Intelligent Systems and Machine Learning

Accurate brain tumor subtypes classification is significant for prognosis and treatment. The aim of this research is to improve the multiclass brain tumor classification using vision transformer as feature extractor. In this study, we first optimized and employed deep learning ResNet101 for feature extraction and fed to machine learning classifiers for multi-class classification. We then optimized and employed vision transformer and fed these features to machine learning decision classifier. We measured the performance with standard performance metrics. The Artificial Intelligence vision transformer with decision tree classifier yielded highest multi-class classification performance with 99.89% accuracy and 1.00 AUC to detect pituitary followed by 97.69% accuracy and AUC of 0.96 to detect meningioma. The results are compared with ResNet101 with transfer learning. ResNet101 deep features by utilizing KNN yielded detection of pituitary tumor (98.80% accuracy, 0.99 AUC), glioma (93.47% accuracy, 0.93 AUC). The results revealed that proposed approach with vision transformer and decision tree features extractor are more robust in detecting multiclass brain tumor prediction. The proposed approach can be better utilized for betterment of treatment and prognosis to obtain improved clinical outcomes.

One of the most thoroughly researched and well-documented non-linear infectious disease dynamical systems is measles. Infants and young children are most likely to contract the immunizable disease measles. Measles is a highly commutable viral infection that has a 90% secondary infection incidence among contacts who are vulnerable. In this study, we have used a deep convolutional neural network to discriminate between various skin diseases and measles rash. The categorization performance of each individually optimized DL model across all of their ensembles has been presented using the specified dataset. We tested four optimizers, namely SGD, ADAM, RMSprop, and RAdam, on three considered models in order to further improve them. These models include VGG16, InceptionV3, and ResNeXt50, on which individual 10-fold cross-validation is done. The maximum average 10-fold cross-validation accuracy of 98.62%, 99.31% recall, and 99.32% F1 score were achieved by the optimised Inception V3 using the SGD optimizer. Finally, our predictive model offers a method for early detection to assist physicians in treating and enforcing new laws and regulations.

The processing and analysis of brain imaging to identify transient ischemic strokes has remained difficult due to the requirement for more precise abnormality identification and the extraction of concealed but essential information from image data. This is necessary in order to diagnose transient ischemic strokes. Because of both of these conditions, identifying people who have had transient ischemic strokes has become more challenging. In order to arrive at a diagnosis of transient ischemic stroke, it is necessary to have fulfilled either one of these conditions. The work that is being done right now has the intention of achieving a higher level of precision in the process of extracting and selecting features from image data. The work that is being done right now places a significant emphasis on this particular aspect. This is being done in order to obtain a more in-depth understanding of the images in terms of the detection of abnormalities, and it is being done so right now. By analysing multiple groups of abnormalities side by side, the purpose of this research is to help advance the development of MRI and CT scans that are more accurate. The comparison of several different types of anomalies is the primary focus of this research.

Stress, sadness and panic have all become major issues in our contemporary culture. Stress has become one of the top ten socioeconomic predictors of health inequalities. The electroencephalogram (EEG) signals and machine learning approaches are utilized to predict the mental state of the person. This has become a significant topic of research in recent times in health care system. There are various ways are used to monitor stress. The primary goal of this study is to identify stress in humans. Because of its potential value, stress detection based on EEG signals has emerged as an interesting study topic. This research looks into brain waves to classify a person’s mental state. Despite the fact that there is no precise way of defining the optimum feature for a classifier, the features utilized as classifier input have a significant impact on the classification outcomes. An algorithm for stress level detection from EEG is proposed in this paper. The Euclidean distance scale is commonly used in the paper for EEG signal identification. In this study, EEG data is separated into EEG rhythms using a band pass filter method, EEG signals are normalized and a k-mean clustering method is used to classify brain wave signals to detect the mental stress.

Obviously, individuals all over the world make a solid effort to stay aware of the hustling scene. Nonetheless, thus, every man and lady is managing interesting wellness issues, one of the most notable of which is misery or stress, which can prompt passing or other horrifying demonstrations. These inconsistencies are alluded to as bipolar problem, which can be treated by following a couple of expert suggested medicines. Victims who have been determined to have psychological wellness issues have their circumstances analyzed to assist them with approaching their regular routines. Positive conditions, such as Schizophrenia and Bipolar Disorder, have a higher likelihood of continuing crises. Mental health professionals are responsible for reducing the risk of patients experiencing crises. Machine learning is being used by neuroscientists and therapists all around the world to widen treatment regimens for patients and to identify some of the key signs for mental health issues before they manifest. One of the benefits is that device learning helps practitioners to predict who might be at risk of a specific condition. For this study, statistics were gathered from working humans, and the dataset was ran through a few machine mastering algorithms, which included all forms of queries for depressed identification. When compared to DNN and Logistic Regression, the Random Forest algorithm delivers the best accuracy of 81.02% after applying a few algorithms to the data set.

