Muhammad Maaz Akhter

Muhammad Maaz Akhter

Auto-generated profile, based on publicly available information

Muhammad Maaz Akhter

Department of Computer Science, FAST National University of Computer and Emerging Sciences, Karachi, Sindh, Pakistan

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
8 May 2024 in VAWKUM Transactions on Computer Sciences
Verified
Journal Article
Open access
PDF file
FAFayyaz Ali
Fayyaz Ali
QAQurban Ali
Qurban Ali
FSFarhan Bashir Shaikh
Farhan Bashir Shaikh
RLRaja Sohail Ahmed Larik
Raja Sohail Ahmed Larik
MSMuhammad Saad
Muhammad Saad
MAMuhammad Maaz Akhter
Muhammad Maaz Akhter
This research presents an advanced Federated Learning approach for privacy-preserving and secure Ecommerce NLP System. Recently, the amount of information is growing very fast, and its importance is also rising quickly. Today, businesses handle huge volumes of data, like emails, which are often key to how they run day to day. It's difficult to find useful information from all this data and keeping it safe is a big worry. Keeping data secure is very important for protecting the trust of an online store and making customers feel safe when they make purchases. Hence, in this paper we have developed a privacy-preserving and secure FedSecureNLP system for text summarization, sentiment analysis and question answering. Within the FL configuration, the local training text summarization, sentiment predication and question answering is performed autonomously on each node local dataset. The Bidirectional Encoder Representations from Transformers (BERT) tokenization is used in the model to create tokens and extract term from amazon review data, after that, hierarchical deep learning for text (HDLTex) is used to predict sentiment rating. The Deep learning network (DeepCNN) is then used to do extractive summarization. Random multimodal deep learning (RMDL) is used for QA prediction. The QA model has different training and testing steps in the Module. During the evaluation phase, pretrained RMDL process the input query and semantically analyzed and generates relevant response. In this case FCDVO is designed to adjust the HDLTex, Deep CNN, and RMDL hyperparameters. The precision, recall, F-measure and Root Mean Square Error (RMSE) of the System that presented were 0.088, 92.896%, 92.481% and 92.688% respectively