rfe_non_lin_class	Decision Tree Recursive Feature Elimination	Feature extraction
isolation_forest_reg	Regression Isolation Forest	Regression anomaly detection
isolation_forest_class	Classification Isolation Forest	Classification anomaly detection
ransac_lin_reg	Regression Random Sample Consensus	Outlier detection
ransac_non_lin_reg	Decision Tree Random Sample Consensus	Outlier detection
pca	Principal Component Analysis	Dimensionality reduction
kernel_pca	Kernel Principal Component Analysis	Dimensionality reduction
fast_ica	Independent Component Analysis	Feature extraction
poly_features	Polynomial Features	Feature engineering
decompose	Diff regression prediction and target for new target	Feature extraction
class_decompose	Diff classification prediction and target for new target	Feature extraction
cntvect	Count Vectorizer	Text feature extraction
text_clean	Lemmatization and Stemming	Text data processing
tfidf	TF-IDF Vectorizer	Text feature extraction
word2vec_pretrained	Text vectorization	Text feature extraction
lagged	Time series to the Hankel matrix transformation	Time series transformation
sparse_lagged	As lagged but with sparsing	Time series transformation
smoothing	Moving average	Time series transformation
gaussian_filter	Gaussian Filter	Time series transformation
diff_filter	Derivative Filter	Time series transformation
cut	Cut timeseries	Timeseries transformation
scaling	Scaling	Feature scaling
normalization	Normalization	Feature normalization
simple_imputation	Imputation	Data imputation
one_hot_encoding	One-Hot Encoder	Feature encoding
label_encoding	Label Encoder	Feature encoding
resample	Imbalanced binary class transformation in classification	Data transformation
topological_features	Calculation of topological features	Time series transformation

Feature transformation operations implementations

API name	Model used	Presets
rfe_lin_reg	sklearn.feature_selection.RFE
rfe_non_lin_reg	sklearn.feature_selection.RFE
rfe_lin_class	sklearn.feature_selection.RFE
rfe_non_lin_class	sklearn.feature_selection.RFE
isolation_forest_reg	sklearn.ensemble.IsolationForest
isolation_forest_class	sklearn.ensemble.IsolationForest
ransac_lin_reg	sklearn.linear_model.RANSACRegressor	fast_train *tree
ransac_non_lin_reg	sklearn.linear_model.RANSACRegressor	*tree
pca	sklearn.decomposition.PCA	fast_train ts *tree
kernel_pca	sklearn.decomposition.KernelPCA	ts *tree
fast_ica	sklearn.decomposition.FastICA	ts *tree
poly_features	sklearn.preprocessing.PolynomialFeatures
decompose	FEDOT model	fast_train ts *tree
class_decompose	FEDOT model	fast_train *tree
cntvect	sklearn.feature_extraction.text.CountVectorizer
text_clean	nltk.stem.WordNetLemmatizer nltk.stem.SnowballStemmer
tfidf	sklearn.feature_extraction.text.TfidfVectorizer
word2vec_pretrained	Gensin-data model
lagged	FEDOT model	fast_train ts
sparse_lagged	FEDOT model	fast_train ts
smoothing	FEDOT model	fast_train ts
gaussian_filter	FEDOT model	fast_train ts
diff_filter	FEDOT model	ts
cut	FEDOT model	fast_train ts
scaling	sklearn.preprocessing.StandardScaler	fast_train ts *tree
normalization	sklearn.preprocessing.MinMaxScaler	fast_train ts *tree
simple_imputation	sklearn.impute.SimpleImputer	fast_train *tree
one_hot_encoding	sklearn.preprocessing.OneHotEncoder
label_encoding	sklearn.preprocessing.LabelEncoder	fast_train *tree
resample	FEDOT model using sklearn.utils.resample
topological_features	FEDOT model	ts
fast_topological_features	FEDOT model	ts

Models used

Using the parameter preset of the main API you can specify what models are available during the learning process.

It influences:

• composing speed and quality

• computational behaviour

• task relevance

For example, 'best_quality' option allows FEDOT to use entire list of available models for a specified task. In contrast 'fast_train' ensures only fast learning models are going to be used.

Apart from that there are other options whose names speak for themselves: 'stable', 'auto', 'gpu', 'ts', 'automl' (the latter uses only AutoML models as pipeline nodes).

Note

To make it simple, FEDOT uses auto by default to identify the best choice for you.