The automated system is now created with excellent accuracy to detect abnormalities in X-ray images. To enhance the appearance of medical photographs, image pre-processing methods are applied, so that high accuracy can be achieved with constrained means. Images are often classified based on their textural properties, which are measured using the Gray Level Co-occurrence Matrix (GLCM). The grey level correlation matrix provides statistical information of the second order on the grey levels of neighboring pixels in a picture (GLCM). In this proposed paper, medical X-ray images are classified and their features are extracted using an ensemble learning model. By extracting image features using the GLCM feature extraction method, this proposed model is able to distinguish between healthy and sick images (Gray level co-occurrence matrix).to improve the efficiency of the Ensemble learning classification method, it is compared against various algorithms using performance indicators, including Logistic regression, Gaussian Naive Bayes, as well as Random Forest. When this approach is compared to existing methods, the proposed ensemble model has an accuracy rate of 97% in classifying normal and diseased images.

The classification accuracy of the high dimensional dataset degrades due to the redundant and irrelevant features. Feature selection (FS) is used to reduce the dimensionality of the dataset by removing the noisy features. Each filter has its statistical approach. So the feature selected by a single filter may ignore the important one. We have presented a multifilter (MF) wrapper hybrid model. The advantage of using the MF method is to select the important feature by one filter which one may ignore by the other. Here, we have used an aggregator approach to combine the most efficacious features among the four individual filter methods (information gain (IG), chi-square (Chi-sq), minimum redundancy maximum relevance (mRMR), and relief). The accuracy assessment is carried out in a multiple filter wrapper (Jaya-SVM, GA-SVM, PSO-SVM, and FA-SVM). The evaluation and prediction of the subset of features are carried out with four classifiers with excellent performance, such as the support vector machine (SVM), Naive Bayes (NB), decision tree (DT), and linear discriminant analysis (LDA) were tested respectively. Four (breast cancer, leukemia, ovarian, and central nervous system (CNS)) cancer datasets are used to implement the model. The performance of the MF wrapper is excellent in comparison to a single filter. According to the findings of this study, the proposed hybrid approach is a more efficient and trustworthy feature selection technique for selecting highly discriminative features.

A vast variety of patient data has been collected and monitored through Electronic Health Records (EHR) using various tools in the healthcare. The objective of the paper is to start data acquisition and data understanding and then create a web interface for data exploration and segmentation and classification. In the data modeling phase, the objective is to create machine learning models for segmentation and classification. The first step is data acquisition from the MIMIC-III v1.4 (Clinical database) data mart. In the data understanding phase, the relationship of multiple tables is evaluated. After data wrangling the combined dataset is then used for k-means clustering techniques for obtaining chest heart failure patients clusters. In the following phase, the diagnosis text data is used for data modeling and for that various text features are created and then multiple classification techniques are applied for predicting the occurrences of death and the best model is considered for the model deployment. In the model evaluation phase, it is observed that six clusters were optimal while training the model and it is incorporated into the application for predicting the segments of the patients based on the risk levels. Few machine learning models were trained on patient’s historic diagnosis text data and the logistic regression model indicated 89% of AUC score in test data and is deployed into the application for the prediction.

There has been an unprecedented rise in the cases of skin diseases since past few decades owing to several factors. Among several skin diseases, skin cancer has also taken a steep rise and resultantly it becomes imperative to devise an efficient model to detect skin cancer. The requirement for automatic detection of skin cancer further grows owing to rise in rate of melanoma skin cancer, its expensive treatment, and its high fatality rate. Treatment of cancer cells frequently necessitates patience and manual inspection. Here, in this work authors propose an image processing and machine learning approach for skin cancer detection. It also uses a feature extraction technique to retrieve the features of the injured skin cells. The proposed model uses convolutional neural network classifier to stratify the extracted data. During the experimental evaluation, it is observed that the proposed system yields an accuracy of 77.03% and a training accuracy of 80% for the datasets available in public domain.

Image of similar object can be taken by using several modalities at same/different time and in various environmental conditions. The human perception can extract same generalized information from captured images through different modalities. It is a need in clinical analysis to scientist, radiologist or practitioner to get right and exact information about captured image of different human body organ. Multimodal medical image fusion (MMIF) technique improves the resolution with minimum redundancy. This paper demonstrated a detail inclusive survey of multimodal image (MI) fusion (MIF) techniques from spatial domain to transform domain with different algorithm in biomedical field. In biomedical field diverse image fusion (IF) applications have been described. Each MIF techniques are evaluated based on output fused image quality by considering evaluation metrics. At last total review conclusion will be stated which will leave new research plan in the era of MIF.

