5. Unsupervised Learning, PCA, Anomaly & Time Series

Sections 21-26: KMeans, DBSCAN, PCA Dimensionality Reduction, Isolation Forest, Recommendation Systems & Time Series ML

Module 1: Clustering, PCA & Anomaly Detection

Uncovering dataset clusters, reducing feature dimensionality, and detecting rare anomalies.

🇬🇧 English Definition

KMeans partitions data into K centroid clusters. PCA (Principal Component Analysis) project features into orthogonal components maximizing variance. Isolation Forest identifies easily isolated anomaly outliers.

🇮🇳 Hinglish Definition

KMeans centroid-based clustering karta hai. PCA high-dimensional data ko compress karke principal components me convert karta hai. Isolation Forest anomalies ko isolate karke detect karta hai.

Core Concepts & Principles

  • KMeans: Requires feature scaling; minimizes within-cluster sum of squares.
  • DBSCAN: Density-based clustering capable of discovering arbitrary shapes and identifying noise points.
  • PCA: Unsupervised linear dimensionality reduction projecting data along directions of maximum variance.

🔍 Line-by-Line Code Breakdown

  1. Line 1-2: Import KMeans clustering and PCA dimensionality reduction classes.
  2. Line 3-4: Fit PCA with n_components=2 to compress feature matrix.

🎯 Interview Focus & Questions

  • Q: What is PCA?
    A: Principal Component Analysis is an unsupervised linear dimensionality reduction technique that projects features into orthogonal components ordered by explained variance.
💻 Practical AI/ML Code Example - Clustering, PCA & Anomaly Detection 
from sklearn.cluster import KMeans
from sklearn.decomposition import PCA
from sklearn.ensemble import IsolationForest

pca = PCA(n_components=2)
print("PCA and Unsupervised Algorithms Ready.")
📤 Execution Output / Interpretation:
PCA and Unsupervised Algorithms Ready.

Module 2: Recommendation Systems & Time Series Machine Learning

Content-based vs collaborative recommendations and chronological time series forecasting.

🇬🇧 English Definition

Recommendation systems use Content-Based (TF-IDF similarity) or Collaborative Filtering (user-item interaction matrix). Time Series ML requires chronological splits, lag features, and rolling window features.

🇮🇳 Hinglish Definition

Content-based item attributes se similarity dekhta hai. Collaborative filtering similar users ke behavior se recommend karta hai. Time Series ML me time order preserve karna mandatory hai.

Core Concepts & Principles

  • Chronological Split: Never randomly shuffle time series data; split past data for training and future data for testing.
  • Lag Features: Past observations (df['sales'].shift(1)) become predictor features for future values.

🔍 Line-by-Line Code Breakdown

  1. Line 1-2: Create lag_1 feature shifting sales column by 1 step: df['lag1'] = df['sales'].shift(1).
  2. Line 3: Create rolling average feature: df['roll7'] = df['sales'].shift(1).rolling(7).mean().

🎯 Interview Focus & Questions

  • Q: Why avoid random split in time series?
    A: Random splitting causes temporal data leakage by using future data to predict past events.
💻 Practical AI/ML Code Example - Recommendation Systems & Time Series Machine Learning 
import pandas as pd

// Time series lag feature generation
// df["lag1"] = df["sales"].shift(1)
// df["roll7"] = df["sales"].shift(1).rolling(7).mean()
print("Time Series Feature Engineering Pipeline Ready.")
📤 Execution Output / Interpretation:
Time Series Feature Engineering Pipeline Ready.