Fundamentals of Machine Learning (4341603) - Summer 2023 Solution

Question 1(a) [3 marks]

Define human learning. List out types of human learning.

Answer:

Human learning is the process by which humans acquire new knowledge, skills, behaviors, or modify existing ones through experience, study, or instruction.

Types of Human Learning:

Mnemonic: "SUR - Supervised, Unsupervised, Reinforcement"

Question 1(b) [4 marks]

Differentiate between qualitative data and quantitative data.

Answer:

Table: Qualitative vs Quantitative Data

Mnemonic: "QUAN-Numbers, QUAL-Words"

Question 1(c) [7 marks]

Compare the different types of machine learning.

Answer:

Table: Types of Machine Learning Comparison

Key Differences:

• Supervised: Uses input-output pairs for training
• Unsupervised: Discovers hidden patterns in data
• Reinforcement: Learns through interaction with environment

Mnemonic: "SUR-LAP: Supervised-Labeled, Unsupervised-Reveal, Reinforcement-Action"

Question 1(c OR) [7 marks]

Define machine learning. Explain any four applications of machine learning in brief.

Answer:

Machine learning is a subset of artificial intelligence that enables computers to learn and make decisions from data without being explicitly programmed.

Four Applications:

Mnemonic: "SIRM - Spam, Image, Recommendation, Medical"

Question 2(a) [3 marks]

Relate the appropriate data type of following examples.

Answer:

Data Type Classification:

Mnemonic: "NCN - Nominal, Categorical, Numerical"

Question 2(b) [4 marks]

Explain data pre-processing in brief.

Answer:

Data pre-processing is the technique of preparing raw data for machine learning algorithms.

Key Steps:

Mnemonic: "CITR - Clean, Integrate, Transform, Reduce"

Question 2(c) [7 marks]

Show K-fold cross validation in detail.

Answer:

K-fold cross validation is a technique to evaluate model performance by dividing data into K equal parts.

Process:

Steps:

• Divide: Split dataset into K equal parts
• Train: Use K-1 folds for training
• Test: Use remaining fold for validation
• Repeat: Perform K iterations
• Average: Calculate mean performance

Advantages:

• Reduces overfitting
• Better use of limited data
• More reliable performance estimate

Mnemonic: "DTRA - Divide, Train, Repeat, Average"

Question 2(a OR) [3 marks]

Define following terms: i) Mean, ii) Outliers, iii) Interquartile range

Answer:

Statistical Terms:

Mnemonic: "MOI - Mean, Outliers, Interquartile"

Question 2(b OR) [4 marks]

Explain structure of confusion matrix.

Answer:

Confusion Matrix Structure:

Components:

• TP: Correctly predicted positive cases
• TN: Correctly predicted negative cases
• FP: Incorrectly predicted as positive
• FN: Incorrectly predicted as negative

Mnemonic: "TTFF - True True, False False"

Question 2(c OR) [7 marks]

Prepare short note on feature subset selection.

Answer:

Feature subset selection is the process of selecting relevant features from the original feature set.

Methods:

Benefits:

• Reduced complexity: Fewer features, simpler models
• Improved performance: Eliminates noise and irrelevant features
• Faster training: Less computational overhead

Popular Techniques:

• Chi-square test
• Recursive Feature Elimination
• LASSO regularization

Mnemonic: "FWE - Filter, Wrapper, Embedded"

Question 3(a) [3 marks]

Give the difference between predictive model and descriptive model.

Answer:

Model Type Comparison:

Mnemonic: "PF-DC: Predictive-Future, Descriptive-Current"

Question 3(b) [4 marks]

Discuss the difference between classification and regression.

Answer:

Classification vs Regression:

Mnemonic: "CCNM - Classification-Categories, Regression-Numbers"

Question 3(c) [7 marks]

Define classification. Illustrate classification learning steps in details.

Answer:

Classification is a supervised learning technique that predicts discrete class labels for input data.

Classification Learning Steps:

Detailed Steps:

• Data Collection: Gather labeled training data
• Preprocessing: Clean and prepare data
• Feature Selection: Choose relevant attributes
• Split Data: Divide into training and testing sets
• Training: Build model using training data
• Evaluation: Test model performance
• Deployment: Use model for predictions

Mnemonic: "DCFSTED - Data, Clean, Features, Split, Train, Evaluate, Deploy"

Question 3(a OR) [3 marks]

Give the difference between bagging and boosting.

Answer:

Bagging vs Boosting:

Mnemonic: "BPV-BSB: Bagging-Parallel-Variance, Boosting-Sequential-Bias"

Question 3(b OR) [4 marks]

Explain different types of logistic regression in brief.

Answer:

Types of Logistic Regression:

Mnemonic: "BMO - Binary, Multinomial, Ordinal"

Question 3(c OR) [7 marks]

Write and show the use of k-NN algorithms.

Answer:

K-Nearest Neighbors (k-NN) is a lazy learning algorithm that classifies data points based on the majority class of k nearest neighbors.

