Skip to content

Latest commit

 

History

History
53 lines (39 loc) · 1.71 KB

SKILL_TEST.en.md

File metadata and controls

53 lines (39 loc) · 1.71 KB
Raw

Background

Suppose you are assigned to a project that develops a price prediction service for an online marketplace. You will build a REST service which returns a predicted price from listing information.

Assignment

Develop a price prediction model and a REST endpoint

POST /v1/price

Request body

  • Any of item information listed in the Data section
  • e.g.
{
  "name":"Hold Alyssa Frye Harness boots 12R, Sz 7",
  "item_condition_id":3,
  "category_name":"Women/Shoes/Boots",
}

Response

  • Json format
  • e.g. when a predicted price is $30, it should be {"price": 30}.

Data

The data is available in the data/ directory. mercari_train.csv and mercari_test.csv consist of a list of product listings

  • id: the id of the listing
  • name: the title of the listing
  • item_condition_id: the condition of the items provided by the seller
  • category_name: category of the listing
  • brand_name: brand of the listing
  • price: the price (USD) that the item was sold for. This column doesn't exist in mercari_test.csv
  • shipping: 1 if shipping fee is paid by seller and 0 by buyer
  • item_description: the full description of the listing
  • seller_id: the seller ID of the listing

Requirements

  • Python 3.7 or higher
  • A Dockerfile is required and your API server should be runnable on Docker
  • Unit tests for the API server
  • A README which describes how to run the unit test and server
  • Training code for your price prediction model
  • Model file size should be smaller than 10MB (You can compress your model file)
  • A csv file for your prediction result (Please refer to sample_submission.csv)
  • Root mean squared logarithmic error (RMSLE) for the mercari_test.csv should be less than 0.5