PRACTICAL MACHINE LEARNING
Once the purview of programmers and statisticians, machine learning has expanded across applications and disciplines into virtually every industry. And to keep pace with an increasingly data-driven marketplace, professionals of all stripes have set their sights on mastering its fundamentals.
UC Berkeley Extension’s Practical Machine Learning course offers a hands-on introduction to machine learning with R-programming that includes real-world datasets that let you solve problems in a variety of industries. Whether it’s business leaders aiming to improve their understanding of data science and machine learning to coordinate better with teams on tactical data-based initiatives, or aspiring data scientists seeking to master the practical aspects of problem framing and model deployment, the course addresses topics, tools, and techniques to help professionals from any background or industry enhance their machine learning skills.
- 111Video Lectures
- 6Application Assignments
- 26R Illustration Videos
- 5Exercises / Quizzes
- 6Live Webinars
- Machine Learning – The What, The Why, and The How?
- Machine Learning – Models and Functions
- Model Training
- Model Scoring
- Classification of Machine Learning Algorithms
- Three Features of Supervised ML Algorithms
- Summary and Key Takeaways
- Overview of Machine Learning Process
- Business Problem Framing
- Define the Final Delivery
- Frame the Problem: App Store Example
- Collect and Shape Data: Reading Data in R
- Collect and Shape Data: Shaping Data in R
- Exploratory Data Analysis
- EDA in R
- Fitting a simple model
- Fitting a simple model in R
- Linear Model Training in R
- Limitations of a Linear Model
- Linear Model Training in R (One Variable)
- Linear Model Training in R (Multiple Variable)
- Stepwise Regression
- Stepwise Regression in R
- Summary and Key Takeaways
- Introduction to Classification (Categorical Modeling)
- Building from Linear Regression
- Loss Function
- Unconstrained Errors
- Logistic Function
- Use of the logistic
- Probability, Odds, and Log-odds
- Training a Logistic Regression Mode
- Stepwise Regression
- Variable Importance
- Multinomial Logistic Regression
- Summary and key Takeaways
- Introduction to Model Evaluation
- Class Separation Plot
- Confusion Matrix
- Evaluating Models Using Confusion Matrix
- Nomenclature of Binomial Metrics
- ROC Curves
- Default ROC Curve
- Generalizing ROC Curve
- Value matrix and curve
- Class Imbalance
- Cohen’s Kappa
- Summary and Key Takeaways
- Introduction to Decision Trees
- Recursive Partitioning
- Partitioning Scoring
- Splitting Continuous Variables
- Splitting Categorical Variables
- How to Evaluate Best Split?
- Measures if Homogeneity
- Evaluating Splits Using Model Performance
- Finding the Best Split
- Stopping Criteria
- Rules for Partitioning
- Handling Missing Data
- Tree Method: Advantages and Disadvantages
- RP Classification & Regression Examples – coding example
- Intro to Resampling
- Variance-bias Tradeoff
- Training and Testing Model Performance
- Resampling Methods
- Resampling Process and Practice
- Tuning Parameter Optimization
- Caret Package in R
- Model Validation
- Tidyverse Coding Example
- Introduction to Model Improvements
- Model Ensembles
- Bagging Models
- Random Forests
- Using Random Forest – Coding Example
- Generalized Ensembles and Model Stacking
- Simple Boosting
- Comparison with CART
- Gradient Boosting
- Weak Learners
- Local Minimums vs Global Minimums
- Stochastic Gradient Boosting Machines
- Stochastic GBMs – Coding Example
- Introduction to Neural Networks
- Analogy to Brain Function
- From Neurons to Neural Networks
- Single-Layer Feed Forward Networks
- Solutions to Overfitting
- MNIST Handwriting Recognition
- MNIST Coding Example
- Back Propagation
- Deep Learning
- Introduction to Unsupervised Learning
- Principal Components Analysis (PCA)
- PCA Coding Example
- K-means Coding Example
- Hierarchical Clustering (HC)
- HC Coding Example
- Clustering Code Example
- Association Rules
- Association Rules Coding Example
- Semi-Supervised Learning
- Introduction to Model Deployment
- Model Development: Roles and Success Factors
- Data Science Project Life Cycle
- Creating a Problem Statement
- Deployment Assets
- Deployment Patterns
- Agile Process for Deployment
- Managing Environments and Assets
- Introduction to Problem Framing
- Recommender Systems 1: Problem Framing
- Recommender Systems 2: Debrief and Approaches
- Recommender Systems 3: Walkthrough with Coding hints
- Customer Lifetime Value 1: Problem Framing
- Customer Lifetime Value 2: Debrief and Approaches
- Customer Lifetime Value 3: Walkthrough with Coding hints
- Deployment and Performance Considerations
Price Predictions Using App Store Data
Participants will perform an exploratory data analysis (EDA) and build a univariate or multivariate linear regression model using data from Apple’s app store.
