BCA 304T Machine Learning Notes cover all 4 units of this GGSIPU BCA Semester 6 paper in 67 pages, with 15 solved numericals, 18 Python programs, 32 answered two-mark questions and 34 ten-mark questions mapped to their sections. They follow the BCA scheme for the batch admitted in 2024-25 onwards.
Written by Prateek Talwar, Founder · Reviewed by Ritik Rawat, B.Tech, Senior Full Stack Developer · Checked against the BCA Scheme of Examination and Syllabi (19.08.2026) · Updated 14 October 2026 · Editorial policy
Scheme: BCA four-year scheme (USICT) | Applies to: batch admitted 2024-25 onwards | Verified against: BCA Scheme of Examination and Syllabi, 19.08.2026
BCA 304T Machine Learning at a glance: GGSIPU BCA Sem 6 credits, units and exam pattern
| Code and title | BCA 304T · Machine Learning |
|---|---|
| Course and semester | BCA (USICT scheme) · Semester 6 |
| Category and credits | Core Course Theory (CCT) · L-T-C 4-0-4 · 4 lecture hours a week |
| Marks | 100: 40 internal + 60 end-term |
| Applies to | Batch admitted in 2024-25 onwards |
| End-term pattern | Q1 compulsory: 10 × 2 marks, at least 2 per unit; then one of two 10-mark questions per unit (20 + 40 = 60 marks) |
| Exam duration | 3 Hours |
| Units | 4, numbered I to IV without titles, 11 hours each |
| Paired lab | BCA 304P Machine Learning Lab · SEC · L-P-C 0-2-1 · 100 marks (40 + 60) |
BCA 304T is one of three core theory papers among the GGSIPU BCA Semester 6 papers; its listed pre-requisite is basic Python.
Evaluation and re-appear rules come from Ordinance 11 of the university.
BCA 304T Machine Learning Notes unit-wise: 15 solved numericals and 18 Python programs across 4 units
These GGSIPU BCA Sem 6 Machine Learning notes give 64 of their 67 pages to the units. The official PDF numbers units only, so headings use the notes' unit titles; the sample PDF above shows one page per unit.
Papers from every other semester are collected under GGSIPU BCA notes.
BCA 304T Unit 1: Introduction to Machine Learning, Regression and Classification
Pages 4 to 21, 11 sections: Mitchell's task-experience-performance definition, four types of learning, binary, multiclass and multilabel classification, the confusion matrix, precision-recall trade-off and ROC curve, then NumPy, Pandas, Scikit-Learn, linear regression and logistic regression. Count: 3 numericals, 6 programs, 8 two-mark and 8 ten-mark questions, plus a Key Formulas page.
| Measure | Working | Answer |
|---|---|---|
| Accuracy | (40 + 45) / 100 | 0.85 |
| Precision | 40 / (40 + 5) | 0.889 |
| Recall | 40 / (40 + 10) | 0.80 |
| F1-score | 2 × 0.889 × 0.80 / 1.689 | 0.842 |
Quick look: confusion matrix for 100 patients, TP 40, FN 10, FP 5, TN 45 (from page 11)
- Measure
- Accuracy
- Working
- (40 + 45) / 100
- Answer
- 0.85
- Measure
- Precision
- Working
- 40 / (40 + 5)
- Answer
- 0.889
- Measure
- Recall
- Working
- 40 / (40 + 10)
- Answer
- 0.80
- Measure
- F1-score
- Working
- 2 × 0.889 × 0.80 / 1.689
- Answer
- 0.842
BCA 304T Unit 2: Supervised Learning Algorithms
Pages 22 to 38, 9 sections: decision trees and pruning, rule-based classification with sequential covering, Naïve Bayes with Laplace smoothing, a Bayesian network on the burglary-alarm example, SVM kernels, k-NN, and bagging, boosting and Random Forest. Count: 4 numericals, 5 programs, 8 two-mark and 9 ten-mark questions, plus a table comparing eight algorithms.
