Mengenal Algoritma Machine Learning dan AI
Posted in Artikel on April 2, 2026 by Roberto Kaban ‐ 5 min read
Di era digital saat ini, kemampuan memahami dan menerapkan algoritma machine learning (ML) dan artificial intelligence (AI) menjadi salah satu kompetensi penting bagi mahasiswa. Algoritma-algoritma ini digunakan untuk analisis data, sistem rekomendasi, pengenalan wajah, prediksi finansial, pengolahan bahasa alami (NLP), dan berbagai aplikasi inovatif lainnya.
Artikel ini menyajikan kumpulan algoritma ML dan AI beserta kategori, domain aplikasi, dan keunggulannya. Informasi ini disusun agar mahasiswa dapat memahami konsep, membandingkan algoritma, dan memilih metode yang sesuai untuk tugas kuliah, proyek riset, atau eksperimen data.
Saya mengelompokkan algoritma berdasarkan pendekatan utama: Supervised Learning, Unsupervised Learning, Reinforcement Learning, Hybrid Models, hingga teknik Few-shot & Self-supervised Learning.
Supervised Learning
| Kategori / Model | Domain / Aplikasi & Kelebihan Utama |
|---|---|
| Linear Regression (Supervised Klasik / Statistical) | Prediksi numerik — Sederhana, interpretasi mudah |
| Logistic Regression (Supervised Klasik / Statistical) | Klasifikasi biner — Cepat, baseline classification |
| Ridge / Lasso / Elastic Net (Supervised Klasik / Statistical) | Tabular regression — Regularisasi, mengurangi overfitting |
| ARIMA / SARIMA / VARIMA (Supervised Klasik / Statistical) | Time series forecasting — Efektif untuk data linear |
| Naive Bayes (Supervised Klasik / Statistical) | NLP, klasifikasi teks — Cepat, sederhana |
| Hidden Markov Model (HMM) (Supervised Klasik / Statistical) | Sequential data — Probabilistic sequence modeling |
| Decision Tree (Supervised Modern ML) | Tabular data — Interpretasi mudah |
| Random Forest (Supervised Modern ML) | Tabular, prediksi bisnis — Robust terhadap overfitting |
| Extra Trees (Supervised Modern ML) | Tabular data — Variance rendah, cepat |
| Gradient Boosting Machine (GBM) (Supervised Modern ML) | Tabular data — Akurat, menangani data numerik/kategorikal |
| General Linear Model (GLM) (Supervised Modern ML) | Regression & classification — Fleksibel untuk berbagai distribusi output |
| XGBoost (Supervised Modern ML) | Tabular data, kompetisi ML — Performa tinggi, cepat |
| LightGBM (Supervised Modern ML) | Tabular data — Cepat & hemat memori |
| CatBoost (Supervised Modern ML) | Data kategorikal — Optimal untuk categorical features |
| CNN (ResNet, EfficientNet, DenseNet, MobileNet) (Supervised Modern ML) | Computer vision — Memahami pola spasial |
| RNN / LSTM / GRU (Supervised Modern ML) | NLP, time series — Memproses data sekuensial |
| Temporal Convolutional Network (TCN) (Supervised Modern ML) | Time series — Alternatif RNN, paralelisasi lebih baik |
| Transformer (BERT, RoBERTa, GPT-4, T5, XLNet) (Supervised Modern ML) | NLP, QA, summarization — Memahami konteks global |
| Vision Transformer (ViT, DeiT, Swin Transformer) (Supervised Modern ML) | Computer vision — Alternatif CNN, performa tinggi |
| CLIP (OpenAI) (Supervised Modern ML) | Multimodal (gambar & teks) — Memahami hubungan teks-gambar |
| Graph Neural Networks (GNN, GraphSAGE, GAT) (Supervised Modern ML) | Graph / social networks — Memanfaatkan struktur graf |
| TabNet (Supervised Modern ML) | Tabular data — Deep learning tabular, interpretabel |
| YOLO (You Only Look Once) (Supervised Modern ML) | Object detection, Computer vision — Cepat, real-time detection, akurat untuk multiple objek |
| Temporal Fusion Transformer (TFT) (Supervised Modern ML) | Time series forecasting — Memproses data sekuensial kompleks |
| Neural Collaborative Filtering (NCF) (Supervised Modern ML) | Recommender system — Personalized recommendations |
| Meta-learning / Few-shot (MAML, Prototypical Networks) (Supervised Modern ML) | Few-shot learning — Prediksi dengan data sangat sedikit |
