Trusted Data Retrieval: A Verifiable Search Model for Multi-Tenant Applications

View at publisher Summarise with VIRAT
Authors: Anjali B, Dr. S. THULASEE KRISHNA
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This research addresses the critical challenge of fraudulent ad clicks in mobile advertising, which causes massive financial losses for advertisers and skews campaign performance analytics. We propose an intelligent, adaptive framework that distinguishes betweengenuine and fraudulent clicks by leveraging both high precision Machine Learning models and sequential Deep Learning architectures. The methodology utilizes thelarge-scale TalkingData dataset, employing feature engineering to extract temporal patterns. A novel contribution of this study is the implementation of a hybridStacking Classifier that integrates Gradient Boosting models like XGBoost with sequential models such as LongShort-Term Memory (LSTM) and Gated Recurrent Units (GRU). Performance ev aluations show that the hybridensemble achieves a superior ROC-AUC of 0.985, outperforming standalone models. Furthermore, the system is deployed via a real-time Flask web application,proving its feasibility for live production environments. Traditional detection systems are often rule-based andstatic, making them unable to generalize across diverse data or adapt to evolving fraud patterns. This article presents a suggested setup for an intelligent, data-driven system that accurately classifies ad clicks. By integrating sequential Deep Learning (DL) with ensemble MachineLearning (ML), we bridge the gap between complex predictive analytics and practical real-time deployment. 2. Related Work Early approaches to fraud detection relied on static heuristics, such as blacklisting IP addresses or setting clicks-per-minute thresholds. While fast, these methods are brittle and easily bypassed by fraudsters using rotating proxies. This end-to-end solution provides a scalable and robustmechanism for modern digital advertising security. Keywords: Ad Click Fraud, Stacking Classifier, LSTM,XGBoost, Deep Learning, Flask.

Fields of Study & Subjects

Spam and Phishing Detection Imbalanced Data Classification Techniques Information Retrieval and Search Behavior

Publication details

Journal / source
International Scientific Journal of Engineering and Management
Volume
05
Issue
03
Pages
1–9
Publication date
21 March 2026
Language
EN