Improving E-commerce Recommendations in the Cloud with Autoencoders and Firefly Algorithm
Author : Sunil Kumar Alavilli and R Pushpakumar
Abstract :
Recommendation engines for online retailers are essential for enhancing the consumer experience through the suggestion of appropriate products according to user interests. Current systems do not handle scalability, accuracy, and dealing with large datasets effectively. Conventional approaches, Comprehensive user-item interactions cannot be adequately captured by techniques like matrix factorisation and collaborative filtering current models tend to miss complex, non-linear user-item relationships, which results in a less-than-ideal recommendation quality. Furthermore, a high computational overhead of processing large-scale data is still a major hindrance in this work, a novel Autoencoder-based recommendation model optimized by the purpose of the Firefly Algorithm (FA) for hyperparameter optimisation is to improve ranking quality and suggestion accuracy. The innovation of this method is in the combination of Autoencoders with the Firefly Algorithm for efficient capture of non-linear interactions and optimization of model parameters to greatly improve recommendation performance and scalability. The generated model yielded an NDCG score of 0.98, 98.2% accuracy, 97.6% precision, 96.9% recall, and 97.2% F1-score, reflecting its capacity for highly accurate and relevant recommendations as well as maintaining optimal ranking quality as compared to the conventional methods the autoencoder-based model with in terms of accuracy, precision, and recall, FA optimisation fared better than any of the baseline models utilising techniques like collaborative filtering and matrix factorisation exhibiting stronger ability to process large-scale datasets and intricate user-item interactions. This methodology brings important gains in recommendation accuracy and ranking quality, especially with large-scale and sparse data. The incorporation of the Firefly Algorithm allows effective hyperparameter optimization, making the model more scalable and suitable for real-time e-commerce scenarios, thereby influencing the future evolution of recommendation systems within cloud-based systems.
Keywords :
E-commerce recommendations, autoencoders, firefly algorithm, hyperparameter optimization, cloud computing.