| Sample (X1, X2) | Class | Distance | Rank |
|---|---|---|---|
| P1 (7, 7) | Bad | 4.00 | 3 |
| P2 (7, 4) | Bad | 5.00 | 4 |
| P3 (3, 4) | Good | 3.00 | 1 |
| P4 (1, 4) | Good | 3.61 | 2 |
Quick look: k-NN with k = 3 for a new tissue at (3, 7) (from page 33)
- Sample (X1, X2)
- P1 (7, 7)
- Class
- Bad
- Distance
- 4.00
- Rank
- 3
- Sample (X1, X2)
- P2 (7, 4)
- Class
- Bad
- Distance
- 5.00
- Rank
- 4
- Sample (X1, X2)
- P3 (3, 4)
- Class
- Good
- Distance
- 3.00
- Rank
- 1
- Sample (X1, X2)
- P4 (1, 4)
- Class
- Good
- Distance
- 3.61
- Rank
- 2
BCA 304T Unit 3: Artificial Neural Networks
Pages 39 to 53, 13 sections: McCulloch-Pitts, HebbNet, Perceptron, Adaline, multilayer architecture, activation and loss functions, gradient descent, backpropagation and its variants, and regularisation with L1, L2, dropout and early stopping. Count: 4 numericals (one is a 2-2-1 backpropagation pass), 2 programs, 8 two-mark and 9 ten-mark questions.
Quick look: perceptron training in four steps (from page 42)
- Start weights and bias at zero; pick learning rate η.
- Compute net = Σ wixi + b and y = step(net).
- Only if y ≠ t: wi += η(t − y)xi, b += η(t − y).
- Stop after an epoch with no change; for AND that is epoch 6, giving w1 = 2, w2 = 1, b = −2.
BCA 304T Unit 4: Unsupervised Learning and Dimensionality Reduction
Pages 54 to 67, 7 sections: six families of clustering, K-means with the elbow and silhouette methods, agglomerative and divisive clustering with dendrograms, the Kohonen SOM, feature selection by filter, wrapper and embedded methods, and PCA in five steps. Count: 4 numericals, 5 programs, 8 two-mark and 8 ten-mark questions.
| Iteration | Cluster 1 | Cluster 2 | Cluster 3 |
|---|---|---|---|
| 1 | A1 | A3, B1, B2, B3, C2 | A2, C1 |
| 2 | A1, C2 | A3, B1, B2, B3 | A2, C1 |
| 3 | A1, B1, C2 | A3, B2, B3 | A2, C1 |
| 4 | A1, B1, C2 | A3, B2, B3 | No change |
Quick look: K-means on eight points converging in four iterations (from page 57)
- Iteration
- 1
- Cluster 1
- A1
- Cluster 2
- A3, B1, B2, B3, C2
- Cluster 3
- A2, C1
- Iteration
- 2
- Cluster 1
- A1, C2
- Cluster 2
- A3, B1, B2, B3
- Cluster 3
- A2, C1
- Iteration
- 3
- Cluster 1
- A1, B1, C2
- Cluster 2
- A3, B2, B3
- Cluster 3
- A2, C1
- Iteration
- 4
- Cluster 1
- A1, B1, C2
- Cluster 2
- A3, B2, B3
- Cluster 3
- No change
BCA 304T exam corner: 32 two-mark answers and 34 ten-mark questions by unit
Each unit ends with "2 Marks" questions with answers and "10 Marks" questions mapped to sections.
| Unit | 2 Marks | 10 Marks | Numericals | Programs |
|---|---|---|---|---|
| 1: Regression and classification | 8 | 8 | 3 | 6 |
| 2: Supervised algorithms | 8 | 9 | 4 | 5 |
| 3: Neural networks | 8 | 9 | 4 | 2 |
| 4: Clustering and PCA | 8 | 8 | 4 | 5 |
BCA 304T question bank in the notes, unit by unit
- Unit
- 1: Regression and classification
- 2 Marks
- 8
- 10 Marks
- 8
- Numericals
- 3
- Programs
- 6
- Unit
- 2: Supervised algorithms
- 2 Marks
- 8
- 10 Marks
- 9
- Numericals
- 4
- Programs
- 5
- Unit
- 3: Neural networks
- 2 Marks
- 8
- 10 Marks
- 9
- Numericals
- 4
- Programs
- 2
- Unit
- 4: Clustering and PCA
- 2 Marks
- 8
- 10 Marks
- 8
- Numericals
- 4
- Programs
- 5
- Unit 1: What does AUC = 0.5 mean? Pointer: the model is no better than random guessing.