This paper discusses the application of machine learning in the healthcare sector for the prediction of heart disease. Because technology is a valuable tool in the healthcare industry, we intend to discuss the development of a machine learning model that measures many health-related characteristics in this study. The algorithm described in this research might identify whether a person is at risk of developing chronic heart disease during the following ten years. After balancing the unbalanced dataset and feature selection, the accuracy attained was 83–84% using various models such as Logistic Regression, Random Forest Classifiers, and Linear Discriminant Analysis. The paper focuses on analyzing a varied and diverse dataset whereas other papers referenced and cited have drawbacks such as the size of the dataset being too small or geographically limited. These anomalies have been kept in consideration while working on this paper.

The aim of this paper is to educate people about technology advancement and allow them to use this technology for a healthy lifestyle along with balancing environmental sustainability, and social and economic needs, letting prosperity for now and future generations. The proposed system “Obesity Healthcare System (OHS)” ensures healthy lives and promotes well-being for all ages. This aim is essential to sustainable development, it is the third goal of the Sustainable Development Goals (SDGs). OHS will include all the knowledge-based information that will guide the user regarding his/her health, and medications and advise them accordingly. The proposed mobile application Obesity Healthcare System (OHS) is designed in Android Studio using the Dart programing language to provide the users with the necessary tools to monitor their current health by checking and updating their health trackers such as weight and BMI, and if any of these trackers reaches abnormal levels, the user will be alerted and given recommendations on how to improve the condition of their trackers. In addition, it’s the ability to set hospital appointments manually, where the user is notified of their appointments. This system is integrated into several hospitals and provides users with access to their medical records. OHS allows the user to enter symptoms they are suffering from and check the stored list of diseases and provide advice accordingly. This system is built using the Flutter framework, tested, and run on an emulator Android SDK build for ×86 (mobile). OHS is a novel system as a digital health approach to healthcare. It aims to BMI calculator, health tracker, symptoms checker, reminders, and access to medical records that would be a full-pack package to help obese people with their weight and health. With the digital health OHS platform, people have quicker access to health services, improving the quality of care provided to them. Simultaneously, this system reduces burdens on healthcare facilities by pioneering the idea of self-care.

Since the onset of Covid-19 there has been a tremendous number of research to help deal better with the disease. This has produced a huge pile of data. The produced data might be contrary to each other and it is a tough task finding a specific data among the huge landscape of available data sources. Also answering a question that requires relating some research papers needs getting through piles of data and reading lots of literature. Knowledge graphs are very helpful tools to handle and structure all sort of data that is being produced every day. It also helps to find the uncovered relations between data. In this paper we reviewed the knowledge graphs that has dealt with Covid-19. Covid-19 has various aspects and each Knowledge graph normally deals with a specific domain of Covid-19 like drug repurposing, drug-drug interaction, etc. As they have different domains, their data sources also differs. In this paper an analysis over the knowledge graph data sources has been done. Mostly knowledge graphs have considered biomedical and medical aspects of Covid-19 and there has been rare knowledge graphs dealing with societal aspect and almost no work on economical and climate change aspects of the disease.

Skin cancer is a common disease that affects mankind significantly every year there are more new cases of skin cancer than the combined incidence of cancers of the breast, prostate, lung, and colon. With over 5,000,000 new cases every year skin cancer is a concerning public health predicament. Melanoma and non-melanoma are the two main kinds of skin cancer, respectively. Melanoma is a malignant tumor. The 19th most common malignancy in both men and women is melanoma. The deadliest types of skin cancer are melanoma, which can spread quickly. The crucial factor in Melanoma cancer treatment is early diagnosis. Doctors usually prefer the biopsy method for skin cancer detection. During a biopsy, a sample from a suspected skin lesion is removed for medical examination to determine if it is cancerous or not a biopsy is a painful, slow, and time-consuming method. This study proposes an end-to-end decision-based system classifiers for example like neural networks. Convolutional Neural Networks (CNN) will be used to classify melanoma or benign. CNN architectures are appropriate classifiers to distinguish between the images of moles on the skin. This study has used images from both clinical and dermoscopic images Med-node and ISIC. The procedure advised in Melanoma detection shall capture images and preprocess. Segment the acquired preprocessed image and extract the desired feature and classify them as Melanoma or benign. The model has given an accuracy of 94%, and Sensitivity and Specificity are at 0.87 and 0.89 respectively.

