Topic 1
Topic 2
Topic 3
Topic 4
You need to have at least some basic programming skills in any programming language.
I’ve never programmed before and I really want to take this course, what should I do?You can take the Introduction to Python Programming short course which prepares you for this one.
I am an experienced software developer but I have not worked with Python before, can I still attend this course?You should be fine to attend this course but it would be helpful if you looked over some of the Python fundamentals just before the course begins.
Will I be able to regard myself as a software engineer after this course?No. This course does not teach you how to write application software, but will instead develop your skills in scripting.
I know how to use Python 2 but not Python 3. Does this matter?Not at all. There are some minor differences which you will pick up quickly as needed during the course.
Which development environment will we use so that I can set this up on my personal machine?Anaconda3-2018-12 version.
Will I have to learn how to write large programs in this course?No. This course focuses on writing small snippets of code which provide you with immediate feedback.
Will this course cover topics like object-oriented programming?You will be using objects all thought the course, but we will not be covering OO principles, nor will we be learning to write Python classes in this course. The focus is on leveraging parts of Python that are most relevant to data science and not end-product development.
Day 1
| Topic | Topic Description |
| Machine Learning Toolkit | A data analyst is only as good as his tools. We introduce you first to Jupyter Notebooks and Python’s Pandas library, which are arguably the best set of tools available for machine learning specialists. |
| Theory of Machine Learning | Machine learning is exciting because it can transform a small amount of input knowledge into a large amount of output knowledge. It isn’t magic, but it can give us a lot for very little. Here we cover what is the actual goal of machine learning and what are some of its pitfalls. |
| First Classification Algorithm | There are machine learning algorithms which mimic intuitive decision-making processes that humans rely on. kNN algorithm is one of those and is a good starting point. We use the challenge of fraud detection as an illustrative example together with Python’s flagship machine learning library: sci-kit learn. |
Day 2
| Topic | Topic Description |
| Assessing Results | Generating machine learning models is typically not difficult. But how good are they? Are they really doing their job? Evaluating the performance of models is an indispensable part of the machine learning and there numerous perspectives to achieve this which will be covered here. |
| Training Classifiers | Robust training strategies are needed to more reliably estimate how well our models will perform once deployed. We cover strategies for creating training, validation and test data sets, as well as designing experiments using cross-fold validation. |
| Refining Classifiers | Machine learning algorithms always find patterns. However, not all patterns are meaningful and some are just hallucinations. We explore methods to reduce this problematic phenomenon called overfitting through techniques like feature selection using loan default risk data as an illustration. |
Day 3
| Topic | Topic Description |
| Expanding the Machine Learning Toolkit | No single machine learning algorithm will always outperform the others. We introduce you to Naïve Bayes, which is often used as a benchmark when selecting and refining solutions. Most algorithms produce models which are not interpretable and being able to make sense of models is becoming increasingly important. Decision Trees possess this quality and serve as a valuable tool on many domains. |
| Wisdom of Crowds | It has been said that all models are wrong, but some are useful. Sometimes putting together all the ‘wrong’ models creates a solution that is more ‘right’. The family of ensemble-based machine learning methods attempts to achieve exactly this. We cover one of the most popular and successful algorithms in this category called the Random Forest classifiers and apply it to data from the financial sector. |
| Category | Full Fee | After SF Funding |
| Singapore Citizen (Below 40) / Singapore PR | $770.40 | |
| Singapore Citizen (40 & above) | $290.40 | |
| Non-Singaporeans | $2,568.00 | Not Eligible |
| Venue | SIT@Dover, 10 Dover Drive S138683 |
| Time | 09:00 AM to 06:00 PM |
| Date |
12 Apr 2021 (Mon) to 14 Apr 2021 (Wed) |
