Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053505
Anjanee Kumar, T. Das
Any organisation’s network infrastructure is insecure as different cyber-attacks have constantly mounted and destabilised these systems. There is a rapid upsurge in the usage of the internet in the modern era. This extensive use of the internet has given a chance to attackers to do malicious activities on the network field. To combat these attacks, we need an Intrusion Detection System (IDS). IDS is a robust technological system that protects the system by detecting any intrusions in it. In this study, different machine learning algorithms, which include Support Vector Machine (SVM), Naive Bayes, Random Forest (RF), and Decision Tree (DT), are compared with the method of Logical Analysis of Data (LAD) on NSL-KDD dataset. NSL-KDD is the benchmark dataset used in the network field. The results have been compared on the basis of accuracy, recall, F1-score, G-mean, detection time and ROC-AUC curve. Based on the result obtained, it is evident that the LAD method has outperformed in comparison with other ML-based methods and also detects intrusions in real time.
{"title":"Rule-based Intrusion Detection System using Logical Analysis of Data","authors":"Anjanee Kumar, T. Das","doi":"10.1109/IDCIoT56793.2023.10053505","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053505","url":null,"abstract":"Any organisation’s network infrastructure is insecure as different cyber-attacks have constantly mounted and destabilised these systems. There is a rapid upsurge in the usage of the internet in the modern era. This extensive use of the internet has given a chance to attackers to do malicious activities on the network field. To combat these attacks, we need an Intrusion Detection System (IDS). IDS is a robust technological system that protects the system by detecting any intrusions in it. In this study, different machine learning algorithms, which include Support Vector Machine (SVM), Naive Bayes, Random Forest (RF), and Decision Tree (DT), are compared with the method of Logical Analysis of Data (LAD) on NSL-KDD dataset. NSL-KDD is the benchmark dataset used in the network field. The results have been compared on the basis of accuracy, recall, F1-score, G-mean, detection time and ROC-AUC curve. Based on the result obtained, it is evident that the LAD method has outperformed in comparison with other ML-based methods and also detects intrusions in real time.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"1 1","pages":"129-135"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77555435","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053391
Mahima R, M. M, Manjari K, Rovenal S, K. S, Sruthi M. P
Traffic-related injuries and deaths are a serious problem that all industrialized nations are dealing with today. Object recognition techniques are employed in this study to develop a low cost and simple solution for automated detection and tracking on highways in order to avoid animal-vehicle collisions. In real-world units, a technique for measuring the animal distances from the camera mounted vehicle is also developed. Wild animal monitoring in their natural settings must be efficient and trustworthy in order to update manage decisions. Because of their effectiveness and accuracy in capturing wildlife data in an inconspicuous, continuous, and massive volume, automatic covert camera traps or cameras are becoming extremely popular as a tool for monitoring wildlife. Hand-taking a massive number of photos and films from camera setups is very costly and tedious. It is a significant barrier for researchers and environmental scientists who want to observe wildlife in a natural setting. This research presents a structure for developing automated animal detection in the wild, with the goal of creating an automated wildlife monitoring system, based on current breakthroughs in deep learning methods. In aspects of recognition, the suggested method attains a total precision of about 85.51 percent.
{"title":"Highway Collision Avoidance by Detection of Animal’s Images","authors":"Mahima R, M. M, Manjari K, Rovenal S, K. S, Sruthi M. P","doi":"10.1109/IDCIoT56793.2023.10053391","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053391","url":null,"abstract":"Traffic-related injuries and deaths are a serious problem that all industrialized nations are dealing with today. Object recognition techniques are employed in this study to develop a low cost and simple solution for automated detection and tracking on highways in order to avoid animal-vehicle collisions. In real-world units, a technique for measuring the animal distances from the camera mounted vehicle is also developed. Wild animal monitoring in their natural settings must be efficient and trustworthy in order to update manage decisions. Because of their effectiveness and accuracy in capturing wildlife data in an inconspicuous, continuous, and massive volume, automatic covert camera traps or cameras are becoming extremely popular as a tool for monitoring wildlife. Hand-taking a massive number of photos and films from camera setups is very costly and tedious. It is a significant barrier for researchers and environmental scientists who want to observe wildlife in a natural setting. This research presents a structure for developing automated animal detection in the wild, with the goal of creating an automated wildlife monitoring system, based on current breakthroughs in deep learning methods. In aspects of recognition, the suggested method attains a total precision of about 85.51 percent.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"11 1","pages":"307-311"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82821596","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053540
S. Kuwelkar, H. G. Virani
In near future, IoT will revolutionize human lifestyle. IoT is categorized as low power lossy network since it employs devices with constrained power, memory and processing capability which are interconnected over lossy links. The efficiency of such networks largely depends on the design of the routing protocol. To cater specific routing needs of such networks, the IETF has proposed IPv6 routing protocol for LLNs (RPL) as a de facto routing standard. In RPL, routing decision is based on a single parameter which leads to the selection of inefficient paths and affects network lifetime. This work primarily focuses on improving the RPL protocol by overcoming the single metric limitation. In this work, a novel version of RPL is proposed which uses a Multilayer Feed Forward Neural Network to make the routing decision based on multiple metrics. Four routing parameters namely, hop count, delay, residual energy and link quality of candidate neighbors are fed as input to ANN in order to compute the fitness of each candidate and the one with highest value is designated as the most suitable parent to route packets towards sink node. This technique lowers energy consumption by 15%, improves Packet Delivery Ratio by 3%, lowers delay by 17% and reduces the control overhead by 48% as compared to standard RPL implementation.