- Unit 2: Differentiate bagging and boosting. Pointer: parallel bootstrap models cut variance; sequential error-fixing models cut bias.
- Unit 3: What is the vanishing gradient problem? Pointer: sigmoid and tanh gradients shrink going backwards; ReLU reduces it.
- Unit 4: What is the curse of dimensionality? Pointer: more features make data sparse and distances lose meaning.
The BCA 304T notes answer 32 two-mark questions and map 34 ten-mark questions across 4 units, plus 15 solved numericals and 18 Python programs.
For a fresh set to practise, open the BCA 304T important questions.
Who writes and checks the BCA 304T Machine Learning notes?
- Prateek Talwar, author: Founder, Unnati Educations; ACS, ICSI; B.Com, University of Delhi. Profile
- Ritik Rawat, subject reviewer: Senior Full Stack Developer, Unnati Educations; B.Tech. Profile
These BCA 304T Machine Learning IPU notes are mapped to the BCA scheme for the 2024-25 batch, whose third and fourth-year syllabi were approved on 19.08.2026. The 2026-27 edition credits the prescribed textbooks where it uses them, such as Géron (TB1) for the Keras model and Fausett (TB3) for the network rules. Last updated 14 October 2026, under our editorial policy.
The same team prepares the GGSIPU important questions for every course we cover.
How to write BCA 304T answers: two-mark definitions, solved numericals and ten-mark algorithm answers
Question 1 carries ten 2-mark parts, so give a definition plus one example or formula in two or three lines. The tip on page 8 adds that for an "applications" question, naming the learning type beside each application shows understanding, not just a list.
Worked example (Unit 4, 2 Marks): What is the elbow method? Plot WCSS against the number of clusters k; the point where the fall in WCSS sharply slows, the elbow, is the best k. Two lines: what you plot, how you read it.
Numericals turn up in every unit, from information gain to PCA. The solved versions follow one layout: formula, values in a table, then a one-line result such as "Outlook becomes the root node" or "PC1 keeps 82.1% of the variance". Carry three decimals as the notes do, so your working matches.
For 10-mark questions, each list gives a "Where to look" column, such as Section 10 for backpropagation. Build the answer as definition, diagram, algorithm steps, a small worked case, then advantages and limitations. The page 36 tip sets the depth: a Random Forest answer names both sources of randomness (bootstrap samples and random feature subsets), the OOB error and feature importance.
Write code only when asked, as in Unit 4's "Write Python code to implement K-means, hierarchical clustering and PCA", using KMeans, AgglomerativeClustering and PCA from Scikit-Learn.
Then practise the format on other papers through the GGSIPU BCA important questions.
BCA 304T and the BCA 304P Machine Learning lab (practical list set by your teacher), plus other GGSIPU BCA Sem 6 papers
Paired lab: BCA 304P Machine Learning lab notes. A Skill Enhancement Course, 0-2-1, whose practical list the teacher notifies in the first week.
Until then, the theory programs match unit by unit: Unit 1's six cover NumPy, Pandas and regression; Unit 2's five cover trees, Naïve Bayes, SVM, k-NN and Random Forest; Unit 3's two build a perceptron and a Keras MLP; Unit 4's five cover K-means, dendrograms, SOM, feature selection and PCA.
Related Semester 6 papers: BCA 302T Distributed Systems and Cloud Computing and BCA 306T Software Project Management (core); among electives, BCA 314T Deep Learning with Python builds on Unit 3 and BCA 312T Data Visualization and Analytics reuses Unit 1's Pandas work.
Entropy, information gain and Gini index in machine learning: excerpt from page 24 of the BCA 304T notes
Excerpt from page 24
1.2 Attribute selection measures
A decision tree is built top-down by choosing, at every node, the attribute that separates the classes best. Three measures decide that choice.
- Entropy (impurity): Entropy(S) = −Σ pi log2 pi. It is 0 for a pure node and 1 for a 50 : 50 split between two classes.
- Information Gain (ID3): Gain(S, A) = Entropy(S) − Σ (|Sv| / |S|) × Entropy(Sv).
- Split Information and Gain Ratio (C4.5): SplitInfo(S, A) = −Σ (|Sv| / |S|) log2(|Sv| / |S|), and Gain Ratio = Gain(S, A) / SplitInfo(S, A).
