Introduction

In order for customers to find our products while browsing the walmart.com site, we need to arrange the products onto virtual shelves, for example, "Smart TVs".

In this problem we ask for a machine learning solution to the task of assigning products to shelves.

The training data consists of detailed product information from 32 shelves. Your algorithm should suggest all shelves that each product should appear for the provided test set, which contains product information from the same 32 shelves. Each product can appear on more than one shelf.

Dataset

We provide a training dataset of 10593 products with various detailed product data, such as "Product Name", "Product Description", and various other attributes about the product. The columns are delimited by tabs and the data is found in train.tsv, with each row describing a different product. These products are found on a total of 32 shelves. The shelves are labeled by integer values, and a list of all shelves that these products appear is given in the tag column for the training data.

We also provide a test set of 10593 products, without the labels, in test.tsv. The goals is to predict the list of shelves that these products are assigned to.

You can download the zip file provided here. The zip file contains both the files train.tsv and test.tsv.

Submission Details

The output file, tags.tsv (max allowed size is 10MB). The file should contain each product from the file test.tsv followed by a list of the shelves for the product contained within square brackets.

A valid output file has the following format:

item_id tag 10593 [4483] 10594 [581514] 10595 [4483] 10596 [4537] 10597 [1229817, 1229821]

Note that:

The first line of the output file should contain the header, with item_id, and tag as the column names separated by a tab. Products can appear on multiple shelves.

There are no limits on the number of shelves a product can be placed.