{"title":"RPL Protocol Enhancement using Artificial Neural Network (ANN) for IoT Applications","authors":"S. Kuwelkar, H. G. Virani","doi":"10.1109/IDCIoT56793.2023.10053540","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053540","url":null,"abstract":"In near future, IoT will revolutionize human lifestyle. IoT is categorized as low power lossy network since it employs devices with constrained power, memory and processing capability which are interconnected over lossy links. The efficiency of such networks largely depends on the design of the routing protocol. To cater specific routing needs of such networks, the IETF has proposed IPv6 routing protocol for LLNs (RPL) as a de facto routing standard. In RPL, routing decision is based on a single parameter which leads to the selection of inefficient paths and affects network lifetime. This work primarily focuses on improving the RPL protocol by overcoming the single metric limitation. In this work, a novel version of RPL is proposed which uses a Multilayer Feed Forward Neural Network to make the routing decision based on multiple metrics. Four routing parameters namely, hop count, delay, residual energy and link quality of candidate neighbors are fed as input to ANN in order to compute the fitness of each candidate and the one with highest value is designated as the most suitable parent to route packets towards sink node. This technique lowers energy consumption by 15%, improves Packet Delivery Ratio by 3%, lowers delay by 17% and reduces the control overhead by 48% as compared to standard RPL implementation.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"21 1","pages":"52-58"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89071402","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053483
K. Laxminarayanamma, R. Krishnaiah, P. Sammulal
Cancer disease prediction based on neurological characteristics of cancer patients is gaining a significant research attention in recent times. The role of data in the processing and analysis of neurological features is critical, and the main goal is to efficiently extract neurological features from cancer patients' data. Random extraction of neurological features from cancer patient data is a new research initiative. Convolutional Neural Networks (CNN) is a promising approach in various healthcare applications to efficiently perform the data processing tasks. Some CNN-based approaches have been proposed to perform efficient cancer disease prediction using remotely sensed neurological features. Cancer disease extraction based on MPDCNN is one of the best CNN approaches used for extracting features and perform disease prediction from Geo-Fan-2 (GF-2) sensing cancer patient data. However, due to its sparse arrangement of optimal boundary, exact neurological features and high amount of training time, it is insufficient to investigate and automate the neurological feature extraction process from the cancer patient's data. A Novel Optimized Multi Feature Contour based Hierarchical Neural Network (NOMFCHNN) is proposed to improve the automatic neurological feature prediction process. NOMFCHNN is made up of expanding neural network features and layers related to inception, which contains the data about network localization, and this approach uses optimal and exact neurological feature matching with extended feature extraction. This method also employs contour map optimization to identify contours based on globalization of cancer patient data along with the output of the identified contour being transmitted to the next identified contour in the selected hierarchical region. Furthermore, the proposed approach evaluates the low- resolution term in cancer patient's data to gain knowledge from the cancer patient's data by obtaining the prediction results of neighbouring optimal and exact neurological features to eliminate small changes or errors. A multi scale feature Prediction module is used to eliminate feature inconsistency between the encoding and decoding phases of the prediction process in order to identify better contours of neurological features from remote sensing cancer patient's data. Extensive experiments on combined repository cancer patient data show that the proposed methodology improves the prediction accuracy and other parameters when compared to the other state-of-the-art methods used to remotely analyze the neurological features.