Predicting Probability of Credit Default
Participants will apply logistic regression to a dataset including features on credit card users and develop a model predicting the probability of default payments based upon previous payment history, bill amount, and customer demographics.
Predicting Probability of Employee Attrition
Participants will examine classification problems and apply what they have learned to an employee attrition data set in order to make predictions about the probability of an employee leaving his/her company.
Neural Nets: Determining Standard vs. Premium Service Levels
Participants will get an introduction to neural networks and make predictions based upon a dataset with information on office supply purchases.
Customer Segmentation with Clustering
Participants will get an introduction to unsupervised learning algorithmic techniques, such as K-means and hierarchical clustering. and employ clustering techniques to develop segments from customer data.
Problem Framing: Recommendation Engines and Customer Lifetime Value
Participants will walk through practical examples of problem framing and identify approaches to modeling customer lifetime value and develop recommendation engines.
Adjunct Professor in UC Berkeley’s Department of Computer Science and a Founding Partner of Decision Patterns
Course Instructor, EMERITUS
EMERITUS follows a unique online model. This model has ensured that nearly 90 percent of our learners complete their course.
- Orientation Week
The first week is orientation week. During this week you will be introduced to the other participants in the class from across the world. You will also learn how to use the learning platform and other learning tools provided.
- Weekly Goals
On other weeks, you have learning goals set for the week. The goals would include watching the video lectures and completing the assignments. All assignments have weekly deadlines.
- Recorded Video Lectures
The recorded video lectures are by faculty from the collaborating university.
- Live Webinars
Every few weeks, there are live webinars conducted by EMERITUS course leaders. Course leaders are highly-experienced industry practitioners who contextualize the video lectures and assist with questions you may have regarding your assignments. Live webinars are usually conducted between 1 pm and 3 pm UTC on Tuesdays and Wednesdays.
- Clarifying Doubts
In addition to the live webinars, for some courses, the course leaders conduct Office Hours, which are webinar sessions that are open to all learners. During Office Hours, learners ask questions and course leaders respond. These are usually conducted every alternate week to help participants clarify their doubts pertaining to the content.
The EMERITUS Program Support team members will follow up and assist over email and via phone calls with learners who are unable to submit their assignments on time.
- Continued Course Access
You will continue to have access to the course videos and learning material for up to 12 months from the course start date.
EMERITUS Program Support Team
If at any point in the course you need tech, content or academic support, you can email program support and you will typically receive a response within 24 working hours or less.
You can access EMERITUS courses on tablets, phones and laptops. You will require a high-speed internet connection.
On completing the course you join a global community of 5000+ learners on the EMERITUS Network. The Network allows you to connect with EMERITUS past participants across the world.
ADMISSION & FEES
DURATION AND COURSE FEE
- Starts 30 April 2019
- 3 Months
- 6-8 hours per week
- Course Fees USD 1400
- The course requires an undergraduate knowledge of statistics (descriptive statistics, regression, sampling distributions, hypothesis testing, interval estimation etc.) and probability. Familiarity with R (importing a data set, assigning variables, working with a variety of data structures like. numeric, character, factor etc., creating and adding columns to data frames) is required.
- Assignments /application projects which require programming will be done using the R programming language.
- This course is designed for working professionals and requires proficiency in English. All videos are recorded in English. All assignments are written in English and are required to be responded to in English.
- The course requires you to have a laptop and 1 Mbps (or more) Internet connection. The laptop should support one of the following browsers: Chrome 71, Firefox 64, IE 11, Edge 42, Safari 11.
- You can pay for the course either with an international debit or credit card (unfortunately we are unable to accept Diners credit cards), or through a bank wire transfer. On clicking the apply now button below, you will be directed to the application form and the payment page.
- We provide deferrals and refunds in specific cases. The deferrals and refund policy is available here.
- You will be provided a course login within 48 hours of making a payment.
- Please provide your work experience and your current employer via the application.
- You can apply by clicking the Apply Now button