Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi)
Not a member yet
1071 research outputs found
Sort by
Implementation of Self-Organizing Map (SOM) Algorithm for Image Classification of Medicinal Weeds
Get PDFWild plants or weeds often become enemies or disturb the main cultivated plants. In its development, wild plants or weeds actually have ingredients that are beneficial to the body and can be used as medicine. However, many people still need knowledge about the types of weed plants that have medicinal properties, especially the leaves. The purpose of this research is to classify the image of weed leaves with medicinal properties based on color and texture characteristics with an artificial neural network using a Self-Organizing Map (SOM). To improve information in feature extraction, RGB and HSV color features are used as well as texture features with Gray Level Co-occurrence Matrix (GLCM). Furthermore, the results of feature extraction will be identified as groups or classes with the Self-Organizing Map (SOM) algorithm which divides the input pattern into several groups so that the network output is in the form of a group that is most similar to the input provided. The test produces a precision value of 91.11%, a recall value of 88.17% and an accuracy value of 89.44%. The results of the accuracy of the SOM model for image classification on medicinal weed leaves are in the good category.
Wild plants or weeds often become enemies or disturb the main cultivated plants. In its development, wild plants or weeds actually have ingredients that are beneficial to the body and can be used as medicine. However, many people still need knowledge about the types of weed plants that have medicinal properties, especially the leaves. The purpose of this research is to classify the image of weed leaves with medicinal properties based on color and texture characteristics with an artificial neural network using a Self-Organizing Map (SOM). To improve information in feature extraction, RGB and HSV color features are used as well as texture features with Gray Level Co-occurrence Matrix (GLCM). Furthermore, the results of feature extraction will be identified as groups or classes with the Self-Organizing Map (SOM) algorithm which divides the input pattern into several groups so that the network output is in the form of a group that is most similar to the input provided. The test produces a precision value of 91.11%, a recall value of 88.17% and an accuracy value of 89.44%. The results of the accuracy of the SOM model for image classification on medicinal weed leaves are in the good category
Real-Time Detection of Face Mask Using Convolutional Neural Network
No full textMasks are a simple barrier that can help us prevent transmission and spread of disease from other people who enter the body, avoid exposure to air pollution, and protect the face from the adverse effects of sunlight. However, many people are still ignorant about the importance of wearing masks for health. This study aims to detect whether or not to use masks in real-time by proposing a deep learning model to reduce illness and death caused by air pollution. The convolutional Neural Network (CNN) method was used in this research to detect facial recognition using a mask and not using a mask. The public dataset used in this research consists of 1300 images with 650 data using masks and 650 data without masks. The results of this study show that the proposed CNN method works well in detecting masked and non-masked faces in real time. The proposed method obtains an accuracy value of 97.5% at epoch 50. Previous research on mask detection using the Eigenface method yielded an accuracy of 88.89%, and another study using the Viola-Jones method yielded an accuracy of 95.5%. It can be concluded that this research can increase the accuracy value of previous studies. So, this research is feasible to be applied to the detection of mask use in real time.
Masks are a simple barrier that can help us prevent transmission and spread of disease from other people who enter the body, avoid exposure to air pollution, and protect the face from the adverse effects of sunlight. However, many people are still ignorant about the importance of wearing masks for health. This study aims to detect whether or not to use masks in real-time by proposing a deep learning model to reduce illness and death caused by air pollution. The convolutional Neural Network (CNN) method was used in this research to detect facial recognition using a mask and not using a mask. The public dataset used in this research consists of 1300 images with 650 data using masks and 650 data without masks. The results of this study show that the proposed CNN method works well in detecting masked and non-masked faces in real time. The proposed method obtains an accuracy value of 97.5% at epoch 50. Previous research on mask detection using the Eigenface method yielded an accuracy of 88.89%, and another study using the Viola-Jones method yielded an accuracy of 95.5%. It can be concluded that this research can increase the accuracy value of previous studies. So, this research is feasible to be applied to the detection of mask use in real time
A Comprehensive Bibliometric Analysis of Deep Learning Techniques for Breast Cancer Segmentation: Trends and Topic Exploration (2019-2023)
Get PDFThe objective of this study is to perform a comprehensive bibliometric analysis of the existing literature on breast cancer segmentation using deep learning techniques. Data for this analysis were obtained from the Web of Science Core Collection (WOS-CC) that spans from 2019 to 2023. The study is based on a comprehensive collection of 985 documents that cover a substantial body of research findings related to the application of deep learning techniques in segmenting breast cancer images. The analysis reveals an annual increase in the number of published works at a rate of 16.69%, indicating a consistent and robust increase in research efforts during the specified time frame. Examining the occurrence of keywords from 2019 to 2023, it is evident that the term "convolutional neural network" exhibited a notable frequency, reaching its peak in 2021. However, the term "machine learning" demonstrated the highest overall frequency, peaking around 2021 as well. This emphasizes the importance of machine learning in the advancement of image segmentation algorithms and convolutional neural networks, which have shown exceptional effectiveness in image analysis tasks. Furthermore, the utilization of latent Dirichlet Allocation (LDA) to identify topics resulted in a relatively uniform distribution, with each topic having an equivalent number of abstracts. This indicates that the data set encompasses a diverse range of topics within the field of deep learning as it relates to breast cancer image segmentation. However, it should be noted that topic 4 has the highest level of significance, suggesting that the application of deep learning for diagnosis was extensively explored in this study