{"title":"Advanced Optimized Counter based Hierarchal Model to Predict Cancer’s Disease from Cancer Patients Neurological Features","authors":"K. Laxminarayanamma, R. Krishnaiah, P. Sammulal","doi":"10.1109/IDCIoT56793.2023.10053483","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053483","url":null,"abstract":"Cancer disease prediction based on neurological characteristics of cancer patients is gaining a significant research attention in recent times. The role of data in the processing and analysis of neurological features is critical, and the main goal is to efficiently extract neurological features from cancer patients' data. Random extraction of neurological features from cancer patient data is a new research initiative. Convolutional Neural Networks (CNN) is a promising approach in various healthcare applications to efficiently perform the data processing tasks. Some CNN-based approaches have been proposed to perform efficient cancer disease prediction using remotely sensed neurological features. Cancer disease extraction based on MPDCNN is one of the best CNN approaches used for extracting features and perform disease prediction from Geo-Fan-2 (GF-2) sensing cancer patient data. However, due to its sparse arrangement of optimal boundary, exact neurological features and high amount of training time, it is insufficient to investigate and automate the neurological feature extraction process from the cancer patient's data. A Novel Optimized Multi Feature Contour based Hierarchical Neural Network (NOMFCHNN) is proposed to improve the automatic neurological feature prediction process. NOMFCHNN is made up of expanding neural network features and layers related to inception, which contains the data about network localization, and this approach uses optimal and exact neurological feature matching with extended feature extraction. This method also employs contour map optimization to identify contours based on globalization of cancer patient data along with the output of the identified contour being transmitted to the next identified contour in the selected hierarchical region. Furthermore, the proposed approach evaluates the low- resolution term in cancer patient's data to gain knowledge from the cancer patient's data by obtaining the prediction results of neighbouring optimal and exact neurological features to eliminate small changes or errors. A multi scale feature Prediction module is used to eliminate feature inconsistency between the encoding and decoding phases of the prediction process in order to identify better contours of neurological features from remote sensing cancer patient's data. Extensive experiments on combined repository cancer patient data show that the proposed methodology improves the prediction accuracy and other parameters when compared to the other state-of-the-art methods used to remotely analyze the neurological features.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"24 1","pages":"613-624"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89093475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053442
Chereddy Spandana, Ippatapu Venkata Srisurya, S. Aasha Nandhini, R. P. Kumar, G. Bharathi Mohan, Parathasarathy Srinivasan
In recent years, AutoML is booming as the time-consuming and iterative tasks involved in developing a machine learning model can be automated using AutoML. It aims to lessen the requirement for skilled individuals to create the ML model. Additionally, it helps to increase productivity and advance machine learning research. Hence, this paper focusses on developing an AutoML model using genetic algorithm to automatically fulfill the function of network architecture search. The proposed methodology has been evaluated in different scenarios such as binary classification and regression. From the results it is observed that the accuracy achieved for binary classification and regression is 98%.
{"title":"An Efficient Genetic Algorithm based Auto ML Approach for Classification and Regression","authors":"Chereddy Spandana, Ippatapu Venkata Srisurya, S. Aasha Nandhini, R. P. Kumar, G. Bharathi Mohan, Parathasarathy Srinivasan","doi":"10.1109/IDCIoT56793.2023.10053442","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053442","url":null,"abstract":"In recent years, AutoML is booming as the time-consuming and iterative tasks involved in developing a machine learning model can be automated using AutoML. It aims to lessen the requirement for skilled individuals to create the ML model. Additionally, it helps to increase productivity and advance machine learning research. Hence, this paper focusses on developing an AutoML model using genetic algorithm to automatically fulfill the function of network architecture search. The proposed methodology has been evaluated in different scenarios such as binary classification and regression. From the results it is observed that the accuracy achieved for binary classification and regression is 98%.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"77 1","pages":"371-376"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89618350","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053539
M. Manvitha, M. Vani Pujitha, N. Prasad, B. Yashitha Anju
1.80 metric tonnes of CO2 are emitted by citizens in India, which is highly detrimental to all living beings. Climate change and glacier melting are the results of CO2 emissions. Sea levels are rising as a result of global warming, which is mostly caused by CO2. In the past, the prediction has been accomplished using statistical approaches including the t-test, ANOVA test, ARIMA, and SARIMAX. The Random Forest, Decision Tree, and Regression Models are increasingly used to forecast CO2 emissions. When several vehicle feature inputs are used, multivariate polynomial regression and multiple linear regression may reliably forecast the emissions. For inputs with a single feature, single linear regression is used for the prediction. Based on factors including engine size, fuel type, cylinder count, vehicle class, and model, CO2 emissions are anticipated. Python Scikit-Learn and the Matplotlib package are used to analyze CO2 emissions. The efficiency of the implemented models is assessed by using performance metrics. The accuracy of each model is predicted by using the Regression Score (R2-Score), MAE (Mean Absolute Error), and MSE (Mean Squared Error).
