Runtime Based Recommendations on Netflix Data using SBE-XGBoost model

Authors

  • Rajeswari Nakka, Dr. G. V. S. N. R. V. Prasad, R.Kiran Kumar

Abstract

Recommendation Systems (RS) are developed and improved to provide meaningful
recommendations of a products or services or items to a set of users who might seek attention towards it. Many
RS were developed for the use by various real world companies like Amazon, Netflix, Spotify and some social
media websites etc, which uses it to make profits. The paper focuses on applying machine learning techniques
to offer recommendations on Netflix dataset by implementing Runtime based Recommendations. A runtime
computation technique was designed to overcome the data sparsity, curse of dimensionality problem, memory
and computational issues for larger datasets. Here the experiments are performed in Google Colab platform for
computing and analysing the large datasets. The proposed approach is a combination of Surprise Baseline
Estimator (SBE) and eXtreme Gradient Boosting (XGBoost). The designed runtime computation technique
outperformed Matrix Factorization approaches Singular Value Decomposition (SVD) and Truncated SVD. The
proposed model SBE-XGBoost evaluated with novelistic approach by combining SBE and XGBoost model to
evaluate the training and test data which gave good prediction results on test data compared to the existing
system.

Published

2020-04-30

Issue

Section

Articles