A Systematic Literature Review: Instagram fake Account Detection Based on Machine Learning Sonal Raj, Rupam Kumari, Monika Verma, Gaurav Kumar, Garima Agarwal, Naseem Ahmad Khan Conference Proceedings 1st International Conference on Advancing Sustainable Solutions Through Technologies Icasst 2026, 2026 Fake Instagram profiles ID are a increasing trouble on the platform. We are identifying the problems in Online social networks (OSNs) of false profiles and providing a scheme to solve this issue with handful algorithms.False or fake profiles can be treated for a different malicious purpose, which is used for spreading misinformation, promoting scams, as well as engaging in cyberbullying. This study proposes a machine learning algorithm for attempt to finding false Instagram profiles. This proposed method uses a variety of attributes to train a model that can recognize real or false profiles. The features enter the number of followers as well as following, the age within the account, the amount of activity on the account, with the type of content which is posted. The suggested approach was estimated on a real dataset as well as False Instagram profiles. The outcome showed that the suggested method was able to succeed an accuracy of 95% in identifying False profiles.To recognition of these types of accounts, machine learning methods such as Naive Bayes, logistic regression, SVM as well as neural networks (NN) is applied. Therefore, for Furthermore, the finding of automated profile, cost sensitive genetic algorithm is applied because of the false bias in the dataset. For managing this type irregularity problem in the False dataset, Smote-ne algorithm is implemented.
"AI-Assisted Early Detection of Fibrodysplasia Ossificans Progressiva (FOP) from Clinical Text and Genomic Patterns: A Rare Disease Diagnosis Framework" Kirti Shukla, Krishna Garg, Gaurav Kumar, Sonal Raj, Richa Agarwal, Garima Agarwal Conference Proceedings 1st International Conference on Advancing Sustainable Solutions Through Technologies Icasst 2026, 2026 This research proposes an artificial intelligence (AI)-assisted agenda for the early detection as well as analysis of Fibrodysplasia Ossificans Progressiva (FOP), an ultra-rare genetic disease disorder categorized by progressive heterotopic ossification. Leveraging Natural Language Processing (NLP) methods on clinical narratives which including case reports as well as electronic health records and Few-Shot Learning on genomic dataset related to ACVR1 gene mutations, the outline aims to recognize subtle early-stage which indicate FOP. Given the shortage of interpreted datasets, the methodology includes anomaly finding to pretend rare clinical scenarios as well as improve sensitivity to outlier patterns. Experimental assessment validates the proposed model which can successfully distinguish FOP cases at early stages, by reducing diagnostic delays as well as reducing misdiagnosis. The mixing of explainable AI modules with further more enhanced clinical trust with the decision support. This work contributes to the growing application of machine learning in precision medicine, particularly in the context of rare disease diagnosis.
Autoencoder Deep Learning Algorithm for Finding Rule Mining Harvendra Kumar, Rakesh Kumar, Amit Kumar Pandey, Gaurav Kumar, Navin Prakash Proceedings of the 9th International Conference on Inventive Systems and Control Icisc 2025, 2025 Day to day life, people deal with enormous amounts of transactional data in one form or another. The transactional data in a shop is picked as a record in different forms, and those records are stored in a database. These records are used further for business purposes or taking out the inside it. So, the size of the database is too large and too difficult to handle. Therefore there is a need for a mechanism or technique through which the hidden information can be converted into new knowledge or business knowledge. The purpose of this paper is to compare with DAENMF and the very renowned algorithms like Apriori and FP growth. The organization is as follows: an introduction, followed by related work then research methodology, and finally the conclusion.
