shopping-cart-completion-problem

Project Objective

The objective of this unsupervised machine learning project was to gain a comprehensive understanding of how recommendation engines work. The focus was on applying various recommendation engine algorithms and techniques to the shopping cart dataset, enabling an in-depth exploration and learning of recommendation systems.

Overview

In the highly competitive online grocery delivery market, ensuring a seamless and customer-centric shopping experience is crucial. A common issue faced by customers is forgetting to purchase essential items. To address this, your task is to design and implement a "Did You Forget?" system that assists users by suggesting items they may have forgotten to add to their cart before checking out.
(kaggle competition link: https://www.kaggle.com/competitions/did-you-forget-b-25/overview)

Problem Statement

Develop a machine learning-based recommendation system that identifies and suggests items likely forgotten by customers based on their shopping behavior, past purchases, and current cart contents.

Task Objective

• Build a recommendation system to predict the remaining items from the last order that the customer might have forgotten to order based on customer's previous order.
• The recommendation system should make recommendations: one that contains a set of 5 SKUs as recommendation.
• Generate the recommendations for all the orders available in last_orders_subset.csv

Data

There are two datasets:
all_except_last_orders.csv
The above file contains all the orders made by customers of an online retail store for a specific duration of time.
last_orders_subset.csv
This file contains the SKUs that contains the last order (most recent) by the customers. But it contains only a subset (around 70% to 80%) of the SKUs from the order. The remaining SKUs are withheld as the items that the customer have forgotten to order.

Evaluation Criteria

On submission of the recommendation submissions will be evaluated with the actual test set (withheld) using recall@5 metrics. recall metrics will be measured using:
$$\frac{1}{N}\sum_{i=1}^{N}\frac{\text{Number of forgotten items identified by the model}}{\text{Number of items forgotten by the customer in order}}$$

Citations

@misc{did-you-forget-b-25,
    author = {manaranjan pradhan},
    title = {Did you forget B23},
    year = {2025},
    howpublished = {\url{https://kaggle.com/competitions/did-you-forget-b-25}},
    note = {Kaggle}
}