Decentralized Smartphone Recommendation with Privacy Preservation

1 NALLURI VENKATA SARANYA, 2 MANUKONDA LAKSHMI THANMAI, 3 CHILUKOTI NAGA SAI,4 NIMMAGADDA CHANDRA SEKHAR

1,2,3 B. Tech Students, Department of CSE, RVR & JC College of Engineering, Chowdavaram, Guntur, A.P, India.

4Assistant Professor, Department of CSE, RVR & JC College of Engineering, Chowdavaram, Guntur, A.P, India,

E-mail:1 y22cs127@rvrjc.ac.in, 2 l23cs210@rvrjc.ac.in, 3 l23cs203@rvrjc.ac.in, 4 nimmagadda65@gmail.com

Abstract—With the rapid growth in mobile phone usage, online platforms now contain a huge number of customer reviews. While these reviews are helpful, they can also overwhelm consumers, making it difficult to choose the right product. To address this issue, this study introduces a new approach for classifying mobile phone ratings using a recommendation system based on federated learning and TF-IDF features.For this research, we created a new dataset by collecting more than 13,000 mobile phone reviews from Flipkart. The proposed method uses a Federated Deep Neural Network (FDNN) to classify ratings. The process includes data cleaning, handling imbalanced data, extracting features using TF-IDF, and making predictions through federated learning.The system is designed with two client models and one central server, and experiments were conducted over three rounds. The results show that the model achieved an accuracy of 96.68% at the server level while ensuring that user data remains secure on local devices.This approach can help customers make better purchasing decisions and can also be applied to other e-commerce platforms with large volumes of reviews.

Index Terms—Flipkart smartphone rating, recommendation system, novel dataset, deep learning, federated learning.