Machine Learning Algorithm Cheat Sheet

This cheat sheet helps you choose the best machine learning algorithm for your predictive analytics solution.
Your decision is driven by both the nature of your data and the goal you want to achieve with your data.

Predict between Multiclass Classification

several categories Answers complex questions with
Text Analytics Extract information from text multiple possible answers
Answers questions like: Is this A or B or C or D?
Derives high-quality information from text
Multiclass Logistic Fast training times,
Answers questions like: What info is in this text? Regression linear model

What do you want to do?

Latent Dirichlet Unsupervised topic modeling, Multiclass Neural Accuracy, long training times
Allocation group texts that are similar Network

Extract N-Gram Creates a dictionary of n-grams Multiclass Decision Accuracy, fast training times
Features from Text from a column of free text Forest
Predict between
One-vs-All Depends on the
Converts text data to integer two categories two-class classifier
Multiclass
Feature Hashing encoded features using the
Vowpal Wabbit library Generate recommendations One-vs-One Depends on binary classifier,
less sensitive to an imbalanced
Multiclass dataset with larger complexity
Performs cleaning operations on text,
Preprocess Text like removal of stop-words, case Recommenders Multiclass Boosted Non-parametric, fast
normalization Decision Tree training times and scalable
Predicts what someone will be interested in
Converts words to values for use in
Word2Vector Answers the question: What will they be interested in? Two-Class Classification
NLP tasks, like recommender, named
entity recognition, machine Use the Train Wide & Deep Hybrid recommender, both collaborative
translation Answers simple two-choice questions,
Recommender module filtering and content-based approach
like yes or no, true or false
Collaborative filtering, better performance Answers questions like: Is this A or B?
SVD Recommender
with lower cost by reducing dimensionality Two-Class Support Under 100 features,
Predict Vector Machine linear model
Regression
values
Makes forecasts by estimating the Two-Class Averaged
Discover structure Perceptron
Fast training, linear model
relationship between values
Answers questions like: How much or how many? Two-Class Decision Accurate, fast training
Clustering Forest
Fast Forest Quantile
Predicts a distribution
Regression Separates similar data points into intuitive groups Two-Class Logistic
Fast training, linear model
Answers questions like: How is this organized? Regression
Poisson Regression Predicts event counts
K-Means Unsupervised learning Classify Two-Class Boosted Accurate, fast training,
Decision Tree large memory footprint
images
Linear Regression Fast training, linear model
Two-Class Neural Accurate, long training
Bayesian Linear Find unusual occurrences Network times
Linear model, small data sets
Regression
Image Classification
Decision Forest
Regression
Accurate, fast training times Anomaly Detection Classifies images with popular networks
Answers questions like: What does this image represent?
Neural Network Identifies and predicts rare or unusual data points
Accurate, long training times
Regression Answers the question: Is this weird? ResNet Modern deep
learning neural
Boosted Decision Accurate, fast training times, Under 100 features, PCA-Based Anomaly
One Class SVM Fast training times DenseNet network
Tree Regression large memory footprint aggressive boundary Detection

© 2021 Microsoft Corporation. All rights reserved. Share this poster: aka.ms/mlcheatsheet