Table 1 A brief description of the most frequently used AI modeling techniques and advantages and disadvantages

Logistic Regression

A statistical model used for predicting binary outcomes based on one or more predictor variables

Easy to implement, interpretability, and computationally efficient

Assumes linearity, not suitable for complex relationships

Decision Trees and Ensemble Variants (Random Forest, Extra Trees, Optimal Classification Trees)

Models that use tree structures for decision-making, including individual trees and ensembles like Random Forest, Extra Trees, and Optimal Classification Trees

Easy to visualize and interpret, reduces overfitting with ensembles, handles high-dimensional data well

Prone to overfitting (single trees), less interpretable (ensembles), computationally intensive

Support Vector Machine (SVM)

A supervised learning model used for classification and regression that finds the optimal hyperplane

Effective in high-dimensional spaces, robust to overfitting

Requires careful parameter tuning, less effective on large datasets

Gradient Boosting (GBM, XGBoost, LightGBM, CatBoost)

An ensemble method that builds models sequentially, correcting previous errors; includes variants like XGBoost, LightGBM, and CatBoost

High accuracy, speed, scalability, automatic handling of categorical variables

Prone to overfitting, requires careful tuning, high computational demands

Neural Networks and Deep Learning (DNN, CNN)

Models inspired by the human brain, consisting of layers of interconnected nodes (neurons); includes DNNs and CNNs for complex representations

Capable of capturing complex patterns, highly flexible, exceptional at processing unstructured data

Requires large datasets, extensive computational resources, can be a "black box"

Naive Bayes

A probabilistic classifier based on Bayes' theorem, assuming independence among predictors

Simple, fast, performs well with small datasets

Assumes feature independence, less accurate with correlated features

K-Nearest Neighbors (KNN)

A non-parametric model that classifies based on the closest training examples in the feature space

Simple to implement, intuitive

Sensitive to irrelevant features and data scale

Adaptive Boosting (AdaBoost)

An ensemble technique that adjusts weights of misclassified instances in successive iterations

Improves weak learners, robust to noise

Sensitive to outliers, may overfit with noisy data

Linear Discriminant Analysis (LDA)

A statistical method for classifying samples based on linear combinations of features

Handles multiclass classification, interpretable

Assumes normal distribution of features, linear boundaries