The use of IoT (Internet of Things) devices such as echo devices, smart locks, hue lights amongst a few, in our daily lives, has increased widely in this era of digitalization. People are gradually becoming dependent on these devices for their work or to store confidential data. This has also led to the concerns of security that arise with the use of these IoT devices. IoT devices are prone to malware attacks because of their dependency on the internet, technical complexity and integration of both hardware and software technology. The use of vulnerabilities in these devices by the cyber criminals is becoming extravagant. Also, the identification and categorization of IoT malware by cybersecurity analysts is further complicated by the diversity of IoT malware and the heterogeneity of IoT platforms. The aim of this paper is to analyze the malwares that are affecting the IoT devices and propose machine learning methodologies to identify these malwares based on various parameters. This paper focused mainly on malwares such as Mirai, Torii, Mushtik and Trojan that have been rampant in IoT devices these days. The models were trained based on algorithms such as SVM, Decision Tree, Naive Bayes, CNN, XG Boosting Classifier and Gradient Boosting Regression. The XG Boosting Classifier model has provided the highest accuracy of 97.4% amongst all other models. Thus, for the dataset used, XG Boosting Classifier is the best classifier that can be used to detect malware traffic in IoT devices.

The rapid growth of internet has led to various security threats. Today, malicious codes has emerged as one of the most serious threats to internet security. Millions of new malicious codes are created every single day. However, with increasing malicious codes, the detection of malicious codes variants with better performance have become a great challenge. In this work, we address the detection of malicious codes using deep as well as machine learning techniques. We proposed a malicious codes detection model with different approaches based on an autoencoder. The paper also, compares different approaches based on their performance. One is an autoencoder based model and the other is PCA based model. After comparing the two models, the experimental results shows that the autoencoder based model have an higher accuracy than the PCA based model. Therefore, the proposed model has better detection performance

Due to the greater flexibility and accessibility of data outsourcing environments such as cloud computing environments, several healthcare organizations have implemented electronic Personal Health Records (PHRs) to ensure that individual patients have such resilience and scalability. It allows users to manage their health information in a safe environment. However, PHRs contain highly sensitive information where security and privacy issues are major concerns. PHR owners should also be able to securely define their own access policies for offsite data. In addition to basic authentication capabilities, existing commercial cloud platforms typically offer symmetric or public key encryption as an optional feature to keep tenants’ data confidential. However, such traditional encryption schemes are not suitable for data outsourcing environments due to the high key management overhead of symmetric encryption and the high maintenance costs of handling multiple copies of ciphertext for public key encryption solutions.The output of this study is to design and development of a secure, fine-grained access control scheme with lightweight updates to outsourced PHR access policies. The proposed scheme is based on Cipher Text Policy Attribute-Based Encryption (CP-ABE) and Proxy Re-Encryption (PRE). Additionally, this study introduces a policy versioning technique that supports full traceability of policy changes using the Elgamal technique and a performance evaluation that demonstrates the efficiency of the proposed scheme.

Blockchain techniques has essential effect in transactions such that it decreases the costs, enhance the trust, and permit decentralized policies for finance network. Blockchain technology allows each participant in the network to have a copy of the transaction ledger where all transactions are stored, and this means that all transactions are verified and viewable by everyone in the decentralized network which offers transparency to all. The Decentralized Finance Transactional Networks (DFTNs) authorizes Peer-to-Peer (P2P) transactions without asking any consent from a third party to accomplish the transaction process. The P2P transactions still need to have a method of maintaining a record of transactions and that is where blockchain technology is introduced. In this paper, the simplest form of the DFTN is implemented using Omnet++ simulation to show how transactions, in form of messages, would be communicated and information would be synced between the network participants. The results showed the ability of the DFTN in implementing the transactions smoothly especially when errors are encountered because the other members would look for the solution while the coder maintains pace in implementing the program logic. Furthermore, the results showed their functionality in avoiding issues with different versions or settings compatibility in the system.

On a global scale, the Web is a platform that allows universal reach via a variety of communication channels. Whenever there is a newly developed multimedia system or a newly designed web site available, it prevents people from losing track of information that has been shared. Moreover, online streaming, which is a billion-dollar industry, offers a unique opportunity to recast viewing of video from a passive activity into a socially engaged activity by transforming it from a passive event into an online network. Our motivation for developing a multimedia system is to create a movie sync system that allows the user to utilize their own experiences. For the development of this synchronization system, we used the web socket protocol, which allows us to create a collaborative control system over the screen. Due to this, a user can use the prebuffer approach for any multimedia content on the web and can use less bandwidth.

