Course: Machine Learning
This "Machine Learning" course provides a comprehensive journey through the field of machine learning, designed for individuals seeking to enhance their skills in developing data-driven solutions. The course begins with an introduction to machine learning fundamentals, explore different machine learning types, and understand the systematic process of implementing machine learning solutions. Practical applications across various industries are highlighted to contextualize the theoretical knowledge and inspire innovative thinking.
The course progresses into more advanced topics, covering essential techniques in regression and classification in supervised learning, unsupervised learning methodologies, and feature engineering. The curriculum also includes a detailed exploration of model tuning and error analysis, equipping learners with the skills required to the continuous improvement in machine learning models. By the end of the course, participants will possess the expertise necessary to tackle complex machine learning challenges and drive data-centric decisions in their respective fields.
Course Outcomes
Upon completion of this course, participants will:
- Clearly understand machine learning process, and explain the different types of machine learning and their applications.
- Apply data validation, feature engineering, and preprocessing methods to prepare datasets for machine learning models.
- Build and evaluate regression and classification models using Scikit-Learn
- Interpret model parameters, and employ techniques to optimize model performance through regularization and hyperparameter tuning.
- Implement clustering algorithms such as KMeans and DBSCAN for tasks like customer segmentation and anomaly detection, and validate clustering results effectively.
Who It's For
This course is ideal for:
- Aspiring data scientists and machine learning engineers
- Programmers and software developers
- IT professionals
- Business and Data Analysts
- Students and Researchers
- Anyone Interested in AI and Machine Learning
Prerequisites
- Fluent in Python programing language.
- Familiar with data wrangling using Pandas
- Completed "Data Analytics using Python" (recommended)
Course Modules
1. Introduction to machine learning
- Comparison of machine learning and traditional programming.
- Overview of the hierarchy of machine learning types.
- Detailed description of the machine learning process.
- Discussion of various applications of machine learning.
2. Data Preprocessing
- Data validation and type transformation techniques.
- Creating feature matrices, target vectors, and performing train-test splits.
- (Optional) Addressing common data issues such as duplicates, missing values, inconsistencies, and outliers.
- Methods for imputing missing values with scikit-learn.
- Encoding categorical features.
- Techniques for scaling and normalizing features.
- Explanation of data leakage and strategies to avoid it.
- Preprocessing data using Scikit-Learn Pipelines and ColumnTransformers.
3. Supervised machine learning - Regression
- Transitioning from scatter plots to regression models.
- Training linear regression models with Scikit-Learn.
- Interpreting model predictions and parameters.
- Evaluating models using regression metrics.
- Understanding the bias and variance tradeoff and methods to mitigate them.
- Applying regularization techniques.
- Hyperparameter tuning with GridSearchCV.
4. Supervised machine learning - Classification
- Training logistic regression models.
- Evaluating models using classification metrics.
- Training KNN and Random Forest classification models.
- Strategies for dealing with unbalanced data.
- Hyperparameter tuning using GridSearchCV.
- Introduction to train-validation-test split and tuning model thresholds.
5. Unsupervised machine learning
- Training KMeans clustering models.
- Overview of the clustering analysis process and cluster validation.
- Implementing clustering analysis using the DBSCAN algorithm.
- Use case 1: RFM customer segmentation.
- Anomaly detection through clustering analysis.
- Use case 2: Detecting fraudulent insurance claims using anomaly detection.
6. Feature engineering
- Addressing dimensionality reduction and the Hughes phenomenon.
- Dimensionality reduction using feature selection based on correlation, variance, feature importance, and multicollinearity.
- Applying principal component analysis (PCA) for dimensionality reduction.
7. Error Analysis Process
- Inspecting and categorizing model errors.
- Prioritizing error groups and developing action items for improvement.
- Invalidating the test set and introducing the concept of a gold set.
Execution approach:
Long run
40 hours
7 weeks
Run over short weekly sessions allowing participants to progressively build their skills and knowledge.
