MATRIK : Jurnal Manajemen, Teknik Informatika dan Rekayasa Komputer

Image Data Acquisition and Classification of Vannamei Shrimp Cultivation Results Based on Deep Learning

2024-06-14T09:08:16+08:00

This research aimed to employ deep learning techniques to address the classification of Litopenaeus vannamei cultivation results in land ponds and tarpaulin ponds. Despite their similar appearance, distinguishable differences exist in various aspects such as color, shape, size, and market price between the two cultivation methods, often leading to consumer confusion and potential exploitation by irresponsible sellers. To mitigate this challenge, the research proposed a classification method utilizing two Convolutional Neural Network (CNN) architectures: Visual Geometry Group-16 (VGG-16) and Residual Network-50 (ResNet-50), renowned for their success in various image recognition applications. The dataset comprised 2,080 images per class of vannamei shrimp from both types of ponds. Augmentation techniques enhanced the dataset’s diversity and sample size, reinforcing the model’s ability to discern shrimp morphology variations. Experiments were conducted with learning rates of 0.001 and 0.0001 on the Stochastic Gradient Descent (SGD) and Adaptive Moment Estimation (ADAM) optimizers to evaluate their effectiveness in model training. The VGG-16 and ResNet-50 models were trained with a learning rate parameter of 0.0001, leveraging the flexibility and reasonable control provided by the SGD optimizer. Lower learning rate values were chosen to prevent overfitting and enhance training stability. The model evaluation demonstrated promising results, with both architectures achieving 100% accuracy in classifying vannamei shrimp from soil ponds and tarpaulin ponds. Furthermore, experimental findings highlight the superiority of using SGD with a learning rate of 0.0001 over 0.001 on both architectures, underscoring the significant impact of optimizer and learning rate selection on model training effectiveness in image classification tasks.

Quality Improvement for Invisible Watermarking using Singular Value Decomposition and Discrete Cosine Transform

2024-06-14T11:16:52+08:00

Image watermarking is a sophisticated method often used to assert ownership and ensure the integrity of digital images. This research aimed to propose and evaluate an advanced watermarking technique that utilizes a combination of singular value decomposition methodology and discrete cosine transformation to embed the Dian Nuswantoro University symbol as proof of ownership into digital images. Specific goals included optimizing the embedding process to ensure high fidelity of the embedded watermark and evaluating the fuzziness of the watermark to maintain the visual quality of the watermarked image. The methods used in this research were singular value decomposition and discrete cosine transformation, which are implemented because of their complementary strengths. Singular value decomposition offers robustness and stability, while discrete cosine transformation provides efficient frequency domain transformation, thereby increasing the overall effectiveness of the watermarking process. The results of this study showed the efficacy of the Lena image technique in gray scale having a mean square error of 0.0001, a high peak signal-to-noise ratio of 89.13 decibels (dB), a universal quality index of 0.9945, and a similarity index structural of 0.999. These findings confirmed that the proposed approach maintains image quality while providing watermarking resistance. In conclusion, this research contributed a new watermarking technique designed to verify institutional ownership in digital images, specifically benefiting Dian Nuswantoro University. It showed significant potential for wider application in digital rights management.

Optimizing Rain Prediction Model Using Random Forest and Grid Search Cross-Validation for Agriculture Sector

2024-07-04T14:27:42+08:00

Agriculture, as a sector that is highly influenced by weather conditions, faces challenges due to increasingly unpredictable changes in weather patterns. The aim of this research is to create an optimal rainfall prediction model to help farmers create irrigation schedules, use fertilizer, and planting schedules, and protect plants from extreme weather events. The method used in this research to obtain the best rain prediction model is to use the random forest algorithm and the grid search cross-validation algorithm. Random Forest, known for its robustness and accuracy, emerged as a suitable algorithm for predicting rain. utilizing a substantial dataset from the West Nusa Tenggara Meteorology, Climatology, and Geophysics Agency covering the period 2000 to 2023. The data is then processed first to ensure its readiness for use. This process involves removing outlier data points, empty data entries, and unused features. After the preprocessing stage, the data underwent training using the Random Forest algorithm, resulting in an R-squared value of 0.1334. To obtain the optimal model, Grid Search Cross Validation is used. The results of this research obtained the best rain prediction model with an R-squared value of 0.0268. This model will be used to predict rain in the agricultural sector. This research concludes that we can get the best rain prediction model by combining Random Forest and Gird Search Cross-Validation. For further research, we can compare other rain prediction methods, add features, and combine datasets from a wider area.

Gender Classification Using Viola Jones, Orthogonal Difference Local Binary Pattern and Principal Component Analysis

2024-07-04T14:43:15+08:00

Facial recognition is currently a widely discussed topic, particularly in the context of gender classification. Facial recognition by computers is more complex and time-consuming compared to humans. There is ongoing research on facial feature extraction for gender classification. Geometry and texture features are effective for gender classification. This study aimed to combine these two features to improve the accuracy of gender classification. This research used the Viola-Jones and Orthogonal Difference Local Binary Pattern (OD-LBP) methods for feature extraction. The Viola-Jones algorithm faces issues in facial detection, leading to outliers in geometry features. At the same time, OD-LBP is a new descriptor capable of addressing pose, lighting, and expression variations. Therefore, this research attempts to utilize OD-LBP for gender classification. The dataset used was FERET, which contained various lighting variations, making OD-LBP suitable for addressing this challenge. Random Forest and Backpropagation were employed for classification. This research demonstrates that combining these two features is effective for gender classification using Backpropagation, achieving an accuracy of 93%. This confirms the superiority of the proposed method over single-feature extraction methods.

