Elective III - Machine Learning Reviewer

Introduction to Machine Learning

• Machine learning (ML) focuses on extracting knowledge from data through research fields like statistics, artificial intelligence, and computer science.
• Early ML applications relied on simple rule-based "if" and "else" decisions.
• The main goal is to automate decision-making processes effectively.

Types of Machine Learning

• Supervised Learning

  • Involves user-provided pairs of inputs and desired outputs.
  • Can generate outputs for unseen inputs without human assistance.
  • Example: Spam classification.

• Unsupervised Learning

  • Deals only with input data, lacking predefined output labels.
  • Typically more challenging to understand and evaluate.
  • Examples include text data analysis, customer segmentation, and anomaly detection.

Essential Libraries for Machine Learning

• Scikit-learn: Offers a variety of machine learning algorithms.
• Numpy: Handles multidimensional arrays and basic linear algebra.
• SciPy: Provides advanced mathematical functions, optimization, and statistical operations.
• Matplotlib: A primary library for scientific plotting, supports high-quality visualizations.
• Pandas: Focused on data wrangling and analysis, organized around the DataFrame structure.

Types of Supervised Machine Learning Tasks

• Classification Tasks

  • Aim to predict a specific class label from defined possibilities (e.g., benign vs. malignant tumors).
  • Binary Classification: Involves distinguishing between two classes.
  • Multiclass Classification: Differentiate among multiple classes.

• Regression Tasks

  • Predict continuous numerical outputs, such as income based on various factors.

Generalization, Overfitting, and Underfitting

• Generalization: The ability of a model to make accurate predictions on new, unseen data.
• Overfitting: Occurs when a model is too closely fit to training data, leading to poor performance on new data.
• Underfitting: A model that fails to capture the training data patterns and also lacks generalization ability.

Data Preprocessing Techniques

• Normalization: Rescaling data to a range between 0 and 1.
• Standardization: Centering data around a mean of zero and a standard deviation of one.
• Label Encoding: Converting categorical labels into numerical format for algorithm compatibility.

Business Intelligence, Analytics, and Decision Support

• Organizations are transitioning to computerized support systems to enhance operations and respond to business pressures.
• The business environment is increasingly complex due to factors like globalization and technology.

Organizational Responses

• Organizations should adopt reactive, anticipative, adaptive, and proactive strategies.
• Strategic planning and innovative business models are vital for success.

Managerial Decision Making

• Management involves utilizing resources to achieve organizational goals, emphasizing decision-making.
• The decision-making process typically follows a structured approach:
  • Define the problem or opportunity.
  • Construct a model for analysis.
  • Identify and evaluate possible solutions.
  • Recommend the best solution.

Early Decision Support Framework

• Decisions are categorized by degrees of structuredness:
  • Highly structured, semi-structured, and highly unstructured.
• Types of managerial control include strategic, management, and operational levels.

Decision Support Systems (DSS)

• DSS are interactive systems that combine human intellect with computational capabilities for solving unstructured problems.
• Business Intelligence (BI) has evolved from these decision support concepts, now emphasizing user-friendly access to data and analysis tools.

Definition and Evolution of Business Intelligence

• BI encompasses various architectures, tools, and methodologies aimed at providing decision-makers with easy data access.
• Key goals include transforming data into actionable insights.
• The term BI originated in the mid-1990s, evolving from earlier information systems like MIS and EIS.

Architecture of Business Intelligence

• A typical BI system consists of:
  • Data Warehouse: Source of consolidated data.
  • Business Analytics: Tools for data manipulation and analysis.
  • Business Performance Management (BPM): For monitoring and performance analysis.
  • User Interface: Facilitates interaction, often via dashboards.

Analytics Overview

• Analytics can be categorized into three types:
  • Descriptive Analytics: Analyzes past data.
  • Predictive Analytics: Forecasts future trends based on data.
  • Prescriptive Analytics: Recommends actions based on data analysis.