Mengembangkan Sistem Pemantauan Detak Jantung Murah untuk perangkat Android
No full textHeart related disease is one of deadly non communicable diseases. In order to detect heart related problems, heart rates must be monitored. Monitoring heart rate can be done with several methods, from stethoscope, to sensors such as PPG and ECG sensors. However, those monitoring can be impractical as it requires doctor visitations. In order to simplify the process, smartwatch is used. Those devices are usually equipped with heart rate monitoring sensors, and many others. However, some previously researched smartwatch-based systems used in monitoring health are cost prohibitive as some requires their own server, high price of smartwatches used, and cumbersome to use. The goal of this research is to build cost effective heart rate monitoring using cheap smartwatches. A proposed system using smartwatch and Android device is used alongside with Whatsapp messaging, with ringtone that plays when emergency situation happened. The message contains GPS location and emergency condition. The resulting system does not require custom made server and nets an acceptable GPS coordinate accuracy outdoors, but not so accurate indoors. It also does not require expensive smartwatches. However, the system requires the device to not be locked due to limitations on Android devices
Integration of Microscopic Image Capturing System for Automatic Detection of Mycobacterium Tuberculosis Bacteria
Get PDFThe Ministry of Health of the Republic of Indonesia is running a program to eliminate Tuberculosis (TB) by 2030. At the Primary Health Care level, AFB (acid-fast bacteria) examination confirms the TB diagnosis. In this process, the patient's sputum is prepared in the form of preparation and observed by the laboratory analyst through the lens of a microscope. The reporting process to establish this diagnosis requires calculating the number of TB bacteria in 100 fields of view per preparation. This manual microscopic observation process is tedious, and the reading results are subjective. This study offers an integrated design for automatic microscopic imaging with a computer-integrated TB bacteria detection system. The process of taking pictures is automatically obtained with the help of a driving motor added to the microscope. With the addition of this motor, the process of taking microscopic images for 100 fields of view takes ±450 seconds. The proposed system integration process can reduce laboratory analysts' work fatigue in conducting microscopic observations manually. The TB bacteria detection system utilizes the working principle of image processing techniques by combining color-deconvolution, segmentation, and contour-detection methods. The comparative value of the TB object detection system with experts resulted in a sensitivity value of 77% and a specificity value of 68%. However, the low detection rate is because the image obtained is still blurry. Thus, further investigation is needed to determine the driving motor's movement rate and the right timing for taking microscopic images so that the resulting image is not blurry. The final result that is the focus of this paper is the successful integration of the system carried out between the motor drive system on the preparation stand and the TB bacteria detection system to become a unified system.The Ministry of Health of the Republic of Indonesia is running a program to eliminate Tuberculosis (TB) by 2030. At the Primary Health Care level, AFB (acid-fast bacteria) examination confirms the TB diagnosis. In this process, the patient's sputum is prepared in the form of preparation and observed by the laboratory analyst through the lens of a microscope. The reporting process to establish this diagnosis requires calculating the number of TB bacteria in 100 fields of view per preparation. This manual microscopic observation process is tedious, and the reading results are subjective. This study offers an integrated design for automatic microscopic imaging with a computer-integrated TB bacteria detection system. The process of taking pictures is automatically obtained with the help of a driving motor added to the microscope. With the addition of this motor, the process of taking microscopic images for 100 fields of view takes ±450 seconds. The proposed system integration process can reduce laboratory analysts' work fatigue in conducting microscopic observations manually. The TB bacteria detection system utilizes the working principle of image processing techniques by combining color-deconvolution, segmentation, and contour-detection methods. The comparative value of the TB object detection system with experts resulted in a sensitivity value of 77% and a specificity value of 68%. However, the low detection rate is because the image obtained is still blurry. Thus, further investigation is needed to determine the driving motor's movement rate and the right timing for taking microscopic images so that the resulting image is not blurry. The final result that is the focus of this paper is the successful integration of the system carried out between the motor drive system on the preparation stand and the TB bacteria detection system to become a unified system
BPNN Optimization With Genetic Algorithm For Classification of Tobacco Leaves With GLCM Extraction Features
Get PDFTobacco leaves are one of the agricultural commodities cultivated by Indonesian farmers. In their application in the field, there are many obstacles in tobacco leaf cultivation, one of which is declining tobacco quality caused by weather factors. In this study, a technology-based analysis step was carried out to determine the classification in determining the quality of tobacco leaves. The research was carried out by applying the classification optimization of the Backpropagation Artificial Neural Network Method and genetic algorithms to determine the weights obtained from extracting GLCM features. You can get the weight value from the genetic algorithm on the homogeneity variable from this analysis step. The variable gets a weight value of 1. The results of this study obtained a classification value with the Backpropagation Artificial Neural Network Method model getting an accuracy value of 53.50% at a hidden layer value of 2,4,5,7. For classification with the Artificial Neural Network Method, Backpropagation, which is optimized with genetic algorithms, you get an accuracy value of 64.50% at the 4th hidden layer value. From this study, the value of optimization accuracy increased by 11% after being optimized with genetic algorithms.
