Repository for Statistical Learning Methods - Summer semester 2020/21

Required software

During the course we'll use RStudio and Jupyter Notebook. To run the code provided during classes you'll need:

• R - https://cran.r-project.org/mirrors.html
• Python - https://www.python.org/downloads/ (for Windows user it's easy to install Python through Anaconda - https://anaconda.org/)
• RStudio - https://rstudio.com/products/rstudio/download/
• Jupyter Notebook or Jupyter Lab - https://jupyter.org/install

Additionally you can install:

• Git - https://git-scm.com/
• PowerShell (for Windows) - https://github.com/PowerShell/PowerShell

Modeling problem report

Use the following dataset hosted on UCI repository: https://archive.ics.uci.edu/ml/datasets/Bank+Marketing (use bank-additional-full.csv file). It contains information related with direct marketing campaigns. All files and description are available under the given URL.

The task is to produce a report with following structure:

5pts 1) Introduction, problem/task description (you should choose at least one supervised learning task, provide description of target variable, rationale behind picking particular task and target), features explanation, dataset description

10pts 2) Cleaning and preprocessing data - removing or imputing missing data, standarization, one-hot encoding, handling outliers, feature extraction, merging categories etc. (some of these points may go to part 3 if EDA/exploration was the cause of particular action)

10pts 3) EDA (both graphical e.g. barplots, scatterplots and descriptive e.g. tables with statistics) with focus on exploring relations between features and target variable; include correlation matrix/heatmap between numerical variables; use one of the clustering (e.g. k-means, hierarchical clustering) or dimensionality reduction technique (e.g. PCA, t-SNE, UMAP) to derive additional insight from data

10pts 4) Create models (at least 3 distinct types of models) and tune hyperparameters for each model

10pts 5) Graphical (e.g. barplots of errors, ROC curves, residual graphs, etc.) and descriptive model assessment (comparing measures picked as performance criteria); remember to choose appropriate measure for the task - RMSE is measure for regression, not classification

5pts 6) Summary, short discussion on encountered problems, actions undertaken to solve them, possible shortcomings of used model/approach/data

Code and descriptions/comments should be in Jupyter or R Markdown notebook. Please send reports to [email protected] or [email protected] with following naming convention <index_number>_SLM_S2021_Report.< extension > before 08.06.2021r. EOD. You can use Python or R language for the report.

Each email should have 2 attachments:

• for Jupyter: .ipynb file and .html/.pdf file generated from notebook,
• for R Markdown: .Rmd file and .html/.pdf file generated from the script.

Contact

Name: Łukasz Kraiński

Email: [email protected] or [email protected]

You can contact me through MS Teams and schedule consultation if needed. You can create an issue in the repository if you have remarks regarding repository content/structure.

Lecturers

• lecturer: Bogumił Kamiński
• laboratories:
  • Groups 1 and 2 – Łukasz Kraiński
  • Group 3 – Michał Kot

Schedule

• lectures: Tuesdays, 8:00-10:35

• laboratories: every second Tuesday (check USOS schedule for your group for details)

Lectures

Laboratories

Literature

Materials distributed on MS Teams

Mykel J. Kochenderfer, Tim A. Wheeler, And Kyle H. Wray (2022), Algorithms for Decision Making (https://algorithmsbook.com/)

Stephen Boyd and Lieven Vandenberghe, Introduction to Applied Linear Algebra (http://vmls-book.stanford.edu/)

Gareth J., Witten D., Hastie T., Tibshirani R. (2013), An Introduction to Statistical Learning with Applications in R (http://www-bcf.usc.edu/~gareth/ISL/)

Hastie T., Tibshirani R., Friedman J. (2013), The Elements of Statistical Learning (http://www-stat.stanford.edu/~tibs/ElemStatLearn/)

Kamiński B., Zawisza M. (2012), Receptury w R. Podręcznik dla ekonomisty, Oficyna Wydawnicza SGH (http://bogumilkaminski.pl/projekty/)

B. Kamiński, P. Szufel: Julia 1.0 Programming Cookbook, Packt Publishing, 2018 (https://www.packtpub.com/application-development/julia-10-programming-cookbook)

Course evaluation criteria

• Modeling problem report (50 points); deadline until last laboratory, subject of modelling will be discussed during first class; for more details see section Modeling problem report above

• Laboratory examination (50 points); during last laboratory performed on-line; it will include both practical (writing R code) and theoretical questions (related to statistical models and machine learning)

• Possible extra points:

  • homeworks
  • competition (7th class) - please form groups up to 3 people for the competition, task will be to obtain highest metric on dataset revealed on the day of competition. Each group will have 4 hours to produce the best model (regressor or classifier) and send prediction outcome to the teacher. More information will be available shortly before 7th class.
  • course: https://juliaacademy.com/p/introduction-to-dataframes-jl1 (5 points) - send an e-mail with screenshot of completion before 08.06.2021r EOD

Grading rules