CommIT (Communication and Information Technology) Journal

Classification Taxonomies Genus of 90 Animals Using Transfer Learning Resnet-152

2023-07-25T03:32:57+00:00

The process of learning theory and the limited ability to remember anything, especially a foreign language, often cause students to have difficulty understanding lessons, especially in determining the type and taxonomy of the animal. With the assistance of computer vision technology, students can more effectively face various challenges, enhance their understanding, and improve their ability to apply the concept of animal classification. The research classifies the taxonomy of 90 animals using Transfer Learning ResNet 152. It aims to analyze the performance of Transfer Learning ResNet 152 on the 90-animal dataset. The results show that in Model A with an architecture with frozen layers in 6 ResNet blocks, the highest evaluation value obtained is 0.9222 on Batch size 4 with Dropout 6, 0.9241 on Batch size 8 with Dropout 7, 0.9259 on Batch size 16 with Dropout 8, and 0.9296 on Batch size 32 with Dropout 4 and Dropout 7. Meanwhile, in model B with an architecture with frozen layers in 5 ResNet blocks and one non-frozen block, the highest evaluation value obtained is 0.7611 on Batch size 4 with Dropout 8, 0.8713 on Batch size 8 with Dropout 2, 0.8852 on Batch size 16 with Dropout 1, and 0.9204 on Batch size 32 with Dropout 3.

Simulating Free-Space Optical Communications to Support a Li-Fi Access Network in a Smart City Concept

2024-01-02T08:12:25+00:00

Smart city development has grown rapidly in the decades since 4G and 5G technologies have been released. Moreover, a highly reliable network is required to support the Internet of Things (IoT) and mobile access within a city. Light Fidelity (Li-Fi) technology can provide huge bitrate transmission and high-speed communications. In the research, a backbone based on Free-Space Optical (FSO) communication (FSO) is designed through simulation to provide a Li-Fi access network with a high capacity data rate. The originality of the proposed method is the implementation of double filtering techniques, which gives an advantage when forwarding the signal to a node and improves the quality of the signal received by the Li-Fi. The FSO as the Optical Relaying Network (ORN) is designed with a configuration of 12 channels of Dense Wavelength Division Multiplexing (DWDM) amplified by optical amplifiers in the transmitter and receiver. The signal output is filtered by a Fiber Bragg Grating (FBG) and a Gaussian filter. In the simulation, the ORN has node spacing in the range of 500 m to 2,000 m. Then, the data transmission rate at 120 Gbps is provided by the implementation of DWDM channels to serve as an access network. From the simulation, the FSO backbone can optimally deliver highly reliable Li-Fi access networks. When the nodes are spaced in a 500–2,000 m range, the Bit-Error-Rate (BER) performance is produced at the order of 10⁻⁶.

The Determinant Factors of Shopping Cart Abandonment Among E-commerce Customers in Indonesia

2023-01-16T07:06:35+00:00

Predicting the non-purchase behavior of potential customers, such as the abandonment of online shopping carts, is a pivotal factor in determining the success of companies. Despite several conducted studies, further investigation is still required to gain a profound understanding of the underlying causes of these phenomena. The research aims to analyze the motivating factors behind shopping cart abandonment among ecommerce customers in Indonesia using a quantitative method. Furthermore, the population size is undefined, and the sample consists of 200 respondents selected through purposive sampling. The sample size is determined by five times the indicator number. The data analysis is conducted using Structural Equation Modeling (SEM) through SmartPLS 4.0.8.5, and the Coefficient of determination (R²) value for shopping cart abandonment is found to be 37.5%. The results show that complicated checkout, information overload, complicated policies, and limited shipping options positively impact shopping cart abandonment. Complicated checkout emerges as the most significant variable. Meanwhile, perceived cost and emotional ambivalence have no impact. The research also provides theoretical contributions and suggests future research for e-commerce companies and merchants. The theoretical contribution is how user emotions, user experience, merchant policies, and e-commerce regulation affect shopping cart abandonment. From the practical implications, e-commerce companies should focus on the user experience during checkout to reduce shopping cart abandonment.

