# AQC Midnight Roast Logistic Regression — Data

来源：https://www.kaggle.com/competitions/aqc-midnight-roast-logistic-regression/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

(midnight-roast-challenge / Kaggle-ready Markdown)
### Files
- `train.csv` – 350 taste-tester records with ground-truth label
- `test.csv` – 150 customers; labels hidden
- `sample_submission.csv` – correct upload format


## Columns

| Column                            | Meaning                                   | Values | Missing    |
| --------------------------------- | ----------------------------------------- | ------ | ---------- |
| Id                                | unique observation id (index)             | int    | no         |
| Age                               | customer age (years)                      | float  | 5 %        |
| Prefer\_Dark\_Chocolate           | likes dark chocolate                      | 0, 1   | no         |
| Usually\_Drinks\_Espresso         | usually drinks espresso                   | 0, 1   | no         |
| Coffee\_Acidity\_Rating           | perceived acidity 1–10                    | float  | 5 %        |
| Roast\_Level                      | roast darkness: 1=Light, 2=Medium, 3=Dark | int    | no         |
| Sugar\_Preference                 | grams sugar typically added               | float  | 5 %        |
| Previous\_Dark\_Roast\_Experience | has tried dark roast before               | 0, 1   | no         |
| Likes\_Coffee                     | target: 1 = enjoyed “Midnight Roast”      | 0, 1   | train only |


#### What am I predicting?
A single binary label Likes_Coffee for each customer in test.csv.
#### Acronyms / Terms
MCAR – Missing Completely At Random (small fraction in Age, Acidity, Sugar)