{"title":"A Predictive Analysis on CO2 Emissions in Automobiles using Machine Learning Techniques","authors":"M. Manvitha, M. Vani Pujitha, N. Prasad, B. Yashitha Anju","doi":"10.1109/IDCIoT56793.2023.10053539","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053539","url":null,"abstract":"1.80 metric tonnes of CO2 are emitted by citizens in India, which is highly detrimental to all living beings. Climate change and glacier melting are the results of CO2 emissions. Sea levels are rising as a result of global warming, which is mostly caused by CO2. In the past, the prediction has been accomplished using statistical approaches including the t-test, ANOVA test, ARIMA, and SARIMAX. The Random Forest, Decision Tree, and Regression Models are increasingly used to forecast CO2 emissions. When several vehicle feature inputs are used, multivariate polynomial regression and multiple linear regression may reliably forecast the emissions. For inputs with a single feature, single linear regression is used for the prediction. Based on factors including engine size, fuel type, cylinder count, vehicle class, and model, CO2 emissions are anticipated. Python Scikit-Learn and the Matplotlib package are used to analyze CO2 emissions. The efficiency of the implemented models is assessed by using performance metrics. The accuracy of each model is predicted by using the Regression Score (R2-Score), MAE (Mean Absolute Error), and MSE (Mean Squared Error).","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"31 1","pages":"394-401"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86413921","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053472
Chinnam Sasidhar Reddy, N. S. K. Deepak Rao, Atkuri Sisir, Vysyaraju Shanmukha Srinivasa Raju, S. S. Aravinth
E-commerce systems have grown in popularity and are now used in almost every business. A platform for online product marketing and customer promotion is an e-commerce system. Customer clustering is defined as the process of categorizing consumers into sections that share resembling characteristics. To maximize each customer's profit to the business, customer clustering’s goal is to help decide how to engage clients in each category. To facilitate customer needs by improvising products and optimizing services, businesses can identify their most profitable customers by segmenting their customer base. As a result, customer clustering assists E-commerce systems in promoting the appropriate product to the appropriate customer to increase profits. Customer clustering factors include geographic, psychological, behavioral, and demographic factors. The consumer’s behavioral factor has been highlighted in this research. As a result, to discover the consumption behavior of the E-shopping system, customers will be analyzed using several clustering algorithms. Clustering seeks to maximize experimental similarity within a cluster while minimizing dissimilarity between clusters. Customers’ age, gender, income, expenditure rate, etc. are correlated in this study. To assist vendors in identifying and concentrating on the most profitable segments of the market as opposed to the least profitable segments, this study compared several clustering techniques to find which technique is more accurate to cluster customer behavior. A significant role for this kind of analysis in business improvement to keep customers for a long time and boost business profits, businesses group their customers based on similar behavioral traits. It also enables the maximum disclosure of online offers to attract the attention of potential customers. A learning algorithm called K-Means and an unsupervised algorithm hierarchical clustering is applied to a customer dataset to compare which strategy gives most accurate clustering.
{"title":"A Comparative Survey on K-Means and Hierarchical Clustering in E-Commerce Systems","authors":"Chinnam Sasidhar Reddy, N. S. K. Deepak Rao, Atkuri Sisir, Vysyaraju Shanmukha Srinivasa Raju, S. S. Aravinth","doi":"10.1109/IDCIoT56793.2023.10053472","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053472","url":null,"abstract":"E-commerce systems have grown in popularity and are now used in almost every business. A platform for online product marketing and customer promotion is an e-commerce system. Customer clustering is defined as the process of categorizing consumers into sections that share resembling characteristics. To maximize each customer's profit to the business, customer clustering’s goal is to help decide how to engage clients in each category. To facilitate customer needs by improvising products and optimizing services, businesses can identify their most profitable customers by segmenting their customer base. As a result, customer clustering assists E-commerce systems in promoting the appropriate product to the appropriate customer to increase profits. Customer clustering factors include geographic, psychological, behavioral, and demographic factors. The consumer’s behavioral factor has been highlighted in this research. As a result, to discover the consumption behavior of the E-shopping system, customers will be analyzed using several clustering algorithms. Clustering seeks to maximize experimental similarity within a cluster while minimizing dissimilarity between clusters. Customers’ age, gender, income, expenditure rate, etc. are correlated in this study. To assist vendors in identifying and concentrating on the most profitable segments of the market as opposed to the least profitable segments, this study compared several clustering techniques to find which technique is more accurate to cluster customer behavior. A significant role for this kind of analysis in business improvement to keep customers for a long time and boost business profits, businesses group their customers based on similar behavioral traits. It also enables the maximum disclosure of online offers to attract the attention of potential customers. A learning algorithm called K-Means and an unsupervised algorithm hierarchical clustering is applied to a customer dataset to compare which strategy gives most accurate clustering.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"36 1","pages":"805-811"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86683493","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053418
Shahnawaz Ayoub, N. Behera, Meena Naga Raju, Pankaj Singh, S. Praveena, R. K.