- Gini Index (CART): Gini(S) = 1 − Σ pi2. It is 0 for a pure node and 0.5 at maximum impurity for two classes.
| Algorithm | Full form / inventor | Split measure | Key features |
|---|---|---|---|
| ID3 | Iterative Dichotomiser 3 (Quinlan, 1986) | Information Gain | Only categorical attributes; multi-way splits; biased toward attributes with many values |
| C4.5 | Successor of ID3 (Quinlan, 1993) | Gain Ratio | Handles continuous attributes and missing values; post-pruning |
| CART | Classification And Regression Trees (Breiman, 1984) | Gini Index (MSE for regression) | Strictly binary splits; classification and regression; used by Scikit-Learn |
ID3, C4.5 and CART compared (page 24)
- Algorithm
- ID3
- Full form / inventor
- Iterative Dichotomiser 3 (Quinlan, 1986)
- Split measure
- Information Gain
- Key features
- Only categorical attributes; multi-way splits; biased toward attributes with many values
- Algorithm
- C4.5
- Full form / inventor
- Successor of ID3 (Quinlan, 1993)
- Split measure
- Gain Ratio
- Key features
- Handles continuous attributes and missing values; post-pruning
- Algorithm
- CART
- Full form / inventor
- Classification And Regression Trees (Breiman, 1984)
- Split measure
- Gini Index (MSE for regression)
- Key features
- Strictly binary splits; classification and regression; used by Scikit-Learn
Advantages and disadvantages
Decision trees are easy to understand, visualise and explain, so they are called white-box models. They need little data preparation and no feature scaling, handle both numerical and categorical data, convert directly into IF-THEN rules, and predict fast, in O(log n) for a balanced tree.
On the other side, deep trees memorise noise and overfit, and a small change in the data can change the whole tree. Greedy splitting may not find the globally best tree, information gain is biased toward attributes with many values, and axis-parallel splits fit diagonal boundaries poorly.
Solved numerical: the Play Tennis data
The page then sets up the classic 14-day Play Tennis dataset (9 Yes, 5 No) with four attributes: Outlook (Sunny, Overcast, Rain), Temperature (Hot, Mild, Cool), Humidity (High, Normal) and Wind (Weak, Strong). The task is to find the root node of the tree using information gain. The step-by-step working continues on page 25 of the full notes.
Do the BCA 304T notes cover the full GGSIPU syllabus?
Yes. Every official BCA 304T topic has its own section in the notes; no gaps.
| Official unit | Official topics | Where in the notes | Status |
|---|---|---|---|
| Unit I | Introduction to Machine Learning, Why Machine learning, Types, Applications. Regression and Classification. Binary, Multiclass, Multilabel Classification. Confusion Matrix, Accuracy, Precision & recall, ROC Curve. NumPy, Pandas, Scikit-Learn. Linear Regression (one and multiple variables), Logistic Regression | Unit 1, Sections 1-11, pages 5-19 | Covered |
| Unit II | Decision Trees, Tree pruning, Rule-base Classification, Naïve Bayes, Bayesian Network, Support Vector Machines, k-Nearest Neighbor, Ensemble Learning and Random Forest algorithm | Unit 2, Sections 1-9, pages 23-36 | Covered |
| Unit III | Artificial Neural Networks, HebbNet, Perceptron, Adaline, Multilayer Neural Network, Architecture, Activation Functions, Loss Function, Hyper parameters, Gradient Descent, Backpropagation and Variants, Regularization, Applications | Unit 3, Sections 1-13, pages 40-51 | Covered |
| Unit IV | Clustering, K-means, Hierarchical Clustering, Kohonen Self-Organizing Maps, Implementation of Unsupervised algorithms, Feature selection and Dimensionality reduction, Principal Component Analysis | Unit 4, Sections 1-7, pages 55-65 | Covered |
BCA 304T official syllabus mapped to the notes
- Official unit
- Unit I
- Official topics
- Introduction to Machine Learning, Why Machine learning, Types, Applications. Regression and Classification. Binary, Multiclass, Multilabel Classification. Confusion Matrix, Accuracy, Precision & recall, ROC Curve. NumPy, Pandas, Scikit-Learn. Linear Regression (one and multiple variables), Logistic Regression
- Where in the notes
- Unit 1, Sections 1-11, pages 5-19
- Status
- Covered
- Official unit
- Unit II
- Official topics
- Decision Trees, Tree pruning, Rule-base Classification, Naïve Bayes, Bayesian Network, Support Vector Machines, k-Nearest Neighbor, Ensemble Learning and Random Forest algorithm
- Where in the notes
- Unit 2, Sections 1-9, pages 23-36
- Status
- Covered
- Official unit
- Unit III
- Official topics
- Artificial Neural Networks, HebbNet, Perceptron, Adaline, Multilayer Neural Network, Architecture, Activation Functions, Loss Function, Hyper parameters, Gradient Descent, Backpropagation and Variants, Regularization, Applications
- Where in the notes
- Unit 3, Sections 1-13, pages 40-51
- Status
- Covered
- Official unit
- Unit IV
- Official topics
- Clustering, K-means, Hierarchical Clustering, Kohonen Self-Organizing Maps, Implementation of Unsupervised algorithms, Feature selection and Dimensionality reduction, Principal Component Analysis
- Where in the notes
- Unit 4, Sections 1-7, pages 55-65
- Status
- Covered
Official source: BCA Scheme of Examination and Syllabi, batch 2024-25 onwards (PDF, ipu.ac.in).