Today, Android has become the most popular operating system because of its salient features. As it is an open-source mobile OS, several developers are developing and publishing their android applications. On the other side, attackers are manipulating those applications in the form of malicious software (Malware) by leveraging the application or functional flow of android OS and those malwares create loss or leakage of confidential sensitive information. Though most anti-virus software affords defence against malware attacks, still the attacks are highly possible in the real time adversarial environment. In this paper, the machine learning-based detection method is designed by combining the features of application namely permission and activity which are obtained during the installation of apps. In our design, permissions and activities of each app are extracted making use of Androguard tool. Using this feature combination, malicious apps are classified as either benign or malicious. The advantage of this method is that there is no need for any dynamic analysis. In our experimentation, we used real-world app samples with 500 malware and 500 benign to train the algorithm for better performance. Based on the experimentation results, highest detection rate is attained by Random Forest (RF) with 95% of accuracy and lowest detection rate is obtained by K-Nearest Neighbors (KNN) with 79% of accuracy.

Reinforcement learning (RL), one of the branches of machine learning, enables a system to learn through trial and error. RL helps in solving control and decision-making tasks. Applying Deep Learning to Reinforcement learning has made it much better at solving many problems. Deep Reinforcement learning, a combination of deep learning and Reinforcement Learning is gaining a lot of interest and application in solving real-world problems. Among Deep Reinforcement learning, Deep Q networks emerged as a popular algorithm. While Deep Q networks have been successfully applied to a lot of scenarios, their application is based on the notion that the agent can completely perceive the environment. In real-time applications, this notion has a fallacy as complete observability is a difficult and sometimes impossible endeavor in real-time and the real world, therefore the use of recurrent networks along with Deep Q networks has been suggested for application to partially observable environments. This paper provides a literature review on various deep recurrent Q network applications. The paper first provides a brief introduction to the concept behind Deep Recurrent Q networks, and the various modifications to improve its performance and then proceeds to review its various applications in different fields.

In the banking sector, credit risk is a significant factor. Banking’s main activities include granting loans, credit cards, investments, mortgages, etc. Credit cards are one of the fastest growing financial services offered by banks in recent years. However, as the number of credit card users increases, banks are facing rising credit card failure rates. Therefore, data analytics can provide solutions to address current phenomena and manage credit risk. This document provides a performance evaluation of credit card default prediction. In this work, a prediction model for credit card defaulters was developed utilising a variety of unconnected decision trees. It helps speculate if someone might be a defaulter and helps the bank decide the credit limit for customers.

At the receiver Channel parameters is an important matter in the wireless communication especially with fast fading Rayleigh channel because of the rapid change of the channel envelop. Data message will suffer during transmission, therefore, there should be a good channel estimation to overcome channel effect. In this paper, modified reduced constellation algorithm MRCA is presented as a solution to estimate Rayleigh fading channel with comb-type tones insertion in Quadrature amplitude modulation modulated signals for OFDM system. The MRCA filter aim to reduce number of tones to estimate Rayleigh fading channel and the number of reduced tone approach to 1/90 (one pilot tones for each 90 data bit) to avoid data rate reduction or bandwidth expansion. The performance of the proposed algorithm was superior on reduced constellation algorithm (RCA) algorithm to estimate channel as showed in the simulation results, as well the performance of the MRCA and RCA, it has been compared with different S/N points in addition to measuring BER verses SNR.

This paper focuses on designing a system to provide the patients with supplemental oxygen to prevent hypoxia for and covid patient. The decrease of oxygen in the tissues in patients with heart diseases, lung related-diseases and elderly people result in hypoxia. During this covid-pandemic, number of the patient died due to poormonitor system and mainly due to insufficient of the oxygen. The main goal of this article is to develop a system to deliver oxygen supplement automatically to patients when required. This system is portable, does not prevent the patients from their life style and gives freedom of mobility for patient. This device can be used both on hospitals and also in houses for patients with not so critical condition to be monitored. The system is designed to read the heart rate in the body using a heartbeat sensor. As the heart-rate decreases eventually the SpO2 levels in the organs start diminishing thus leading to hypoxia. Here the Arduino Uno operates a solenoid valve using a relay to regulate the required oxygen supply automatically. It also has a temperature sensor to record the body temperature. The system uses a portable light-weighted oxygen cylinder to deliver the oxygen via nose. The system uses LCD device to display the readings recorded. This system also uses GSM to communicate with others in case of emergency. The system is provided with a buzzer to intimate the patient in case of emergency.

Images are created in a variety of ways in various industries. These images are tough to work with, and as a result, they can’t be used effectively in a variety of fields. In this paper, Image Resolution is improved to carry out the process of generic image enhancement tasks. In this process, the low-resolution image is enhanced so that the high-resolution image is achieved. With the help of Image enhancement, the perception or in other words the process of interpreting information present in images by the human viewers is enhanced and the quality is improved to a large extent. Image resolution augmentation has traditionally been accomplished using a variety of classic image processing approaches. However, these methods are not as robust as they should be in dealing with any form of noise signal associated with the image and unable to handle the problems of Error Control Mechanism, Optimization and some other problems. Therefore, this paper presents a method of image resolution enhancement using Advanced Hybrid Neural Network architecture which brings about significant improvements in the entire process.