Medical image classifiers roles a crucial play in medical service and teaching tasks. But the classical approach obtained its ceiling on performance. Besides, from their use, much longer and more effort require spent on extracted and selected classifier features. The Deep Neural Network (DNN) is a developing Machine Learning (ML) approach which is verified their potential for distinct classifier tasks. Especially, the Convolutional Neural Network (CNN) leads to optimum outcomes on distinct image classifier tasks. But medical image databases can be hard for collecting as it requires several professional skills to categorize them. This study develops a new Hyperparameter Tuned Deep Learning Model for Healthcare Monitoring Systems (HPTDLM-HMS) in big data environment. The presented HPTDLM-HMS technique concentrates on the examination of medical images in the decision-making process. Initially, the presented HPTDLM-HMS technique derives features using EfficientNet model with Manta Ray Foraging Optimization (MRFO) algorithm as hyperparameter tuner. At last, the classification of medical images takes place by Long Short-Term Memory (LSTM) method. To handle big data, Hadoop MapReduce is utilized. The result analysis of the HPTDLM-HMS technique is tested on medical imaging dataset. The comprehensive study of the HPTDLM-HMS technique highlighted and gives recall value of 87.46% is higher when compared to its promising outcomes over other models.
{"title":"Hyperparameter Tuned Deep Learning Model for Healthcare Monitoring System in Big Data","authors":"Shahnawaz Ayoub, N. Behera, Meena Naga Raju, Pankaj Singh, S. Praveena, R. K.","doi":"10.1109/IDCIoT56793.2023.10053418","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053418","url":null,"abstract":"Medical image classifiers roles a crucial play in medical service and teaching tasks. But the classical approach obtained its ceiling on performance. Besides, from their use, much longer and more effort require spent on extracted and selected classifier features. The Deep Neural Network (DNN) is a developing Machine Learning (ML) approach which is verified their potential for distinct classifier tasks. Especially, the Convolutional Neural Network (CNN) leads to optimum outcomes on distinct image classifier tasks. But medical image databases can be hard for collecting as it requires several professional skills to categorize them. This study develops a new Hyperparameter Tuned Deep Learning Model for Healthcare Monitoring Systems (HPTDLM-HMS) in big data environment. The presented HPTDLM-HMS technique concentrates on the examination of medical images in the decision-making process. Initially, the presented HPTDLM-HMS technique derives features using EfficientNet model with Manta Ray Foraging Optimization (MRFO) algorithm as hyperparameter tuner. At last, the classification of medical images takes place by Long Short-Term Memory (LSTM) method. To handle big data, Hadoop MapReduce is utilized. The result analysis of the HPTDLM-HMS technique is tested on medical imaging dataset. The comprehensive study of the HPTDLM-HMS technique highlighted and gives recall value of 87.46% is higher when compared to its promising outcomes over other models.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"13 1","pages":"281-287"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86761580","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053429
C. M. Raj, W. R. Babu, Karparthi Uday, R. Senthilkumar, V. Sudha, V. Anandhakumar
In logistics services, measuring of volumetric weight is done manually. In order to reduce manpower, automation is required. Here we use Programmable Logic Controller (PLC) and RV-2FB-Q Series robotic arm for automation. PLC system helps in operating pick & place robotic arm, processing the data acquired through ultrasonic sensors, Light Detecting Resistor, load cell and also controlling (ON/OFF) PMDC motor attached with conveyor belt based on inductive proximity sensor signal. The charges for customers are calculated based on different volumetric weights. Using HMI (Human Machine Interface) a bill is projected to customers. In the end, the right amount for proper volumetric weight can be calculated and collected from the customers. Furthermore, with the use of PLC, the logistics management can be monitored and being connected to other hardware application which improve the operation of parcelling. Hence, customer reliability can be improved and also computation of volumetric weight can be done precisely. Finally, our project leads to achieve industry 4.0
