E-commerce product recommendation system

• Main Author: Mody, Kimisha Piyush.
• Other Authors: Vitali Zagorodnov
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering::Information systems
• Online Access: http://hdl.handle.net/10356/52030
• Institution: Nanyang Technological University

In today’s world, filtering vast amount of information has become an important part of the daily life of an increasing number of people. The amount of information available through books, movies, news and advertisements has become enormous. It is no longer feasible for the customer to filter through the vast array of products available unaided, to quickly find those that are most interesting and relevant. In response to this challenge, recommender systems based on information-filtering techniques are used. Recommender systems assist and augment this natural social process to help people sift through available books, articles, web pages, movies, music, restaurants, jokes, grocery products, etc. This report will cover the investigation and implementation of recommendation algorithms for an e-commerce product website, www.pupsikstudio.com. The author specifically explores and evaluates one such recommendation technique: Collaborative filtering. Collaborative filtering is the process of making recommendations for the potential preference of a user based on the preference of the user as well as a number of other users for various items. In this report, the author has compared two different collaborative filtering approaches and used them to make recommendations for specific users in the website database: one, using the Pearson Correlation similarity, henceforth referred to as Method 1, and the other, an intuitive method which considers two items to be similar if purchased together in the same order, referred to as Method 2. All implementation has been performed using Python as a programming language. The results of the author’s investigation indicate that Method 1 produces more accurate results, while Method 2 performs better in terms of time complexity. The results and analysis have been detailed in this report. Recommendations made to users are in the form of emails, providing them with a list of suggested products.