Midterm Assignment

class: center, middle, inverse, title-slide

# Midterm Assignment
## Kaggle Prediction Competition
### Itamar Caspi
### May 9, 2021 (updated: 2021-05-10)

---

# What is Kaggle?

.pull-left[
- Kaggle is a huge data science community where machine learning practitioners around the world compete against each other.

- The datasets used in Kaggle are uploaded by public companies as well as private users.

- A "kaggler" wins if her algorithm is the most accurate on a particular data set.

- Kaggle competitions are one of the best places to practice your ML skills and learn about state-of-the-art ML method.  
]

.pull-right[

<img src="figs/titanic.jpg" width="50%" style="display: block; margin: auto;" />
]

---
# Introduce yourself to Kaggle

.pull-left[
1. Visit [www.kaggle.com](www.kaggle.com) and sing-up.

2. Go to the ml4econ course competition [webpage](https://www.kaggle.com/t/7d925d886da049b88e99e4d2eb3a9add).

3. Review competition details: objectives, deadline, data, evaluation, submission rules, etc. 
]
.pull-right[
<img src="figs/ml4econ-kaggle.png" width="1265" style="display: block; margin: auto;" />
]

---
# Kaggle competition data structure

.pull-left[
- MSE for the public test set (30%) immediately available at submission.

- MSE for the private test set (70%) available only once the competition closes.

- The split between public and private test sets is arbitrary and unknown in advance to all competitors.

Your final ranking is based on how well you perform on the *private* test set.
]
.pull-right[
<img src="figs/mse.png" width="865" style="display: block; margin: auto;" />
]

---
# The basic Kaggle competition workflow

1. Acquire domain knowledge.

2. Explore the data.

3. Preprocessing (standardization, dummies, interactions, etc.).

4. Choose a model class (Lasso, ridge, trees, etc.).

5. Tune complexity (Cross validation).

6. Submit your prediction.

7. __Document your workflow (R Markdown)__

---
# Tracking your performance

.pull-left[
- Use the public lead-board to track your performance.

- Your ranking ("scores" column) is based on your MSE on the public test set.

- Once the competition is closed, the final ranking will be based on the MSE on the private test set.

- Your can submit multiple predictions but be careful not to overfit the public test set!
]
.pull-right[
<img src="figs/tracking.png" width="1085" style="display: block; margin: auto;" />
]

---
# Getting started

Running the following code chunk will automatically download the data (train, test, and a sample submission file) you'll need for our Kaggle competition:

```r
library(tidyverse)

train <- read.csv("https://raw.githubusercontent.com/ml4econ/lecture-notes-2021/master/a-kaggle/data/train.csv")

test <- read.csv("https://raw.githubusercontent.com/ml4econ/lecture-notes-2021/master/a-kaggle/data/test.csv")

sample_submission <- read.csv("https://raw.githubusercontent.com/ml4econ/lecture-notes-2021/master/a-kaggle/data/sample_submission.csv")
```

__NOTE:__ By default, a new project will be created on your desktop.

---
class: .title-slide-final, center, inverse, middle

# `slides %>% end()`

[<i class="fa fa-github"></i> Source code](https://raw.githack.com/ml4econ/notes-spring2021/master/a-kaggle/a-kaggle.html)