DA3-Coding-Examples

Coding classes for Data Analysis 3 on MSc in Business Analytics on the Central European University

How to run the codes?

Open an R project as given on the book's home page: How to set up your computer for R.

Download the appropriate files, or fork this repo, clone it, and open the code from your R project's environment. Once you're done, you are good to go.

Contents

1. class 13, used cars
   Basic data manipulation and exploratory data analysis.
   Basic visualizations; plotting logged values in ggplot.
   Multiple linear regression.
   Model selection by goodness-of-fit metrics.
   Cross-validation and model comparison.

2. class 14, airbnb
   Handling missing data; integrating missing data information in the analytics.
   Model setup.
   Interactions and dummies.
   Train, test and holdout sets.
   Cross-validation, train and test metrics.
   Lasso:

   • running a lasso optimization
   • interpreting the results
   • RMSE
     

   Diagnostics on the holdout set.
   Plotting prediction results.

3. class 15, used cars
   Data manipulation as in class 13
   Basic regression trees
   Plotting trees and regression results as step functions
   Building more complex regression trees with control parameters
   Pruning
   Comparing tree-based and OLS models
   Variable importance: with final only and with competing variables
   

4. class 16, airbnb, hitters
   Setting up grid for grid search in caret::train
   Running random forest model using the ranger package in caret
   Getting and plotting individual and grouped variable importances
   Partial dependence plots for rf models Predictions and RMSE for subsets of data
   Comparing OLS, LASSO, CART, and random forest
   Gradient Boosting Machines: tuning and model run
   Hitters: parameter grid search on a smaller and easier-to-handle dataset
   The airbnb analysis is implemented both in R and in Python using a Jupyter notebook
   

5. class 17, bisnode
   Modelling probabilities with simple & lasso logit
   CV RMSE & AUC for probability models
   Classification using logit with no loss function
   Calibration plot, confusion matrix, ROC, AUC
   Classification (logit) with user-defined loss function
   Classification with CART
   Random forest for probabilities, with and without a loss function
   Classification with random forest
   

6. class 18, swimming pool, Case-Shiller
   Managing time series data with the tsibble package
   Deterministic modelling: OLS, trend, seasonal & other dummies
   Introducing fbProphet
   Stochastic modelling with the fable package
   ARIMA, auto-arima
   Vector Autoregressions