Aim
Learn how to build machine learning models in R (using tidymodels), interpret them, and how to 'improve' model evaluation with cross-validation.
Content
The focus will be on building and evaluating machine learning models in R rather than an in-depth breakdown of specific algorithms. We will be building models to distinguish between different categories of text based on linguistic features (including number of nouns, adjectives, etc.) using XGBoost.
- Exploratory data analysis
- Binary classification
- Feature importance
- Multiclass classification
- Cross-validation
- *Extra (if enough time)*
- Hyperparameter tuning
- PCA
- Cluster analysis
Target audience
This is a course for UiO-affiliated students or researchers those that want to learn more about machine learning, how it can be used in research, but do not have a strong background in mathematics or data science. This is a hands-on course and it is an advantage but not necessary that you are accustomed to writing code in R. Basic knowledge of descriptive statistics and tidyverse is a plus.
A video (approximately 25 minutes) has been prepared that might be useful for those that are completely new to machine learning, with example use-cases in research.
Duration
2 x 3 hours
Signing up
Fill out the signup form if you want to attend the course. The course is full. If you want to be put on the waiting list or to be updated about the next time the course will be help, sign this form.
Important: Participants must use their own PC or Mac (laptop) with both R and RStudio installed. Both R (¡Ý 3.3.0) and RStudio are free and do not require a licence. R can be installed from https://cran.r-project.org and RStudio from https://www.rstudio.com/products/rstudio/download/.
Contact IT-support from your faculty or department if you need help with installation. You can use UiO Programkiosk ("Statistikk fullskjerm") if it is not possible to install either R or RStudio on your own computer.
Install the following packages in R(studio) before the start of the course:
tidyverse, tidymodels, xgboost, vip, patchwork,
*extra packages* doParallel, discrim, factoextra
A second screen/monitor is an advantage (i.e. one for zoom, the other for coding)
Number of participants
30
Language
The course will be held in english
Instructor
Luigi Maglanoc PhD
Contact information
If you have any questions about the course, send us an email: statistikk@usit.uio.no
Links to course material
- Dataset
- R-code (to be added)