Traffic Congestion has been a very big issue due to the large number of vehicles being utilized in cities. It is necessary to control the number of vehicles getting in and out of the crossing such that the city traffic congestion decreases in an efficient way. In this paper, we try to improvise the timings of green signals using fuzzy logic based on the number of vehicles, speed limit, and length of the crossing. We find out the number of vehicles using computer vision dynamically. We validate our system in two different locations in Chennai against the manual green signal duration that is set manually and we prove that our method improves the waiting time significantly.

Rheumatoid joint pain is ongoing severe infection that can influence joints as it were. This condition can harm different organ frameworks like skin, eyes, lungs, heart, and blood vessels. Rheumatoid joint pain, an immune system sickness, happens when the safe framework incidentally goes after its body tissues. Rheumatoid joint inflammation goes after the coating of the joints, causing a difficult expanding that can prompt bone deterioration and joint deformity. Ultrasound imaging is utilized for the early discovery of rheumatoid joint pain in the wrist. RA causes disfigurements in the patient's wrists, causing unbearable delicacy and enlarging in the joints. Clinical pictures, for example, X-beams, Ultrasound Power Dopplers assume a fundamental part in the early recognition of RA on the wrist. In any case, the as of late presented simulated intelligence (computerized reasoning) innovation further improves the capacities of imaging apparatuses. We support the exact determination of clinical experts. Picture catch utilizing simulated intelligence can extraordinarily assist with mechanizing the checking system, accelerate patient finding, and empower early identification and treatment to forestall distortions. Computer-based intelligence can likewise further develop work productivity through precise determination. Furthermore, the computerised platform helps radiologists with chasing after clinical decisions (i.e.) about contamination following and prognosticate. We have in like manner analysed the different segmentation methodologies used in clinical image management.

Cyberattacks are occurring increasingly frequently as cyber science advances and people utilize the internet and other technology on a regular basis. Digital forensics is used to assess malicious evidence found in a network or system and compile it in a fashion that may be used in court. Network forensic analysis is a method for looking through intrusion data received from a networked environment in order to spot suspicious entities. Utilizing intrusion detection systems (IDS), such as Snort and Wireshark, is the initial step in spotting and reporting a network flooding attack.As technology has advanced and its use has significantly expanded, there is a higher likelihood of attacks on computer networks. In order to help with the identification and/or prevention of such assaults, many techniques have been developed. One well-liked technique is the use of network intrusion detection and prevention systems or NIDS. Businesses can choose from a variety of open-source and commercial intrusion detection systems nowadays, but the fundamental problem is still their performance. An intrusion detection system's job is to safeguard a network against risks posed by security experts, hackers, and crackers as well as the possibility of unlawful activities. A network administrator needs to develop their signature and keep up with new attack types because issues might arise when new attacks appear quickly. IDS would monitor network traffic and only compare packets that included signatures from its own signature database or traits of known failed attacks in the past.

Digital communication of any form must provide data confidentiality as the threats are increasing in today’s rapid world. Data privacy and security are crucial factors as data is considered gold in the modern era. The 128-bit Advanced Encryption Standard algorithm, commonly known as AES, has been implemented in several designs, focusing on specific purposes and is used widely. The 256-bit variant uses the same fundamental cipher blocks as the 128-bit version but differs in key size, the key expansion function and the number of cipher rounds. This paper investigates the 256-bit AES algorithm targeted at FPGA-Field Programmable Gate Arrays architectures and compares it with the 128-bit implementation, reporting performance and resource utilization. Also, the security offered is discussed. The security is determined by the complexity of recovering the key using cryptanalytic attacks. Both encryption and decryption processes are handled by this implementation and are tested in Verilog language using the Xilinx Vivado software on the Xilinx Zynq-7000 (xc7z020-clg484-1) FPGA.

In this study the various spectral estimation of the Electrogastrogram Signals are discussed. The Main objective of the study is to identify the gastroparesis condition using various signal processing techniques. Gastroparesis is one of the pathetic condition in which the stomach fails to digest. Gastroparesis is also called as stomach paralysis. Due to damage or dysfunction of vagal nerve, kajal cells at stomach lining temporary stomach paralysis happen. The immediate stimulation can make the pacemaker of the stomach to get activated and back to digestion process. In this study the stomach signals are preprocessed and spectral estimation techniques are applied to study about the signal spectrum characteristics. The determination of dominant frequencies and features extracted from frequency domain gives vivid description about the abnormalities of the gastric signal.

