sujata Chakravarty - Academia.edu (original) (raw)

Papers by sujata Chakravarty

International journal of web and grid services, 2024

Technologies, Mar 19, 2024

The International journal of computational intelligence systems/International journal of computational intelligence systems, Jan 29, 2024

In recent years, deep learning techniques have presented a major role in hyperspectral image (HSI... more In recent years, deep learning techniques have presented a major role in hyperspectral image (HSI) classification. Most commonly Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) has greatly advanced the accuracy of hyperspectral image classification, making it powerful tool for remote sensing applications. Deep structure learning, which involves multiple layers of neural network, has shown promising results in effectively addressing nonlinear problems and improving classification accuracy and reduce execution time. The exact categorization of ground topographies from hyperspectral data is a crucial and current research topic that has gotten a lot of attention. This research work focuses on hyperspectral image categorization utilizing several machine learning approaches such as support vector machine (SVM), K-Nearest Neighbour (KNN), CNN and LSTM. To reduce the number of superfluous and noisy bands in the dataset, Principal Component Analysis (PCA) and Minimum Noise Fraction (MNF) were utilized. Different performance evaluation measures like time taken for testing, classification accuracy, kappa accuracy, precision, recall, specificity, F1_score, and Gmean have been taken to prove the efficacy of the models. Based on the simulation results, it is observed that the LSTM model outperforms the other models in terms of accuracy percentage and time consumption, making it the most effective model for classifying hyperspectral imaging datasets.

$Research paper thumbnail of A systematic review on approach and analysis of bone fracture classification$

Materials Today: Proceedings, 2023

Abstract As a result of accidents or other injuries, bone fractures are becoming more common in o... more Abstract As a result of accidents or other injuries, bone fractures are becoming more common in our country. According to the India market survey study, fracture cases are becoming more common in Indian hospital records. The incidence rising by over 4.4 lakh in the last three decades and projected to reach over 6 lakh by 2020. The authors of this paper attempted to explain various forms of fracture detection techniques. This paper is folded into six halves. First, we will go through the introduction and data preparation step. Second, we discussed related work on fracture detection so far. Third, we look at different feature extraction methods that may be used to diagnose bone fractures. Fourth, we look at both traditional and deep learning-based methods for detecting bone fractures. Fifth, we looked at performance evolution approaches for determining the correctness of various algorithms. Sixth, we go through the many concerns and obstacles that researchers confront when working with fracture detection. The majority of authors are only concerned about whether the bone is broken or not, with very few concentrating on the classification of bone fractures. This paper aims to aid researchers in developing models that can automatically detect and classify fractures in human bones by providing a preliminary decision support system.

2020 International Conference on Computer Science, Engineering and Applications (ICCSEA), Mar 1, 2020

Healthcare is an important field where image classification has an excellent value. An alarming h... more Healthcare is an important field where image classification has an excellent value. An alarming healthcare problem recognized by the WHO that the world suffers is diabetic retinopathy (DR). DR is a global epidemic which leads to the vision loss. Diagnosing the disease using fundus images is a timeconsuming task and needs experience clinicians to detect the small changes. Here, we are proposing an approach to diagnose the DR and its severity levels from fundus images using convolutional neural network algorithm (CNN). Using CNN, we are developing a training model which identifies the features through iterations. Later, this training model will classify the retina images of patients according to the severity levels. In healthcare field, efficiency and accuracy is important, so using deep learning algorithms for image classification can address these problems efficiently.

2020 4th International Conference on Trends in Electronics and Informatics (ICOEI)(48184), Jun 1, 2020

Android malware is a ubiquitous threat to the information security of mobile users. Android users... more Android malware is a ubiquitous threat to the information security of mobile users. Android users often download applications from unauthorized and untrusted sources. Such applications may request several permissions from the user, and due to unawareness, the user may grant the required permissions. Android permissions are one of the significant sources of malware infection. By analyzing the permissions database classification of malware and benign applications can be done with the help of machine learning tools. There are a total of 330 permissions in Android applications. However, all of them may not contribute to the classification. In this paper, the proposed system investigates identifying the most influential permissions using feature reduction. The gain ratio is used for feature reduction and J48, Random Committee, Multilayer Perceptron, Sequential Minimal Optimization (SMO), and Randomizable filtered classifiers are used for evaluation of the selected features. The experimentation results show that five permissions can produce near full feature accuracy, thereby optimizing the malware detection system.

