ML Summary

Supervised Learning

Definition: A machine learning paradigm where the model is trained on labeled data (input-output pairs).

Applications:

• Classification: Predicting categorical outcomes (e.g., customer churn prediction, fraud detection).
• Regression: Predicting continuous outcomes (e.g., sales forecasting, pricing models).

Key Algorithms:

• Linear regression, logistic regression
• decision trees, support vector machines (SVM)
• neural networks. (Neural networks used both in supervised and unsupervised learning)

Unsupervised Learning

Definition: A paradigm where the model identifies patterns in unlabeled data.

Applications:

• Clustering: Grouping similar data points (e.g., customer segmentation, market research).
• Dimensionality Reduction: Simplifying data while retaining essential features (e.g., Principal Component Analysis, t-SNE).

Key Algorithms:

• K-means clustering
• hierarchical clustering,
• DBSCAN.

Semi-Supervised Learning

Definition: Combines a small amount of labeled data with a large amount of unlabeled data.

Applications: Useful when labeling data is expensive or time-consuming (e.g., text classification in NLP)

Reinforcement Learning

Definition: A learning paradigm where an agent learns by interacting with an environment to maximize rewards.

Applications: Optimizing supply chains, dynamic pricing, and recommendation systems.

Statistical Analysis and Methods

• Exploratory Data Analysis (EDA):
  • Using statistical techniques to understand the structure of data (e.g., summary statistics, correlation analysis).
• Hypothesis Testing: Validating assumptions or relationships in data using tests like t-tests, chi-square tests, and ANOVA.
• Time Series Analysis: Analyzing sequential data points for trends, seasonality, and forecasting (e.g., ARIMA models).

Basic Machine Learning Approaches

• Ensemble Methods: Combining multiple models to improve performance:
  • Random Forests,
  • Gradient Boosting Machines like XGBoost or LightGBM.
• Feature Engineering: Enhancing model input through domain knowledge (e.g., creating ratios, aggregating data).
• Model Evaluation: Validating model performance using metrics like accuracy, precision, recall, F1 score, or ROC-AUC.

Natural Language Processing (NLP)

Applications:

• Text analytics for sentiment analysis
• topic modeling (Identify underlying themes or topics in a large corpus of text - Unsupervised learning) useful in indexing, semantic analysis and visualization.
• Information extraction from unstructured data.

Deep Learning

Applications: Leveraging neural networks for tasks like:

• image recognition, speech analysis and complex pattern identification.

Visualization and Communication

For visual insights tools like:

• Tableau, Power BI, and Python libraries (e.g., Matplotlib, Seaborn) are used
• Storytelling with data is crucial for translating analytical findings into actionable business strategies.

Business Intelligence Integration

Combining predictive analytics, dashboards, and automated reporting systems to facilitate data-driven decision-making.

NumPy

• Scope: Fundamental library for numerical computations.
• Key Goals: Efficient handling of large arrays and matrices.
• Key Features: Mathematical functions, linear algebra, random number generation.

Pandas

• Scope: Data manipulation and analysis.
• Key Goals: Handling tabular data with DataFrames, cleaning and preprocessing data.
• Key Features: Data filtering, grouping, merging, reshaping.

Matplotlib

• Scope: Low-level static visualizations.
• Key Goals: Create custom plots, charts, and visualizations.
• Key Algorithms: Plotting primitives, customization of visual elements.

Seaborn

• Scope: Statistical data visualization built on Matplotlib.
• Key Goals: Simplify the creation of attractive and informative graphics.
• Key Features: Heatmaps, pair plots, violin plots, categorical plots.

Plotly

• Scope: Interactive visualizations for the web.
• Key Goals: Interactive dashboards and 3D plots.
• Key Features: Zooming, hover labels, interactive sliders.

SciPy

• Scope: Advanced statistical computations and optimization.
• Key Goals: Hypothesis testing, curve fitting, integration, and signal processing.
• Key Features: Statistical distributions, t-tests, ANOVA, optimization solvers.

Statsmodels

• Scope: Statistical modeling and testing.
• Key Goals: Perform linear and generalized linear regression, time-series analysis.
• Key Features: Regression models, ARIMA, hypothesis testing.

Scikit-learn

• Scope: Comprehensive toolkit for classical machine learning.
• Key Goals: Train and evaluate machine learning models.
• Key Algorithms:
  • Classification: SVMs, Random Forests, kNN, Logistic Regression.
  • Regression: Linear Regression, Ridge, Lasso.
  • Clustering: K-Means, DBSCAN, Hierarchical Clustering.

XGBoost, LightGBM, CatBoost

• Scope: Gradient boosting frameworks for structured data.
• Key Goals: High-performance ensemble models for tabular data.

Key Features:

• XGBoost: Handles sparsity, regularization support.
• LightGBM: Faster training, optimized for large datasets.
• CatBoost: Native handling of categorical variables.

TensorFlow

• Scope: Full-stack deep learning framework.
• Key Goals: Develop scalable and production-ready models.
• Key Algorithms: Neural networks (CNNs, RNNs, Transformers), reinforcement learning.

PyTorch

• Scope: Flexible framework for research-oriented deep learning.
• Key Goals: Enable dynamic graph computation and experimentation.
• Key Algorithms: Similar to TensorFlow, with more flexibility in prototyping.

NLTK (NLP Toolkit Library)

• Scope: Foundational NLP tasks.
• Key Goals: Text tokenization, stemming, and parsing.
• Key Features: POS tagging, N-gram models.

Gensim: NLP Library

• Scope: Topic modeling and vector representation.
• Key Goals: Perform LDA, Word2Vec, and document similarity analysis.

spaCy: NLP Library

• Scope: Industrial-strength NLP pipelines.
• Key Goals: Fast and accurate linguistic analysis.
• Key Features: Named Entity Recognition (NER), Dependency Parsing.
• Complexity: Moderate to High.

Hugging Face Transformers Library

• Scope: State-of-the-art transformer-based NLP.
• Key Goals: Enable tasks like text classification, summarization, and translation.
• Key Algorithms: BERT, GPT, T5, etc.
• Complexity: High.