Available models definitions

API name	Definition	Problem
adareg	AdaBoost regressor	Regression
catboostreg	Catboost regressor	Regression
dtreg	Decision Tree regressor	Regression
gbr	Gradient Boosting regressor	Regression
knnreg	K-nearest neighbors regressor	Regression
lasso	Lasso Linear regressor	Regression
lgbmreg	Light Gradient Boosting Machine regressor	Regression
linear	Linear Regression regressor	Regression
rfr	Random Forest regressor	Regression
ridge	Ridge Linear regressor	Regression
sgdr	Stochastic Gradient Descent regressor	Regression
svr	Linear Support Vector regressor	Regression
treg	Extra Trees regressor	Regression
xgbreg	Extreme Gradient Boosting regressor	Regression
bernb	Naive Bayes classifier (multivariate Bernoulli)	Classification
catboost	Catboost classifier	Classification
cnn	Convolutional Neural Network	Classification
dt	Decision Tree classifier	Classification
knn	K-nearest neighbors classifier	Classification
lda	Linear Discriminant Analysis	Classification
lgbm	Light Gradient Boosting Machine classifier	Classification
logit	Logistic Regression classifier	Classification
mlp	Multi-layer Perceptron classifier	Classification
multinb	Naive Bayes classifier (multinomial)	Classification
qda	Quadratic Discriminant Analysis	Classification
rf	Random Forest classifier	Classification
svc	Support Vector classifier	Classification
xgboost	Extreme Gradient Boosting classifier	Classification
kmeans	K-Means clustering	Clustering
ar	AutoRegression	Forecasting
arima	ARIMA	Forecasting
cgru	Convolutional Gated Recurrent Unit	Forecasting
ets	Exponential Smoothing	Forecasting
glm	Generalized Linear Models	Forecasting
locf	Last Observation Carried Forward	Forecasting
polyfit	Polynomial approximation	Forecasting
stl_arima	STL Decomposition with ARIMA	Forecasting
ts_naive_average	Naive Average	Forecasting

Available models implementations

API name	Model used	Presets
adareg	sklearn.ensemble.AdaBoostRegressor	fast_train ts *tree
catboostreg	catboost.CatBoostRegressor	*tree
dtreg	sklearn.tree.DecisionTreeRegressor	fast_train ts *tree
gbr	sklearn.ensemble.GradientBoostingRegressor	*tree
knnreg	sklearn.neighbors.KNeighborsRegressor	fast_train ts
lasso	sklearn.linear_model.Lasso	fast_train ts
lgbmreg	lightgbm.sklearn.LGBMRegressor	*tree
linear	sklearn.linear_model.LinearRegression	fast_train ts
rfr	sklearn.ensemble.RandomForestRegressor	fast_train *tree
ridge	sklearn.linear_model.Ridge	fast_train ts
sgdr	sklearn.linear_model.SGDRegressor	fast_train ts
svr	sklearn.svm.LinearSVR
treg	sklearn.ensemble.ExtraTreesRegressor	*tree
xgbreg	xgboost.XGBRegressor	*tree
bernb	sklearn.naive_bayes.BernoulliNB	fast_train
catboost	catboost.CatBoostClassifier	*tree
cnn	FEDOT model
dt	sklearn.tree.DecisionTreeClassifier	fast_train *tree
knn	sklearn.neighbors.KNeighborsClassifier	fast_train
lda	sklearn.discriminant_analysis.LinearDiscriminantAnalysis	fast_train
lgbm	lightgbm.sklearn.LGBMClassifier
logit	sklearn.linear_model.LogisticRegression	fast_train
mlp	sklearn.neural_network.MLPClassifier
multinb	sklearn.naive_bayes.MultinomialNB	fast_train
qda	sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis	fast_train
rf	sklearn.ensemble.RandomForestClassifier	fast_train *tree
svc	sklearn.svm.SVC
xgboost	xgboost.XGBClassifier	*tree
kmeans	sklearn.cluster.Kmeans	fast_train
ar	statsmodels.tsa.ar_model.AutoReg	fast_train ts
arima	statsmodels.tsa.arima.model.ARIMA	ts
cgru	FEDOT model	ts
ets	statsmodels.tsa.exponential_smoothing.ets.ETSModel	fast_train ts
glm	statsmodels.genmod.generalized_linear_model.GLM	fast_train ts
locf	FEDOT model	fast_train ts
polyfit	FEDOT model	fast_train ts
stl_arima	statsmodels.tsa.api.STLForecast	ts
ts_naive_average	FEDOT model	fast_train ts