Machine Learning Unit III Quiz

Data Science Course Information

• Course: Data Science for engineers
• Course Credit: 3 (Theory-2hr, Lab-2hr)
• Course Instructor: Dr. Ankita Agarwal
• Level: T. Y. (B.Tech. Bio Engineering)
• University: MIT World Peace University, Pune

Unit III: Machine Learning

• Introduction to Machine Learning: Supervised and Unsupervised Learning
• Splitting datasets: Training and Testing
• Regression: Simple Linear Regression
• Classification: Naïve Bayes classifier
• Clustering: K-means
• Evaluating model performance, Python libraries for ML

Introduction to Machine Learning

• Artificial Intelligence (AI): Computer acts/thinks like a human
• Data Science: Al subset dealing with data methods, scientific analysis, and statistics to gain insight from data
• Machine Learning (ML): Al subset that teaches computers to learn from provided data:
  • "Machine Learning allows the machines to learn and make predictions based on its experience(data)."
  • Machine Learning (by Tom Mitchell, 1998): the study of algorithms that improve their performance at a given task with experience. This is represented as <P, T, E>, where P is performance, T is task, and E is experience.
  • Example learning tasks. Examples are given for handwritten word recognition and spam filters.

Machine Learning Applications

• Recognizing patterns: handwritten/spoken words, medical images
• Generating patterns: generating images or motion sequences
• Recognizing anomalies: unusual credit card transactions, unusual sensor patterns in a nuclear plant
• Prediction: future stock prices, currency rates, personalized medicine (individual genetic profiles for medicine prediction)

Machine Learning Types

• Supervised Learning: Learning with labeled data.
  • The machine learns from a labeled dataset to learn a relationship and predict output values for new data.
  • Types:
    • Regression (predicting real-valued outputs)
    • Classification (predicting categorical outputs)
      • Logistic regression
      • Binary classification
      • Multi-class classification
      • Naïve Bayes classifiers
      • k-NN (k-nearest neighbors) classifiers
      • Decision trees (Random Forest, Gradient Boosting, AdaBoost)
      • Support Vector Machine (SVM)
• Unsupervised Learning: Learning with unlabeled data.
  • The machine explores the data to discover patterns and relationships between data without any labeled knowledge.
  • Types:
    • Clustering (grouping similar data points together)
      • Exclusive Clustering (each item is part of only one subset)
      • Overlapping Clustering (items can be part of one or more subsets)
      • Agglomerative Clustering (set of nested clusters)
      • Probabilistic Clustering (model based on probability distribution function)
      • K-means clustering (partitioning-based method)
      • Hierarchical clustering (agglomerative clustering)
      • Principal Component Analysis (PCA)
      • Singular Value Decomposition (SVD)
  • Advantages: Dimensionality reduction, finding previously unknown patterns, flexibility (wide applicability to problems, such as anomaly detection and association rule mining), cost-effectiveness (does not require labeled data).
  • Disadvantages: Difficult to measure accuracy, may produce less accurate results, lacks guidance and feedback, sensitive to data quality (missing values, outliers), and scalability issues for large, complex datasets.

Regression

• Given data points (x1, y1), (x2, y2), ..., (xn, yn)
• Learn a function f(x) to predict y from x
• y is real-valued

Classification

• Given data points (x1, y1), (x2, y2), ..., (xn, yn)
• Learn a function f(x) to predict y from x
• y is categorical

Model Performance

• Classifier Accuracy: Percentage of correctly classified test set tuples
• Error Rate: 1 - Accuracy
• Confusion Matrix: A table used in model evaluation to show performance
  • TP (True Positives)
  • TN (True Negatives)
  • FP (False Positives)
  • FN (False Negatives)
  • Metrics: Precision, Recall, F1-score, Mathew's correlation coefficient (MCC), Specificity, Sensitivity (Recall)

Workflow/Pipeline

• Data collection and preparation
• Choosing algorithm
• Model training
• Model evaluation and testing
• Candidate models
• Chosen trained model
• Tested model
• Model deployment
• Monitoring

Python Libraries 

• Scikit-Learn: Provides machine learning algorithms (classification, regression, clustering). Built on NumPy, SciPy and matplotlib.
  • Naïve Bayes Classifier
  • Linear Regression
  • K-means clustering

Steps in Machine Learning

• Understand the problem and goals
• Gather prior knowledge and data
• Data integration, selection, and cleaning
• Split data into training and testing sets
• Train models
• Interpret results
• Consolidate and deploy discovered knowledge