ML algorithm types

ML algorithm types:

After going over that terminology -the question then is then how is an algorithm type chosen?

Supervised Learning (target variable exists)

if the variable is continuous 👍🏽 : Regression

1. Linear; Penalized Regression/LASSO
2. Logistic
3. Classification and regression tree (cart)
4. Random forest

if the variable is categorical 👍🏽 : Classification

1. Logistic
2. Support Vector Machine (if linear)
3. KNN K nearest neighbor (if linear)
4. CART - Classification and regression tree (non linear)

Unsupervised learning (target variable does not exist)

if continuous/continuous : Dimensionality reduction (A) or clustering (#) A) Principal Components analysis (PCA)

1. K-means
2. Hierarchical

General Steps for choosing ML algorithm

1. Decide if the data is complex (aka too many features -x variables). If so, apply dimension reduction algorithm

2. Decide if problem is classification - if yes, go to step 3. If this is a numerical prediction : -use penalized regression if the data is linear -or if nonlinear or complex use CART, Random forest, or neural network

3. Is it supervised classification, if not: go to step 4. if it is supervised: -for linear data use KNN, or SVM -for complex data use CART, Random forest, neural network

4. For unsupervised classification -for non linear data - use k means if the number categories is unknown. If the number categories is known then hierarchical networks -if complex data- use neural networks