Advanced Certification in Data Science & Analytics

Machine Learing

1. Introduction to Machine Learning

   • What is ML? Difference between AI, ML and Deep Learning
   • Types of ML: Supervised, Unsupervised
   • ML Workflow: Data → Train → Evaluate → Predict
   • Key terms: Features, Labels, Training set, Test set
   • Overfitting vs Underfitting — visual intuition

2. Data Preprocessing & Feature Engineering

   • Handling missing values: drop vs mean/median/mode imputation
   • Encoding categoricals: Label Encoding and One-Hot Encoding
   • Feature Scaling: MinMax Normalization vs Standardization
   • Train-Test Split and why shuffling matters
   • Hands-on: Build a full preprocessing pipeline in scikit-learn

3. Regression – Linear & Multiple

   • What is Regression? Predicting continuous values
   • Simple Linear Regression: equation, slope and intercept
   • Multiple Linear Regression: using multiple features
   • Evaluation metrics: MAE, RMSE, R-squared
   • Hands-on: Train and evaluate a Linear Regression model

4. Regression – Polynomial & Regularization

   • Polynomial Regression: fitting curves to non-linear data
   • Problem of overfitting with high-degree polynomials
   • Ridge Regression (L2): shrinking large coefficients
   • Lasso Regression (L1): coefficient shrinkage + feature selection
   • Hands-on: Compare Linear vs Ridge vs Lasso on the same dataset

5. Classification – Logistic Regression

   • What is Classification? Binary vs Multi-class problems
   • Logistic Regression: Sigmoid function and decision boundary
   • Confusion Matrix: TP, TN, FP, FN explained
   • Evaluation metrics: Accuracy, Precision, Recall, F1-Score
   • Hands-on: Binary and multi-class classification with Logistic Regression

6. Classification – KNN & Naive Bayes

   • K-Nearest Neighbours (KNN): distance-based prediction intuition
   • Choosing the right K with cross-validation
   • Naive Bayes: applying Bayes’ theorem to classification
   • Gaussian vs Multinomial Naive Bayes — when to use each
   • Hands-on: KNN vs Naive Bayes comparison on same dataset

7. Classification – SVM

   • Support Vector Machine: hyperplane and margin concept
   • Hard Margin vs Soft Margin SVM
   • Kernel Trick: Linear, RBF and Polynomial kernels (intuition)
   • SVM for multi-class classification
   • Hands-on: SVM with different kernels — compare decision boundaries

8. Classification – Decision Trees & Random Forest

   • Decision Tree: splitting criteria — Gini Impurity and Information Gain
   • Visualising a trained Decision Tree
   • Overfitting in trees and pruning concept
   • Random Forest: ensemble of trees and feature importance
     Hands-on: Decision Tree vs Random Forest — accuracy and feature importance

9. Clustering – K-Means

   • What is Clustering? Unsupervised learning use cases
   • K-Means algorithm: initialise → assign → update → repeat
   • Choosing K: Elbow Method and Inertia
   • Limitations of K-Means: outliers and non-spherical clusters
   • Hands-on: K-Means clustering with scatter plot visualisation

10. Clustering – Hierarchical & DBSCAN

    • Hierarchical Clustering: Agglomerative approach step-by-step
    • Dendrogram: reading and cutting to select cluster count
    • DBSCAN: Core, Border and Noise points concept
    • When to use DBSCAN vs K-Means vs Hierarchical
    • Hands-on: Compare all three clustering methods on one dataset

11. Dimensionality Reduction – PCA

    • Curse of Dimensionality: why too many features hurt models
    • Principal Component Analysis (PCA): intuition and purpose
    • Explained Variance Ratio: choosing number of components
    • Applying PCA before classification and clustering
    • Hands-on: PCA on a high-dimensional dataset — visualise in 2D

12. Model Evaluation & Hyperparameter Tuning

    • K-Fold and Stratified K-Fold Cross Validation
    • ROC Curve and AUC Score for classifier comparison
    • Handling class imbalance: class_weight parameter
    • Hyperparameter tuning with GridSearchCV
    • Hands-on: Full model selection pipeline — tune, validate and compare

DL-AI

1. Introduction to Neural Networks

   • Biological neuron vs Artificial neuron analogy
     Perceptron: inputs, weights, bias and activation
     Limitation of single Perceptron — XOR problem
     Multi-Layer Perceptron (MLP): hidden layers intuition
     Activation functions: Sigmoid, Tanh, ReLU — when to use which

2. ANN Architecture & Forward Propagation

   • ANN layers: Input, Hidden and Output
   • Weights, Biases and forward pass walkthrough
   • Output activations: Sigmoid (binary), Softmax (multi-class)
   • Loss Functions: MSE for regression, Cross-Entropy for classification
   • Hands-on: Build and run a forward pass using Keras Sequential API

3. Training an ANN – Backpropagation & Optimizers

   • Gradient Descent: rolling downhill analogy
   • Backpropagation: how errors flow backwards (concept only)
   • Learning Rate: effect of too high vs too low
   • Optimizers: SGD, Adam — intuition and practical difference
   • Hands-on: Train an ANN and read training vs validation loss curves

4. ANN Regularization & Applied Tasks

   • Overfitting in ANNs: signs from loss curves
   • Dropout: randomly disabling neurons during training
   • Early Stopping: halt when validation loss stops improving
   • ANN for Classification and Regression — architecture differences
   • Hands-on: Add Dropout + Early Stopping; save and reload model

NLP

1. NLP – Text Preprocessing

   • What is NLP? Real-world applications overview
   • Text cleaning: lowercase, remove punctuation and noise
   • Tokenization: word and sentence tokenization with NLTK
   • Stop Word Removal
   • Hands-on: Build a reusable text cleaning pipeline in Python

2. NLP – Stemming, Lemmatization & POS Tagging

   • Stemming: PorterStemmer — reducing words to root form
   • Lemmatization: WordNetLemmatizer — linguistically correct roots
   • Stemming vs Lemmatization — when to use which
   • POS Tagging: labelling Nouns, Verbs, Adjectives with NLTK
   • Named Entity Recognition (NER): extracting people, places and orgs with spaCy

3. NLP – Text Representation (BoW & TF-IDF)

   • Why machines need numbers — text to vectors
   • Bag of Words (BoW): vocabulary, sparse matrix concept
   • CountVectorizer in scikit-learn
   • TF-IDF: rewarding rare important words, penalising common ones
   • Hands-on: Build BoW and TF-IDF matrices and compare them

4. NLP – Sentiment Analysis & Text Classification

   • Sentiment Analysis: Positive, Negative, Neutral
   • Rule-based approach: VADER sentiment analyser
   • ML-based approach: Logistic Regression on TF-IDF features
   • Evaluation: Accuracy, Precision, Recall, F1 for text tasks
   • Hands-on: Sentiment classifier on a real product reviews dataset