🇬🇧 English | 🇹🇼 繁體中文

Machine Learning Note

• Data Preprocessing
• Regularized Regression & MARS & PLS & kernel overview
• Ensemble Learning
• Evaluation Metrics

Useful package

Feature engineering

• Category Encoders:encode categorical variable
• Featuretools: be utilized with compose & EvalML
  • Featuretools automates the feature engineering process
  • EvalML automates model building, includes data checks, and even offers tools for model understanding (Tutorial)
  • Compose automates prediction engineering
• feature-engine
  • Variable transformation, selection, preprocessing
  • Imputatoin, Encoding, Discretization, Outlier Handling
  • Time series features
• imbalance: deal with imbalance data issue
  • sklearn resample
  • tensorflow tf.data_sampler
  • pytorch torch.utils.data.WeightedRandomSampler

Feature selection

• feature-engine
• mlxtend: wrapper, greedy algorithm, etc (also include some data visualization tool)

Modeling

• sklearn
• xgboost
• lightgbm
• pyearch: Multivariate Adaptive regression spline
• scikit-multilearn: Multi-label classification with focus on label space manipulation
• seglearn: Time series and sequence learning using sliding window segmentation
• pomegranate: Probabilistic modelling for Python, with an emphasis on hidden Markov models. (GMM, HMM, Naive Bayes and Bayes Classifiers, Markov Chains, Discrete Bayesian Networks, Discrete Markov Networks)

Hyperparameter

• sklearn-deap: Use evolutionary algorithms instead of gridsearch in scikit-learn.
• hyperopt: bayesian hyperparameter
• scikit-optimize: scikit learn hyperparamter
• Optuna: Guidance for Optuna

Time series

• tslearn: time series preprocessing, feature extraction, classification, regression, clustering
• sktime: time series classification, regression, clustering, annotation (also can be used in data that is univariate, multivariate or panel)
• HMMLearn: Implementation of hidden markov models
• pytorchforecasting: time series forecasting model implemented by pytorch

AutoML

• auto-sklearn
• Compose, Featuretools, EvalML
• tpot

Anomaly detection

• PyOD for tabular dataset
• TODS for time seires
• Anomaly detection for different types of dataset

MLOps

• mlflow
• clearml
• wandb
• dvc: command line usage (not only included data version control)
• comet
• Neptune.ai

A/B testing

• Learning Resource

Lazyprediction for lists of models

Scikit Learn related projects