SUK · Third Year B.Tech. Computer Science and Engineering (Data Science) · NEP 2020
Feature Engineering
DS3203 · Semester VI · 3 credits · ISE 10 + mid-semester 30 + ESE 60
Unit 1: Introduction to Feature Engineering
6 hoursWhy feature engineering matters; evaluating ML algorithms and feature-engineering procedures; feature understanding, improvement (cleaning), selection, construction, transformation and learning.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Feature engineering
- Data quality dimensions
- Data integration and transformation
Unit 2: Basics of Feature Representation
7 hoursScalars, vectors and spaces; counts, binarisation, quantisation or binning and log transformation; feature scaling: min-max, standardisation and ℓ2 normalisation; interaction features and feature selection.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Binning, binarisation and interaction features
- Square-root, arcsine, log and logit transforms
- Normalisation and standardisation
Unit 3: Features of Text and Categorical Data
8 hoursBag-of-words vectors, filtering, n-grams and phrases, TF-IDF; encoding categorical variables and handling large categorical variables.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Word representations
- TF-IDF and vector space models
- Encoding categorical variables: one-hot, ordinal and target encoding
Unit 4: Feature Selection
7 hoursGoals and classes of feature-selection methods; the effect of irrelevant features; overfitting to predictors and external validation; greedy search: simple filters, recursive feature elimination and stepwise selection.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Feature selection: filter, wrapper and embedded methods
Unit 5: Feature Transformations
8 hoursLinear projection, variance and the derivation of principal components; implementing PCA, whitening and ZCA; limitations and use cases of PCA.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Feature extraction and PCA
- Principal component analysis
- Linear discriminant analysis
Unit 6: Feature Learning
6 hoursParametric assumptions and the non-parametric fallacy; feature-learning algorithms; the Bernoulli RBM and extracting PCA and RBM components from MNIST; RBMs in an ML pipeline; word vectorisations and embeddings and their use in information retrieval.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Boltzmann machines and deep belief networks
- Word embeddings
- Self-supervised learning and AutoML
Unit contents are summarised from the official SUK syllabus . AiCanCode.org is not affiliated with Shivaji University, Kolhapur.