Uncovering the Risk of Academic Information System Vulnerability through PTES and OWASP Method

2023-03-20T05:03:15+00:00

The security of academic information systems needs consideration to anticipate various threats, resulting in data leakage, misuse of information, modification, and data destruction. There are 36 public and private universities that utilize the academic information system provided by the software developed by Company XYZ. Limited resources in universities contribute to the weak handling of vulnerabilities in academic information systems. The research aims to determine the vulnerability level of academic information systems developed by Company XYZ through penetration testing. The research employs a deductive approach to explore academic system vulnerabilities based on incidents related to system security issues at a university. The research utilizes a combination of two testing methods: Penetration Testing Execution Standard (PTES) and Open Web Application Security Project (OWASP), chosen for their reliability, ease of use, and support by penetration testing tools. Penetration testing follows the PTES, involving seven steps: pre-engagement interaction, information collection, threat modeling, vulnerability analysis, exploitation, postexploitation, and reporting. The threat focus in the research aligns with the top 10 of 2021 OWASP, ranking the ten most critical security risks. Results reveal eight critical security issues based on measurements using the Common Vulnerability Scoring System (CVSS) method. There are two high-level vulnerabilities, five medium-level vulnerabilities, and one low-level vulnerability. Moreover, the three principal vulnerabilities are Structured Query Language (SQL) Injection, broken access control, and weak encryption. Universities can enhance data integrity by independently remediating vulnerabilities discovered in the research. Furthermore, universities are encouraged to raise awareness within the academic community regarding the security of academic data.

Insights into Mobile Government Adoption Factors: A Comprehensive Analysis of Peduli Lindungi Application in Indonesia

2023-01-09T04:26:13+00:00

Information and Communication Technology (ICT) progression has notably impacted the shift from traditional public services to digital alternatives. Among the various digital services, m-government services, provided by smartphone technology, have gained widespread popularity. Unfortunately, the broader adoption of digital technology encounters several challenges, including insufficient user interest and acceptance, as well as concerns regarding security and user privacy. The primary goal of the research is to address the existing gap in the literature by examining the factors that contribute to the effective implementation of m-government services. A mix of key components is employed, incorporating the Information Systems (IS) Success model and Technology Acceptance Model (TAM) as research variables. The research applies a quantitative approach in the form of an online survey. Furthermore, a Partial Least Square- Structure Equational Modeling (PLS-SEM) analytic approach is performed to evaluate 230 data points. The research findings support five hypotheses while rejecting three hypotheses. Significantly, the findings suggest that perceived usefulness and ease of use influence behavioral intention considerably. Additionally, constructions related to service quality significantly impact behavioral intention. Meanwhile, both system quality and information quality do not contribute to affecting behavioral intention. Furthermore, information quality exerts a substantial impact on perceived usefulness, but it does not influence perceived ease of use. Finally, it is observed that system quality significantly affects the perceived ease of use.

Program Evaluation and Review Technique (PERT) Analysis to Predict Completion Time and Project Risk Using Discrete Event System Simulation Method

2022-11-21T07:50:15+00:00

The prediction of project completion time, which is important in project management, is only based on an estimate of three numbers, namely the fastest, slowest, and presumably time. The common practice of applying normal distribution through Monte Carlo simulation in Program Evaluation and Review Technique (PERT) research often fails to accurately represent project activity durations, leading to potentially biased project completion prediction. Based on these problems, a different method is proposed, namely, Discrete Event Simulation (DES). The research aims to evaluate the effectiveness of the simmer package in R in conducting PERT analysis. Specifically, there are three objectives in the research: 1) develop a simulation model to predict how long a project will take and find the critical path, 2) create an R script to simulate discrete events on a PERT network, and 3) explore the simulation output using the simmer package in the form of summary statistics and estimation of project risk. Then, a library research with a descriptive and exploratory method is used for data collection. The hypothetical network is used to obtain the numerical results, which provide the predicted value of the project completion, the critical path, and the risk level. Simulation, including 100 replications, results in a predicted project completion time and a standard deviation of 20.7 and 2.2 weeks, respectively. The DES method has been proven highly effective in predicting the completion time of a project described by the PERT network. In addition, it offers increased flexibility.

Leaf Temperature Measurement Using Low-Resolution Thermal Camera Based on Thresholding and Clustering Techniques

2024-02-22T02:37:38+00:00

Leaf temperature can indicate photosynthetic rates, leaf water status, and stomata conductance. Leaf temperature can be measured using thermal resistance sensors, thermocouple devices, infrared thermometers, or infrared thermal imaging devices. Additionally, measuring leaf temperature using a thermal camera is simple and efficient. Therefore, the research proposes a leaf temperature measurement method using AMG8833, a low-resolution (64 pixels) thermal camera. The proposed system adopts an image segmentation technique to extract the leaf area from a thermal image. The leaf temperature is then calculated by averaging the temperature values on the leaf area. The proposed system aims to utilize a low-cost and low-resolution thermal camera for measuring the leaf temperature. The proposed approach is evaluated using real images of the Dieffenbachia plant, a popular ornamental plant that can be easily planted. In the experiments, fourteen segmentation methods consisting of eight thresholding techniques and six clustering techniques are evaluated. The experimental findings on the Dieffenbachia plant indicate that the most accurate leaf temperature measurements are obtained using local thresholding with an absolute error of 0.0109 and k-means clustering with an absolute error of 0.0134. The proposed method provides a simple, effective, and low-cost leaf temperature measurement system compared to the existing systems which employ high-cost commercial thermal cameras and complex measurement methods.