Designers have scaled down feature sizes, decreasing threshold voltage and allowing the integration of on a single chip, the functionality is becoming increasingly sophisticated. Due to the rapid improvement of semiconductor technology and the growing need for battery-powered portable electronics. To increase the number of devices in a concert, three critical factors are required: system speed, small space, and low power consumption. The entire power consumption of integrated devices is determined by leakage current dissipation in particular. In order to minimize power consumption, the leakage current must be lowered. This study article examines and analyses numerous leakage power reduction techniques, including SC-CMOS and Sleepy keeper. Inverter and Full adder reduction approaches are evaluated in terms of static and dynamic power. In this work, a new strategy for reducing leakage power in 90nm technology is suggested. The suggested approaches will be compared to other leakage reduction strategies that have been used in the past.

Multipliers are the utmost commonly used elements in today’s digital electronics. In digital signal processing systems, hardware multiplication is critical for obtaining high data throughput. Based on the increasing applications of electronic devices, various types of multipliers have emerged. Array Multiplier is the most fundamental multiplier of all. The main goal of our project is to create an optimized and fast 4-bit array multiplier. Parallel Array Multipliers are used in DSP applications to do multiplication at high speeds while meeting performance criteria. The simulations are run on the Xilinx 14.7 ISE Design Suite.

Traffic is a major issue in any major city, and Hyderabad is no different. The Hyderabad traffic police commission is doing its best to control the traffic in the country with 2nd largest population in extreme conditions like heat and pollution. To help the traffic commission, who are controlling the traffic every day under the scorching sun, we have built a multi-purpose helmet that can help the traffic police while on duty. The helmet is installed with a camera and raspberry pi. Instead of using a handheld smartphone to take the picture of vehicles, the helmet snaps a high-resolution picture using a button. These pictures are sent to the AWS cloud using a pre-configured raspberry pi. These images are processed in the cloud where information like the number plate, color, and vehicle types is extracted. All these images can now be filtered according to date and time and viewed on a user-friendly interface.

The internet of things’ (IOT) applications are expanding quickly. IOT is widely employed in the medical industry. The use of IOT in interconnected hospitals is covered in this paper. When patients are unable to schedule an appointment or reserve an ambulance from their local hospitals, they will be given one at the next closest hospital or given an ambulance from that hospital with the help of interconnected hospitals. In places with high population densities or where healthcare facilities are far away, interconnected hospitals might be extensively employed. In the case study, we learned how IOT and interconnected hospitals may enable a mobile application to guide patients to the closest hospital and also help them book ambulances from the nearest available hospitals. In a later section of this paper, we discussed the difficulties interconnected hospitals face now and their potential future.

The difficulties faced by the country in the medical system during COVID-19 have increased tremendously. Shortage of resources to patients if generated can help them to come out of the situation and can even save their lives. The proposed system can help to provide primary resources which generates oxygen based on the heartbeat of the patient. In this paper, an oxygen generation ventilation system is designed which assist patients who are suffering from breathing problems. This system generates oxygen based on the heartbeat of a patient where the minimum and maximum values of a heartbeat are taken as threshold values and these values are processed to drive the motor through the relay driver circuit. IoT-based monitoring and alert system will send the status of the patient’s health condition to a doctor at regular intervals. Additionally sensing devices like the temperature sensor and vibration sensors are also used to monitor the patients. Simulation results show that the proposed system will work efficiently and save the life of the patient.

Compared to more traditional social media channels, Facebook and Twitter are far more effective at spreading information. Social media has developed into a great data origin for businesses or researchers to create models to analyse this repository and harvest practical insights for marketing policy for word-of-mouth (WOM) trading. However, the vocabulary used in social media is rather condensed and includes specialised words and symbols. Such brief communications are not well suited for the majority of natural language processing (NLP) techniques, which concentrate on processing formal phrases. In this paper, we suggest a brand-new paradigm for social media sentiment analysis based on deep learning models. We gather information from which we create a dataset. We aim to create a semantic dataset after processing these particular phrases in order to support future study. Future applications will benefit greatly from the retrieved data. Several social media platforms have been crawled to gather the trial data.

A recommender system is one that tries to anticipate or filter preferences based on the user’s selections. Films, music, journalism, publications, scientific papers and items in general all make use of recommender systems. We’re building a recommendation system with Python and Pandas in this model. Utilizing an approach called content-based filtering, which is based on the information provided about the items, the system makes suggestions for movies that are comparable to those that a user has previously enjoyed. The information we know about the items and the user’s previous choices are used to calculate this similarity. Combining collaborative filtering and content-based filtering is used to overcome the shortcomings of these two types of filtering generally so that a better recommendation system may be created.

