---
tags: Meeting
---

# Trivago
> https://recsys.trivago.cloud/challenge/dataset/

## Session

* 這裡的session都是紀錄有登入的使用者，故單一使用者會有不同天的session紀錄
* Split date由比賽方決定
* X是觸發點擊
* ?是要predict的click item
* validation group => public leaderboard
* confirmation group => private leaderboard

## Example

Query 1:
impressions = [100, 101, 102, 103, 104, 105]
clicked_item_id = 102
submission = [101, 103, 104, 102, 105, 100]
reciprocal rank = 0.25

Query 2:
impression = [101, 103, 104, 100, 105]
clicked_item_id = 105
submission = [103, 105, 101, 100, 104]
reciprocal rank = 0.5
mrr = (0.25 + 0.5) / 2 = 0.375

Evaluation: MRR@25
> https://github.com/keyblade95/recsys2019/blob/master/visualize_data/NN%20impressions_features_input.ipynb

## Features
Session actions
> https://github.com/keyblade95/recsys2019/blob/master/visualize_data/click_out%20accommodations.ipynb

* user_id: identifier of the user
* session_id: identifier of each session
* timestamp: UNIX timestamp for the time of the interaction
* step: step in the sequence of actions within the session


## Observations
### User session clicks on more than one hotel
> https://github.com/keyblade95/recsys2019/blob/master/visualize_data/same_clickout_user_more_clickouts.ipynb

* 約44趴的人


### Sessions amount of per user
> https://github.com/keyblade95/recsys2019/blob/master/visualize_data/sessions_per_user.ipynb

* 有163609的user有超過一個session, 總過有948041個user, 表示有17趴的user有超過1個session
* 一人平均1.2個