Tobacco leaves are one of the agricultural commodities cultivated by Indonesian farmers. In their application in the field, there are many obstacles in tobacco leaf cultivation, one of which is declining tobacco quality caused by weather factors. In this study, a technology-based analysis step was carried out to determine the classification in determining the quality of tobacco leaves. The research was carried out by applying the classification optimization of the Backpropagation Artificial Neural Network Method and genetic algorithms to determine the weights obtained from extracting GLCM features. You can get the weight value from the genetic algorithm on the homogeneity variable from this analysis step. The variable gets a weight value of 1. The results of this study obtained a classification value with the Backpropagation Artificial Neural Network Method model getting an accuracy value of 53.50% at a hidden layer value of 2,4,5,7. For classification with the Artificial Neural Network Method, Backpropagation, which is optimized with genetic algorithms, you get an accuracy value of 64.50% at the 4th hidden layer value. From this study, the value of optimization accuracy increased by 11% after being optimized with genetic algorithms
Implementation of Enhanced Spray Routing Protocol for VDTN On Surabaya Smart City Scenario
No full textThe application of smart-city, which promises better city management in helping to improve people's quality of life, is still inhibited due to the high cost of infrastructure investment. In several Smart Cities, it takes at least 30 - 40 billion to convert a conventional town into a smart city, including for data collection infrastructure. Alternatively, low-power wide-area networks (LPWANs) could be considered, but they need more bandwidth to serve data transmission in a smart city. Vehicle Delay Tolerant Network (VDTN) is one part of DTN that employs vehicles as a communication infrastructure that allows communication in challenging conditions and could make it an alternative network for Data Collection in a Smart City. This paper proposes a Surabaya Smart City scenario with VDTN as data collection. The scenario consists of 40 wireless sensors and 50 to 200 vehicles (car and bus) with five Road Side Units that forward data from the sensor to the monitoring server. Furthermore, to increase the VDTN performance, we improve our proposed routing protocol, Spray and Hop Distance (SNHD), with two sprays method (Adaptive and Simple) and data collection support from multiple sources and destinations. The evaluation was carried out using a simulation-based comparison with an increase in the number of vehicles to determine the impact of vehicle density on data collection performance in terms of delivery probability, latency average, and overhead ratio. Based on the simulation results, the simple spray method in SNHD and A-SNHD outperformed the well-known VDTN routing protocol, i.e., Epidemic and Spray and Wait. Furthermore, when the number of cars increases from 50 to 200, the performance of VDTN does not increase significantly as the density of the network increases. It means that VDTN only requires a small number of vehicles to be used as a low-cost alternative network for smart cities.
The application of smart-city, which promises better city management in helping to improve people's quality of life, is still inhibited due to the high cost of infrastructure investment. In several Smart cities, it takes at least 30 - 40 billion to convert a conventional town into a smart city, Include for Data collection infrastructure. Alternatively, low-power wide-area networks (LPWAN) could be considered, but it needs more bandwidth to serve data transmission in a smart city. Vehicle Delay Tolerant Network (VDTN) is one part of DTN that employs vehicles as a communication infrastructure that allows communication in challenging conditions and could make it an alternative network for Data Collection in Smart City. This paper proposes a Surabaya Smart City scenario with VDTN as a data collection. The scenario consists of 40 wireless sensors and 50 to 200 vehicles (car and bus) with 5 Road Side Units that forward data from the sensor to the monitoring server. Furthermore, to increase the VDTN performance, we improve our proposed routing protocol, Spray and Hop Distance (SNHD), with two sprays method (Adaptive and Simple) and multiple sources and destinations data collection support. The evaluation was done by simulation-based comparison with an increase in the number of vehicles to determine the impact of vehicle density on data collection performance in terms of delivery probability, Latency Average, and Overhead Ratio. Based on the simulation results, the simple spray method in SNHD and A-SNHD outperformed the well-known VDTN routing protocol, i.e., Epidemic and Spray and Wait. Moreover, when the number of cars is increased from 50 to 200, the performance of VDTN does not increase significantly as network density increases. It means that VDTN only requires a small number of vehicles for use as a low-cost alternative network for smart city.