Hybrid Ensemble Modeling for Customer Churn Prediction: Integrating CatBoost with XGBoost and Random Forest Kirti Shukla, Jyoti Gaur, Neha Sharma, Gaurav Kumar, Richa Verma, Ajay Pratap Singh Yadav Proceedings 2025 IEEE 3rd International Symposium on Sustainable Energy Signal Processing and Cybersecurity Isssc 2025, 2025 Effectiveness and client lifetime value are in a straight line impacted by customer attrition, which is still a challenge for businesses across all industries. Proper and timely estimation of client loss enables businesses to implement targeted retention efforts. This research proposes a hybrid collective model that incorporates the Random Forest, XGBoost, and CatBoost algorithms on a stacking architecture to improve churn prediction systems' performance. XGBoost offers respectable prediction accuracy by gradient boosting, Random Forest ensures robustness through bagging, and CatBoost efficiently controls unconditional variables that minimizes overfitting. There are distinct benefits to each fundamental model. To train and evaluate the model, a publicly available telecommunications dataset with 7,043 customer records was used. This dataset includes demographic, transactional, and service usage features that reflect real-world churn behavior. A hard training procedure assurances data of superior quality & wide applicability. which includes a structured preprocessing pipeline consisting of feature selection, stratified data splitting, and missing value resolution. Standard classification actions such as AUC-ROC, F1-score, recall, accuracy, precision, & others are used to evaluate the ensemble model, leaving individual learners behind. The crossbreed model enhances accuracy and recall without compromising computational efficiency, based on training data. The interpretable and computationally efficient churn detection methodology in this study advances the investigation of analytics and customer relationship management in dynamic corporate situations. It is made to be seamlessly integrated into real-time business systems.
System for Recommending Crops Through Machine Learning Kirti Shukla, Megha Garg, Sourabh Pal, Gaurav Kumar, Monika Verma, Deepesh Kumar 2025 IEEE 2nd International Conference on Green Industrial Electronics and Sustainable Technologies Giest 2025, 2025 The expansion of automated agricultural techniques is vulnerable by a quantity of factors. India is a good sample because of its diverse landscapes, which include plains, grasslands, rivers, mountains, and deserts. Finding the greatest crop to grow on a given plot of land is still a tough mission, despite a great deal of study in this area. Environmental elements cannot be fully analyzed by the systems established due to a lack of sophisticated technology. A machine learning method for creating a crop recommendation system is presented in this research study. This study will inspect some machine learning algorithms, Random Forest (RF), Decision Trees (DT), Support Vector Machine (SVM), Naïve Bayes, & XGBoost (Gradient Boosting), to recommend crop that is most suitable for a certain plot of land and environmental conditions. Furthermore, the approach may aid control whether crops are fitting for non-farming sites, which vary in terms of moisture content, solar exposure, soil nutrient levels, and water levels. A proposal scheme like this might be useful in undertaking the difficulties of growing crops in non-traditional agricultural settings and on a variety of terrains.
Regulatory Frameworks and Ethical Governance of Neural Networks in Computer Science Applications Gaurav Kumar, Rupam Kumari, Amit Shukla, Ajay Pratap Singh Yadav, Richa Verma, Jyoti Gaur Proceedings 2025 IEEE 3rd International Symposium on Sustainable Energy Signal Processing and Cybersecurity Isssc 2025, 2025 The growing use of neural networks in vital industries including healthcare, banking, and government has increased worries about transparency, accountability, and adherence to international laws. There are a number of Responsible AI methods, but they frequently don't have a cohesive system in place to match technological design with moral and legal requirements at every stage of the model lifespan. In this research, we present the Regulation-Aware Neural Network Lifecycle (RANNL), a new five-stage paradigm that directly integrates explainability, bias mitigation, and real-time compliance tracking into neural network development. The Compliance Scoring System (CS) in RANNL, in contrast to previous methods, statistically assesses models according to their interpretability, fairness, and regulatory preparedness. The EU AI Act, UNESCO AI Ethics Recommendations, and OECD AI Principles are just a few of the international policies that are linked to each step of the lifespan by a dynamic Regulation Mapping Matrix.Using deep learning architectures and regulatory toolkits like AI Fairness 360, Captum, and SHAP, we apply RANNL to a real-world healthcare prediction problem in order to validate the scheme. Model openness, fairness metrics, and compliance scores are all improved without sacrificing predictive performance in comparison to traditional ML pipelines. These findings demonstrate how well RANNL may bridge the gap between global AI governance and neural network engineering. This paper offers an interdisciplinary, quantifiable, and scalable method for creating reliable AI systems by fusing technical and policy viewpoints.