The accelerated tremendous reach of web applications substantially raises the demand for effective recommender systems to examine and filter required content from the vast quantity of information. Recommender systems have evolved in this digital arena as a way to aid users by giving them possibilities among acceptable and relevant items by analyzing user interests. In this system, the preferences as well as prior behavior patterns of the users, have been utilized to give a recommendation. The utilization of recommendation models became a crucial component in digital marketing strategy. It also plays a vital role in areas such as streaming services (movies, music, and books), social networking systems, e-governance, e-commerce (shopping), e-library, e-learning, tourism, resource services, any group activities and much more. Recently it has been inducted into healthcare, education and a wide variety of user’s needs, to help the users in discovering and fetching related interests. However, the main challenges like cold start, sparsity, grey sheep, starvation, and shilling can degrade the performance of the recommender system. Research on recommender system has raised significantly to make the system overcome the challenges and enhance the accuracy of predictions. This article aims to provide a comprehensive review on the main models, challenges, evaluation methods, and metrics of the recommender system. Also aimed to provide a glimpse of the domains and tools concerning the recommender system. Future prospects were also to explore additional insights, and unresolved concerns in the area of RS to support future researchers.

In today’s world, abundant digital content like e-books, movies, videos and articles are available for consumption. It is daunting to review everything accessible and decide what to watch next. Consequently, digital media providers want to capitalise on this confusion and tackle it to increase user engagement, eventually leading to higher revenues. Content providers often utilise recommendation systems as an efficacious approach for combating such information overload. This paper concentrates on developing a synthetic approach for recommending movies. Traditionally, movie recommendation systems use either collaborative filtering, which utilises user interaction with the media, or content-based filtering, which makes use of the movie’s available metadata. Technological advancements have also introduced a hybrid technique that integrates both systems. However, our approach deals solely with content-based recommendations, further enhancing it with a ranking algorithm based on content similarity metrics. The three metrics contributing to the ranking are similarity in metadata, visual content, and user reviews of the movies. We use text vectorization followed by cosine similarity for metadata, feature extraction by a pre-trained VGG19 followed by K-means clustering for visual content, and a comparison of sentiments for user reviews. Such a system allows viewers to know movies that “feel” the same.

In 2019, people are getting sick from the coronavirus. We can only stay safe from the epidemic if we wear masks and stay away from each other. Airports, hotels, hospitals, and train stations, among other places, require users to wear the Mask and stay away from other people. Manually checking people to see if they follow the mask-and-distance rule is hard because it costs a lot. The COVID-19 Face Mask and Social Distancing Detector System uses machine learning to find face masks and social distances at the same time. It does this by combining high-level contextual features with feature maps and an artificial neural network. IP address and CCTV cameras with computer vision would be used in the technology to identify people without masks or social isolation. This solution keeps things safe even when no one is watching. The technology could help hospitals, offices, schools, building sites, airports, and more. People may be safer if they use our face mask and social distance detecting device.

A product trial is a great way of marketing. It offers a first-hand experience that a customer can use to evaluate the product and decide whether to buy it. However, there are expenses involved in every trial. If the buyer decides not to purchase the product after the trial, this expense gets wasted. The lower the conversion rate, the higher the loss that the company must face. To reduce these losses, a strategy should be developed on which studies should be conducted. One such business case is the subject of this paper. A software development company has launched a new product and given its current customers access to the product's trial version. A loss of $8.5M resulted from 56% of trial customers not buying the product. The business is therefore looking for strategies to reduce this loss. This paper aims to present the work which is done to solve this business problem. The goal is accomplished by employing analytical methods to determine the most likely clients. The data provided by the company is used to build binary classification models. Another way to find the likeliest customer is based on the purchase pattern of the customers. Affinity analysis is done on the data to identify the set of products with which the new product is frequently sold to target the customers who have purchased those products for the sale of the new product. Among the binary classification models that were built, the Random Forest model outperformed other models with an accuracy of 83%. This enables the business to take calculated decisions while extending the trial to the customers. Alternatively, Market Basket Analysis, with an accuracy of 88%, discovered a set of two products, the existing buyers of which are more likely to buy the newly launched product. This information not only helped find the right customers but also paved the path for cross-selling of the new product.

Television is both a source of information and a means of communication, and it plays an important role in everyone's life. It broadcasts news, documentaries, sporting events, and other events, among other things. In the market, different models of televisions having different features are available based on the user requirement. This paper tries to develop a model that can offer a client with a fair pricing estimate based on a tradeoff between features and price. A four-step process is devised for this objective, which includes real-time data scraping from an eCommerce website and creation of a model using machine learning algorithms. The algorithms like Multi Linear Regression, SVM (Regressor), Decision Tree Regressor are used for price prediction. Decision Tree Regression was found to be more accurate in predicting television prices in this study.