Object Detection Model for Web-Based Physical Distancing Detector Using Deep Learning

2022-09-12T03:51:41+00:00

The pandemic has changed the way people interact with each other in the public setting. As a result, social distancing has been implemented in public society to reduce the virus’s spread. Automatically detecting social distancing is paramount in reducing menial manual tasks. There are several methods to detect social distance in public, and one is through a surveillance camera. However, detecting social distance through a camera is not an easy task. Problems, such as lighting, occlusion, and camera resolution, can occur during detection. The research aims to develop a physical distancing detector system that is adjusted to work with Indonesian rules and conditions, especially in Jakarta, using deep learning (i.e., YOLOv4 architecture with the Darknet framework) and the CrowdHuman dataset. The detection is done by reading the source video, detecting the distance between individuals, and determining the crowd of individuals close to each other. In order to accomplish the detection, the training is done with CSPDarknet53 and VGG16 backbone in YOLOv4 and YOLOv4 Tiny architecture using various hyperparameters in the training process. Several explorations are made in the research to find the best combination of architectures and fine-tune them. The research successfully detects crowds at the 16th training, with mAP50 of 71.59% (74.04% AP50) and 16.2 Frame per Second (FPS) displayed on the web. The input size is essential for determining the model’s accuracy and speed. The model can be implemented in a web-based application.

Analyzing the Effects of Combining Gradient Conflict Mitigation Methods in Multi-Task Learning

2022-09-26T03:36:04+00:00

Multi-task machine learning approaches involve training a single model on multiple tasks at once to increase performance and efficiency over multiple singletask models trained individually on each task. When such a multi-task model is trained to perform multiple unrelated tasks, performance can degrade significantly since unrelated tasks often have gradients that vary widely in direction. These conflicting gradients may destructively interfere with each other, causing weights learned during the training of some tasks to become unlearned during the training of others. The research selects three existing methods to mitigate this problem: Project Conflicting Gradients (PCGrad), Modulation Module, and Language-Specific Subnetworks (LaSS). It explores how the application of different combinations of these methods affects the performance of a convolutional neural network on a multi-task image classification problem. The image classification problem used as a benchmark utilizes a dataset of 4,503 leaf images to create two separate tasks: the classification of plants and the detection of disease from leaf images. Experiment results on this problem show performance benefits over singular mitigation methods, with a combination of PCGrad and LaSS obtaining a task-averaged F1 score of 0.84686. This combination outperforms individual mitigation approaches by 0.01870, 0.02682, and 0.02434 for PCGrad, Modulation Module, and LaSS, respectively in terms of F1 score.

An Adaptive Heading Estimation Method based on Holding Styles Recognition Using Smartphone Sensors

2023-01-16T06:49:56+00:00

Pedestrian Dead Reckoning (PDR), which comes with many sensors integrated into widely available smartphones, is known as one of the most popular indoor positioning techniques. Sensors such as accelerometers, gyroscopes, and magnetometers are used to determine three important components in PDR: step detection, step length estimation, and heading estimation. Among them, the last component is the most challenging since a small heading error accumulates to produce a very large positioning error, especially when the pedestrian holds the smartphone in unconstrained styles such as swinging the phone freely along the pedestrian’s walking direction or putting the phone into the pants’ front pockets. The research proposes an adaptive heading estimation method to deal with heading errors caused by smartphone holding styles. The novelties are described as follows. Firstly, the proposed method attempts to classify the four basic smartphone holding styles using a machine learning algorithm based on simple features of acceleration values to give pedestrians more freedom during the walking period. Secondly, the proposed method adaptively combines the two heading estimation methods, which are calculated from the integrated sensors, to determine the walking direction for different smartphone holding styles. The experimental results show that the proposed heading estimation method achieves average heading errors of less than 30 degrees when testing in two different walking paths with the smartphone held in dynamic styles. It helps to reduce the heading errors by more than 15% compared to previous heading estimation methods.

Editorial Page and Table of Content

2024-04-30T03:05:23+00:00