A Comparative Study of Different Cloud Computing Security Models Kirti Shukla, Rupam Kumari, Mr. Deepesh Kumar, Richa Verma, Deepika Khurana, Gaurav Kumar Proceedings of the 4th International Conference on Innovative Mechanisms for Industry Applications Icimia 2025, 2025 Within the boundaries of the Internet, users may now manage portable, distributed computer systems thanks to a concept known as cloud computing. The cloud provides competitors with a competitive platform because of its vibrant source communities, virtualization, and higher-level accessibility. Cloud computing offers a concise and useful description of everyday computing. To address typical computing problems including hardware, software, and the possibility of resource access through cunning computer manipulation, the concept of cloud computing was developed. Cloud computing reservations are a recurring problem, and the network was shut down under the applicable cloud laws. The technology under consideration is referred to as blockchain technology. It is used by cloud platforms, among many others, for security purposes. It is a broad area that encompasses all possible approaches to protecting consumer or cloud computing systems. It is crucial to make clear the support layer of the cloud computing platform, which is already well-established and included into the cloud architecture. This study reviews or investigates cloud computing security across one or more cloud platforms.
Crop Recommendation System using Machine Learning Algorithm Siddhant Goel, Shubham Dhapola, Gaurav Kumar, Vihan Singh Bhakuni, Anil Kumar Sethi, Aditya Verma Proceeding of 2024 International Conference on Communication Computing and Energy Efficient Technologies I3ceet 2024, 2024 Developing automated agricultural practices presents many challenges. Especially in India, with its range of terrains including mountains, desserts, rivers, plains and grasslands. Extensive research has been done in the field of farming but determining the best crop for cultivation on a particular land is still not an easy task. Due to the lack of advanced machinery the systems developed are not capable of properly analyzing environmental factors. This research paper offers a machine learning algorithm to develop a recommendation system for crops. In this research paper, we are going to compare different machine learning algorithms such as Random Forest, Decision Trees, Support Vector Machine, Naïve Bayes, and XGBoost (Gradient Boosting) to suggest the crop which is suitable for a particular piece of land and environmental factors. Additionally, the system can assist in identifying crops suitable for non-farming lands, which differ in environmental conditions such as water levels, soil nutrient levels, sunlight exposure, and moisture. Such a recommendation system holds promise in addressing the challenges associated with cultivating crops on diverse terrains and non-traditional farming areas.
Evaluating the Effectiveness of Deep Learning Models in Network Intrusion Detection Gaurav Kumar, Pankaj Gupta, Ganesh Kumar Yadav, Richa Verma, Jai Prakash Bhati, Vihan Singh Bhakuni 2024 International Conference on Cybernation and Computation Cybercom 2024, 2024 Network security is now a significant concern in the current digital era since the increasing volume of cyber-attacks in day-to-day life has become a potential threat for the integrity and privacy of data. Most traditional intrusion detection systems face problems when it comes to detecting new and sophisticated attacks. The paper goes ahead to elaborate on how the concepts of machine learning and deep learning can be utilized to enhance the process of detecting and classifying network attacks. For instance, we experimented on the UNSW-NB15 dataset with different models: Decision Tree, Random Forest, SVM, KNN, MLP, and DNN. We have proved through our results that such kinds of models can considerably enhance the accuracy and robustness of IDS. With an average accuracy of 94.91%, the Decision Tree model took second place as the Deep Neural Network was closely followed. We will outline performance metrics based on the respective strengths and weaknesses of the models. This research is used to construct more effective and reliable network intrusion detection systems, which lead further into advancements of cybersecurity.