{"title":"Smart Dimensional Measurement and Material Transportation (SDMMT) System using Artificial Intelligence","authors":"C. M. Raj, W. R. Babu, Karparthi Uday, R. Senthilkumar, V. Sudha, V. Anandhakumar","doi":"10.1109/IDCIoT56793.2023.10053429","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053429","url":null,"abstract":"In logistics services, measuring of volumetric weight is done manually. In order to reduce manpower, automation is required. Here we use Programmable Logic Controller (PLC) and RV-2FB-Q Series robotic arm for automation. PLC system helps in operating pick & place robotic arm, processing the data acquired through ultrasonic sensors, Light Detecting Resistor, load cell and also controlling (ON/OFF) PMDC motor attached with conveyor belt based on inductive proximity sensor signal. The charges for customers are calculated based on different volumetric weights. Using HMI (Human Machine Interface) a bill is projected to customers. In the end, the right amount for proper volumetric weight can be calculated and collected from the customers. Furthermore, with the use of PLC, the logistics management can be monitored and being connected to other hardware application which improve the operation of parcelling. Hence, customer reliability can be improved and also computation of volumetric weight can be done precisely. Finally, our project leads to achieve industry 4.0","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"29 1","pages":"493-497"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81059712","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-05DOI: 10.1109/IDCIoT56793.2023.10053547
Swaminathan K, V. Ravindran, R. Ponraj, S. Venkatasubramanian, K. Chandrasekaran, S. Ragunathan
Modern wireless technology demands the implementation of preset Sensor nodes for a structured wireless network. The network has sensor nodes for surveillance or environmental sensing, which wirelessly transmit data to a collection point. Therefore, data transfer must be protected by preventing external intrusion attacks. This will be handled by designing an effective intrusion detection system proposed as a Composite Intrusion detection system (CIDS). It is suitable for a network in heterogeneous network structure with a capable of identifying externals attacks like flooding of data's, sending unwanted data packets and changing the destination node. For routing of data packets between the nodes, minimum power utilization with changeable cluster heading method is used. The activities of sensor nodes will be monitored and a dataset is formed on the basis of the node’s activity. It is known as Network Databases (NDB). Using this dataset, the intrusion attacks will be identified by using Artificial Neural Network (ANN). ANN will be trained with a predefined dataset for the effective identification of external attacks. The proposed CIDS methodology shows the high accuracy of identifying the external attacks on the sensor networks when comparing to the previous designed system in all the types of attacks.
{"title":"A Novel Composite Intrusion Detection System (CIDS) for Wireless Sensor Network","authors":"Swaminathan K, V. Ravindran, R. Ponraj, S. Venkatasubramanian, K. Chandrasekaran, S. Ragunathan","doi":"10.1109/IDCIoT56793.2023.10053547","DOIUrl":"https://doi.org/10.1109/IDCIoT56793.2023.10053547","url":null,"abstract":"Modern wireless technology demands the implementation of preset Sensor nodes for a structured wireless network. The network has sensor nodes for surveillance or environmental sensing, which wirelessly transmit data to a collection point. Therefore, data transfer must be protected by preventing external intrusion attacks. This will be handled by designing an effective intrusion detection system proposed as a Composite Intrusion detection system (CIDS). It is suitable for a network in heterogeneous network structure with a capable of identifying externals attacks like flooding of data's, sending unwanted data packets and changing the destination node. For routing of data packets between the nodes, minimum power utilization with changeable cluster heading method is used. The activities of sensor nodes will be monitored and a dataset is formed on the basis of the node’s activity. It is known as Network Databases (NDB). Using this dataset, the intrusion attacks will be identified by using Artificial Neural Network (ANN). ANN will be trained with a predefined dataset for the effective identification of external attacks. The proposed CIDS methodology shows the high accuracy of identifying the external attacks on the sensor networks when comparing to the previous designed system in all the types of attacks.","PeriodicalId":60583,"journal":{"name":"物联网技术","volume":"R-34 1","pages":"112-117"},"PeriodicalIF":0.0,"publicationDate":"2023-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84560241","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：