Algorithm:

goat

Distance Calculation:

• Euclidean Distance: √[(x₁-x₂)² + (y₁-y₂)²]

Applications:

• Recommendation systems: Similar user preferences
• Image recognition: Pattern matching
• Medical diagnosis: Symptom similarity

Advantages:

• Simple to implement
• No training required
• Works well with small datasets

Mnemonic: "CDSA - Choose, Distance, Select, Assign"

Question 4(a) [3 marks]

List out applications of support vector machine.

Answer:

SVM Applications:

Mnemonic: "TIB - Text, Image, Bio"

Question 4(b) [4 marks]

Create pseudo code for k-means algorithm.

Answer:

K-means Pseudo Code:

BEGIN K-means
1. Initialize k cluster centroids randomly
2. REPEAT
   a. Assign each point to nearest centroid
   b. Update centroids as mean of assigned points
   c. Calculate total within-cluster sum of squares
3. UNTIL convergence or max iterations
4. RETURN final clusters and centroids
END

Mnemonic: "IAUC - Initialize, Assign, Update, Check"

Question 4(c) [7 marks]

Write and explain applications of unsupervised learning.

Answer:

Unsupervised learning discovers hidden patterns in data without labeled examples.

Major Applications:

Clustering Applications:

• Market research: Customer grouping
• Social network analysis: Community detection
• Gene sequencing: Biological classification

Dimensionality Reduction:

• Visualization: High-dimensional data plotting
• Feature extraction: Noise reduction

Mnemonic: "CADA - Customer, Anomaly, Data, Association"

Question 4(a OR) [3 marks]

List out applications of regression.

Answer:

Regression Applications:

Mnemonic: "SSM - Stock, Sales, Medical"

Question 4(b OR) [4 marks]

Define following terms: i) Support ii) Confidence

Answer:

Association Rule Terms:

Example:

• If 30% transactions contain bread and milk: Support = 0.3
• If 80% of bread buyers also buy milk: Confidence = 0.8

Mnemonic: "SF-CP: Support-Frequency, Confidence-Probability"

Question 4(c OR) [7 marks]

Explain apriori algorithm in detail.

Answer:

Apriori algorithm finds frequent itemsets in transactional data using the apriori property.

Algorithm Steps:

Apriori Property:

• If an itemset is frequent, all its subsets are frequent
• If an itemset is infrequent, all its supersets are infrequent

Steps:

1. Scan database: Count 1-item support
2. Generate candidates: Create k+1 itemsets from frequent k-itemsets
3. Prune: Remove candidates with infrequent subsets
4. Count support: Scan database for candidate frequencies
5. Repeat: Until no new frequent itemsets found

Applications:

• Market basket analysis
• Web usage patterns
• Protein sequences

Mnemonic: "SGPCR - Scan, Generate, Prune, Count, Repeat"

Question 5(a) [3 marks]

List out the major features of matplotlib.

Answer:

Matplotlib Features:

Mnemonic: "MCE - Multiple, Customization, Export"

Question 5(b) [4 marks]

How to load iris dataset in Numpy program? Explain.

Answer:

Loading Iris Dataset in NumPy:

Python

Steps:

• Import: Import required libraries
• Load: Use sklearn's load_iris() function
• Extract: Get features and target arrays
• Access: Use .data and .target attributes

Mnemonic: "ILEA - Import, Load, Extract, Access"

Question 5(c) [7 marks]

Explain features and applications of Pandas.

Answer:

Pandas is a powerful data manipulation and analysis library for Python.

Key Features:

Applications:

Common Operations:

• Reading data: pd.read_csv(), pd.read_excel()
• Filtering: df[df['column'] > value]
• Grouping: df.groupby('column').mean()

Mnemonic: "DSDCG - DataFrame, Series, Data I/O, Cleaning, Grouping"

Question 5(a OR) [3 marks]

List out the applications of matplotlib.

Answer:

Matplotlib Applications:

Mnemonic: "SBE - Scientific, Business, Educational"

Question 5(b OR) [4 marks]

Develop and explain the steps to import csv file in Pandas.

Answer:

Steps to Import CSV in Pandas:

Python

Process:

• Import: Import pandas library
• Read: Use pd.read_csv() function
• Specify: Add file path and parameters
• Store: Assign to DataFrame variable

Mnemonic: "IRSS - Import, Read, Specify, Store"

Question 5(c OR) [7 marks]

Explain features and applications of Scikit-Learn.

Answer:

Scikit-Learn is a comprehensive machine learning library for Python.

Key Features:

Applications:

Algorithm Categories:

• Supervised: SVM, Random Forest, Linear Regression
• Unsupervised: K-means, DBSCAN, PCA
• Ensemble: Bagging, Boosting

Workflow:

1. Data preparation: Preprocessing
2. Model selection: Choose algorithm
3. Training: Fit model to data
4. Evaluation: Assess performance
5. Prediction: Make forecasts

Mnemonic: "APME - Algorithms, Preprocessing, Metrics, Evaluation"