| Adaptive Neuro-Fuzzy Inference System (ANFIS) (Supervised / Hybrid) | Regression, classification, time series — Neural network + fuzzy logic |
| Fuzzy Rule Learning (FRL) (Supervised / Rule Learning) | Classification, decision-making — Menemukan aturan fuzzy otomatis, interpretable |
Unsupervised Learning & Probabilistic / Bayesian Models
| Kategori / Model | Domain / Aplikasi & Kelebihan Utama |
|---|---|
| Fuzzy c-means (FCM) (Unsupervised / Fuzzy Clustering) | Clustering, pattern recognition — Membership probabilistik, fleksibel dari k-Means |
| Neuro-Fuzzy Deep Learning (Hybrid Neuro-Fuzzy) | Computer vision, time series, NLP — Representasi deep learning + aturan fuzzy |
| k-Means / k-Medoids (Unsupervised Klasik / Statistical) | Clustering, segmentasi — Cepat, sederhana |
| Hierarchical Clustering (Unsupervised Klasik / Statistical) | Customer profiling — Melihat struktur cluster bertingkat |
| Gaussian Mixture Model (GMM) (Unsupervised Klasik / Statistical) | Probabilistic clustering — Probabilitas anggota cluster |
| PCA (Principal Component Analysis) (Unsupervised Klasik / Statistical) | Dimensionality reduction — Menyederhanakan data |
| ICA (Independent Component Analysis) (Unsupervised Klasik / Statistical) | Signal processing, EEG — Memisahkan sumber independen |
| Self-Organizing Map (SOM) (Unsupervised Klasik / Statistical) | Visualisasi & clustering — Memetakan high-dimensional ke 2D |
| Autoencoder / VAE (Unsupervised Modern / Representation / Generative) | Feature extraction, anomaly detection — Representasi latent, fleksibel |
| Masked Autoencoder (MAE) | Self-supervised image representation — Memanfaatkan patch masking |
| GANs (StyleGAN, BigGAN, CycleGAN, InfoGAN) | Generative citra/video — Menghasilkan data realistis |
| Diffusion Models (Stable Diffusion, DDPM, Latent Diffusion) | Generative citra — Kualitas citra tinggi |
| Contrastive Learning (SimCLR, MoCo v2, BYOL, SwAV, SimSiam, DINO) | Self-supervised representation — Representasi kuat tanpa label |
| Deep Graph Infomax (DGI) | Graph embedding — Representasi node tanpa label |
| Node2Vec | Graph embedding — Representasi node/graf |
| Graph Contrastive Learning (GraphCL, GCA) | Graph embedding — Self-supervised, transfer learning |
| Deep Clustering (DEC, DeepCluster, IIC) | Image clustering — Feature learning + clustering |
| Bayesian Neural Networks (BNN) | Regression, classification — Uncertainty estimation |
| Gaussian Processes (GP) | Regression, forecasting — Non-parametric, probabilistic |
| Hidden Markov Model (HMM) | Sequential / time series — Probabilistic sequence modeling |
| Variational Bayesian methods | Probabilistic inference — Approximate Bayesian inference |
Reinforcement Learning, Hybrid Advanced, Few-shot / Self-supervised
| Kategori / Model | Domain / Aplikasi & Kelebihan Utama |
|---|---|
| Q-Learning / Deep Q-Network (DQN) (RL) | Game AI, robotics — Value-based RL, belajar policy |
| Policy Gradient / REINFORCE (RL) | Continuous action RL — Policy-based optimization |
| Actor-Critic (A3C, PPO, SAC) (RL) | Robotics, games — Kombinasi value & policy |
| AlphaZero / MuZero (RL) | Game AI, simulasi kompleks — RL + search, performa tinggi |
| Neural ODEs (Hybrid Models / Advanced) | Continuous-time modeling — Model dinamik, sekuensial |
| GBM + Neural Network hybrid | Tabular + deep learning — Performa maksimal tabular |
| Diffusion + Transformer hybrids | Generative multimodal — Generasi citra + teks berkualitas tinggi |
| Prototypical Networks / Matching Networks (Few-shot / Meta-learning) | Few-shot classification — Prediksi dengan data sangat sedikit |
| Masked Language Models (MLM, BERT pretraining) | Self-supervised NLP — Representasi teks tanpa label |
| Prompt-based LLMs (GPT, T5, LLaMA, Falcon) | Zero-shot / few-shot NLP — Prediksi dengan data minimal |
Semoga bermanfaat!
Machine Learning
Artificial Intelligence
Algoritma
Supervised
Unsupervised
Reinforcement Learning
Hybrid Models
comments powered by Disqus