 
Tree Algorithm Model on Size Classification Data Mining
No full textThe goal of this research is to use a tree algorithm to categorize student clothing in order to acquire an accurate size. This research is qualitative through descriptive analysis, while the analysis used C.45 tree algorithm classification. Manual calculations utilizing the tree algorithm formula revealed that most students require XL-sized clothing. On the characteristic of X5 (length of the shoulder), the maximum entropy and information gain values were obtained at 0.212642462. According to the forecast, the shoulder length attribute is the first calculation in developing a decision tree scheme since it has the largest entropy and the value of information gain. Lastly, the findings of this study analysis can be used as a mapping prediction to make decisions on the size of the student group's clothing.
The goal of this research is to use a tree algorithm to categorize student clothing in order to acquire an accurate size. This research is qualitative approach through descriptive analysis, while the analysis employed C.45 Tree algorithm classification. Manual calculations utilizing the tree algorithm formula revealed that the majority of students require XL-sized clothing. On the X5 (Shoulder length) characteristic, the maximum entropy and information gain values were obtained at 0.212642462. According to the forecast, the shoulder length attribute is the first calculation in developing a decision tree scheme since it has the largest entropy and information gain value. Lastly, the findings of this study analysis can be used as a mapping prediction to make decisions on the size of the student group's clothing
Fatigue Detection Through Car Driver’s Face Using Boosting Local Binary Patterns
Get PDFThe general population is concerned with traffic accidents. Driver fatigue is one of the leading causes of car accidents. Several factors, including nighttime driving, sleep deprivation, alcohol consumption, driving on monotonous roads, and drowsy and fatigue-inducing drugs, can contribute to fatigue. This study proposes a facial appearance-based driver fatigue detection system. This is based on the assumption that facial features can be used to identify driver fatigue. We categorize driver conditions into three groups: normal, talking, and yawning. In this study, we used Adaboost to propose Boosting Local Binary Patterns (LBP) to improve the image features of fatigue drivers in the Support Vector Machine (SVM) model. The experimental results indicate that the system's optimal performance achieves an accuracy value of 93.68%, a recall value of 94%, and a precision value of 94%.
 
Detecting Fake News on Social Media Combined with the CNN Methods
Get PDFSocial media platforms are created to facilitate human social life as technology develops. Twitter is one of the most popular and frequently used social media for exchanging information. This social media platform disseminates real-time and complete information. Unfortunately, there are not a few tweets that contain false information or are often referred to as hoaxes. Those hoaxes that existed on Twitter are very troubling for society. Fake news or hoaxes can cause misunderstandings in receiving information. Therefore, this research aimed at developing a system that can detect hoaxes on Twitter to anticipate their spread, which can be detrimental to related parties. The system being developed uses a deep learning approach with the Convolutional Neural Network (CNN), Term Frequency-Inverse Document Frequency (TF-IDF), Bidirectional Encoder Representations from Transformers (BERT), and Global Vectors (GloVe). The results of this study display the fake news detected by the system using the CNN method with baseline, BERT, and GloVe. The data have been adjusted to the keywords related to fake news and spread on online media, such as Hoax or Not from Detik.com, CekFakta from Kompas.com, etc. The results show the highest accuracy of 98.57% using CNN with a split ratio of 90:10, baseline unigram-bigram, BERT, and Top10 corpus tweet+IndoNews with an increase of 4.7%.Social media platforms are created to facilitate human social life as technology develops. Twitter is one of the most popular and frequently used social media for exchanging information. This social media platform disseminates real-time and complete information. Unfortunately, there are not a few tweets that contain false information or are often referred to as hoaxes. Those hoaxes that existed on Twitter are very troubling for society. Fake news or hoaxes can cause misunderstandings in receiving information. Therefore, this research aimed at developing a system that can detect hoaxes on Twitter to anticipate their spread, which can be detrimental to related parties. The system being developed uses a deep learning approach with the Convolutional Neural Network (CNN), Term Frequency-Inverse Document Frequency (TF-IDF), Bidirectional Encoder Representations from Transformers (BERT), and Global Vectors (GloVe). The results of this study display the fake news detected by the system using the CNN method with baseline, BERT, and GloVe. The data have been adjusted to the keywords related to fake news and spread on online media, such as Hoax or Not from Detik.com, CekFakta from Kompas.com, etc. The results show the highest accuracy of 98.57% using CNN with a split ratio of 90:10, baseline unigram-bigram, BERT, and Top10 corpus tweet+IndoNews with an increase of 4.7%