$Research paper thumbnail of A statistical model approach based on the Gaussian Mixture Model for the diagnosis and classification of bone fractures$

International journal of healthcare management, Jan 10, 2023

$Research paper thumbnail of Feature extraction and classification of hyperspectral imaging using minimum noise fraction and deep convolutional neural network$

Journal of Electronic Imaging, Nov 15, 2022

In recent years, researchers have frequently utilized convolutional neural networks (CNNs) to cla... more In recent years, researchers have frequently utilized convolutional neural networks (CNNs) to classify hyperspectral images and have, indeed, embraced exciting achievements. However, most of the existing approaches tend to handle images block by block, which is less efficient as image blocks need to be fed into the network for many times. With this in mind, this letter presents a novel hierarchical CNN that adopts raw images as the input and extracts useful features for classification. Specifically, we adopt several hierarchical convolutional neural layers as a feature extractor and adopt the support vector machine instead of the classifying layer in the original network as the final classifier. Experiments show the proposed approach can work efficiently and exhibit competitive performance when compared to some other approaches based on deep networks.

$Research paper thumbnail of A systematic approach to diagnosis and categorization of bone fractures in X-Ray imagery$

International Journal of Healthcare Management

In the last few years, anonymized credit card transactions have grown more threats that have caus... more In the last few years, anonymized credit card transactions have grown more threats that have caused serious consequences in the finance and banking sectors. Due to the dramatical growth of the online payment system, now many banks and financial sectors are implementing various types of automatic fraud detection system to analyze the fraud transactions; machine learning (ML) is one of the promising approaches to find out the fraud transactions. Machine learning methodologies have proved the most promising solution for anonymized transactions. This paper comparatively analyzes the basic machine learning algorithms which include SVM, LDA, QDA, DT, and RF for fraud detection. At the same time, some of the modern open-sourced boosting machine learning algorithms which include XGBoost, LGBoost, and CatBoost are also implemented.

Electronics

Cyber threat intelligence includes analysis of applications and their metadata for potential thre... more Cyber threat intelligence includes analysis of applications and their metadata for potential threats. Static malware detection of Windows executable files can be done through the analysis of Portable Executable (PE) application file headers. Benchmark datasets are available with PE file attributes; however, there is scope for updating the data and also to research novel attribute reduction and performance improvement algorithms. The existing benchmark dataset contains non-PE header attributes, and few ignored attributes. In this work, a critical analysis was conducted to develop a new dataset called SOMLAP (Swarm Optimization and Machine Learning Applied to PE Malware Detection) with a value addition to the existing benchmark dataset. The SOMLAP data contains 51,409 samples that include both benign and malware files, with a total of 108 pure PE file header attributes. Further research was carried out to improve the performance of the Malware Detection System (MDS) by feature minimiz...

International Journal of Web and Grid Services

International journal of modern agriculture, Dec 1, 2020

Materials Today: Proceedings, 2021

Abstract Intrusion Detection System (IDS) is one of the best ways to combat several cybercrimes, ... more Abstract Intrusion Detection System (IDS) is one of the best ways to combat several cybercrimes, both at the edge of the network and inside the segments of the internal network. This article proposes a hybrid learning approach namely Improved Harmony Search Extreme Learning Machine based IDS (IHSELMIDS) to classify NSL KDD dataset. This dataset is a polished version of its predecessor i.e., Knowledge Discovery in Databases (KDD). Since 1999, many researchers have relied on the dataset of KDD for the evaluation of anomaly detection system. Thus, it is popularly known as KDD”99 dataset. Improved Harmony Search has been used to boost the weights of input and latent biases for a more robust and stable Extreme Learning Machine (ELM). In addition, the generalized inverse Moore – Penrose is used to systematically evaluate the weights of output. Additionally, to address the curse of high dimensionality in the dataset, correlation-based feature selection with greedy hill climbing is proposed which reduces time complexity while increases computational efficiency. A series of performance evaluation measures such as training and testing accuracy, True Positive Rate (TPR), True Negative Rate (TNR), G-mean, F-score, False Alarm Rate (FAR), Receiver Operating Characteristic curve (ROC) and Confusion Matrix is into consideration to contrast and examine the efficiency, flexibility and reliability of the proposed model. The experimental result showed that IHSELMIDS outperforms all the benchmark models considered in this study.