Every course in GGSIPU notes is mapped to its official PDF the same way.
BCA 304T Machine Learning notes FAQs
How do I get the full BCA 304T notes?
Message Unnati Educations on WhatsApp at 9899436384 using the "Get Full Notes" button on this page. The full BCA 304T notes are sold as a printed book only, not as a PDF. The book has all 4 units in 67 pages, with 15 solved numericals and 18 Python programs. Our team confirms delivery for your address on WhatsApp.
Which BCA students should use these BCA 304T notes?
These notes follow the BCA scheme for students admitted in 2024-25 onwards, where BCA 304T Machine Learning is a 4-credit core theory paper in Semester 6. They suit students at affiliated colleges under the USICT BCA scheme. If your batch follows an older scheme, compare your syllabus with the coverage table on this page before you order.
What is the passing mark for BCA 304T?
BCA 304T is marked out of 100: 40 internal and 60 end-term. The BCA scheme of examination does not print a pass percentage for this theory paper; its only stated 40% pass mark is for the bridge course. End-term minimums, re-appear rules and result rules are set by GGSIPU Ordinance 11, published on the university website, ipu.ac.in.
Which BCA 304T unit has the most solved numericals?
Units 2, 3 and 4 of the BCA 304T notes have four solved numericals each, while Unit 1 has three. Unit 2 covers information gain, Naïve Bayes, a Bayesian network and k-NN; Unit 3 covers Hebb net, perceptron, Adaline and backpropagation; Unit 4 covers K-means, agglomerative clustering, SOM and PCA. Unit 1 leads on Python programs, with six.
Do the BCA 304T notes include Python code?
Yes. The BCA 304T notes contain 18 Python programs placed beside the theory they implement. They use NumPy, Pandas and Scikit-Learn throughout, plus one Keras model in Unit 3 for regularisation with dropout and early stopping. Many programs carry comments with the expected output, such as the final K-means centroids, so you can check your own run line by line.
How is BCA 304T different from the BCA 304P lab?
BCA 304T is the 4-credit theory paper with a 60-mark written end-term exam. BCA 304P Machine Learning Lab is a 1-credit Skill Enhancement Course with two practical hours a week, also marked out of 100. The scheme says the lab shares the objectives of BCA 304T, and your teacher announces its practical list in the first week of classes.
Get the full BCA 304T Machine Learning Notes on WhatsApp
The full BCA 304T Machine Learning Notes come as a printed book only, not a PDF. These are Machine Learning notes in English, built from short sections, tables, flowcharts and formula boxes.
Ordering is on WhatsApp only; our team confirms dispatch and delivery for your address. [CONFIRM: book dispatch time and delivery area]
Order the BCA 304T notes book on WhatsApp (4 units, 67 pages)
Unnati Educations is an independent education provider based in Vikaspuri, Delhi. We are not affiliated with, authorised by, or endorsed by Guru Gobind Singh Indraprastha University (GGSIPU). 'GGSIPU' and 'IPU' are used only to describe the syllabus our notes follow. For official information, visit ipu.ac.in.
Ready for the full notes?
Discuss coverage and delivery with our team.