A Comparison of Enhanced Ensemble Learning Techniques for Internet of Things Network Attack Detection

2024-07-12T15:33:05+08:00

Over the past few decades, the Internet of Things (IoT) has become increasingly significant due to its capacity to enable low-cost device and sensor communication. Implementation has opened up many new opportunities in terms of efficiency, productivity, convenience, and security. However, it has also brought about new privacy and data security challenges, interoperability, and network reliability. The research issue is that IoT devices are frequently open to attacks. Certain machine learning (ML) algorithms still struggle to handle imbalanced data and have weak generalization skills when compared to ensemble learning. The research aims to develop security for IoT networks based on enhanced ensemble learning by using Grid Search and Random Search techniques. The method used is the ensemble learning approach, which consists of Random Forest (RF), Adaptive Boosting (AdaBoost), Gradient Boosting Machine (GBM), and Extreme Gradient Boosting (XGBoost). This study uses the UNSW-NB15 IoT dataset. The study's findings demonstrate that XGBoost performs better than other methods at identifying IoT network attacks. By employing Grid Search and Random Search optimization, XGBoost achieves an accuracy rate of 98.56% in binary model measurements and 97.47% on multi-class data. The findings underscore the efficacy of XGBoost in bolstering security within IoT networks.

Normalization Layer Enhancement in Convolutional Neural Network for Parking Space Classification

2024-07-12T15:39:36+08:00

The research problem of this study is the urgent need for real-time parking availability information to assist drivers in quickly and accurately locating available parking spaces, aiming to improve upon the accuracy not achieved by previous studies. The objective of this research is to enhance the classification accuracy of parking spaces using a Convolutional Neural Network (CNN) model, specifically by integrating an effective normalizing function into the CNN architecture. The research method employed involves the application of four distinct normalizing functions to the EfficientParkingNet, a tailored CNN architecture designed for the precise classification of parking spaces. The results indicate that the EfficientParkingNet model, when equipped with the Group Normalization function, outperforms other models using Batch Normalization, Inter-Channel Local Response Normalization, and Intra-Channel Local Response Normalization in terms of classification accuracy. Furthermore, it surpasses other similar CNN models such as mAlexnet, you only look once (Yolo)+mobilenet, and CarNet in the same classification task. This demonstrates that EfficientParkingNet with Group Normalization significantly enhances parking space classification, thus providing drivers with more reliable and accurate parking availability information.

Deep Learning Model Compression Techniques Performance on Edge Devices

2024-07-12T15:44:43+08:00

Artificial intelligence at the edge can help solve complex tasks faced by various sectors such as automotive, healthcare and surveillance. However, challenged by the lack of computational power from the edge devices, artificial intelligence models are forced to adapt. Many have developed and quantified model compres-sion approaches over the years to tackle this problem. However, not many have considered the overhead of on-device model compression, even though model compression can take a considerable amount of time. With the added metric, we provide a more complete view on the efficiency of model compression on the edge. The objective of this research is identifying the benefit of compression methods and it’s tradeoff between size and latency reduction versus the accuracy loss as well as compression time in edge devices. In this work, quantitative method is used to analyze and rank three common ways of model compression: post-training quantization, unstructured pruning and knowledge distillation on the basis of accuracy, latency, model size and time to compress overhead. We concluded that knowledge distillation is the best, with potential of up to 11.4x model size reduction, and 78.67% latency speed up, with moderate loss of accura-cy and compression time.

New Method for Identification and Response to Infectious Disease Patterns Based on Comprehensive Health Service Data

2024-07-12T15:53:04+08:00

Infectious diseases continue to pose a major threat to global public health and require early detection and effective response strategies. Despite advances in information technology and data analysis, the full potential of health data in identifying disease patterns and trends remains underutilised. This study aims to propose a comprehensive new mathematical model (new method) that utilises health data to identify infectious disease patterns and trends by exploring the potential of data-driven care approaches in addressing public health challenges associated with infectious diseases. The research methods used are exploratory data collection and analytical model development. The research results obtained mathematical models and algorithms that consider data of period, time, patterns, and trends of dangerous diseases, statistical analysis, and recommendations. Data visualisation and in-depth analysis were conducted in the research to improve the ability to respond to infectious disease threats and provide better decision-making solutions in improving outbreak response, as well as improving preparedness in addressing public health challenges. This research contributes to health practitioners and decision-makers.

Accuracy of K-Nearest Neighbors Algorithm Classification For Archiving Research Publications

2024-07-12T15:57:08+08:00

The Archives and Research Publication Information System plays an important role in managing academic research and scientific publications efficiently. With the increasing volume of research and publications carried out each year by university researchers, the Research Archives and Publications Information System is essential for organizing and processing these materials. However, managing large amounts of data poses challenges, including the need to accurately classify a researcher's field of study. To overcome these challenges, this research focuses on implementing the K-Nearest Neighbors classification algorithm in the Archives and Research Publications Information System application. This research aims to improve the accuracy of classification systems and facilitate better decision-making in the management of academic research. This research method is systematic involving data acquisition, pre-processing, algorithm implementation, and evaluation. The results of this research show that integrating Chi-Square feature selection significantly improves K-Nearest Neighbors performance, achieving 86% precision, 84.3% recall, 89.2% F1 Score, and 93.3% accuracy. This research contributes to increasing the efficiency of the Archives and Research Publication Information System in managing research and academic publications.

Development of the Multi-Channel Clustering Hierarchy Method for Increasing Performance in Wireless Sensor Network

2024-07-12T16:02:22+08:00

Wireless Sensor Networks are technologies that make it possible to observe phenomena. The problem is data delays in covering the distance from the origin to the destination. Packet Loss is a condition that shows the number of lost packets and the total queue length caused by data processing time. This research aims to develop a cluster-based protocol. This research uses a multichannel hierarchical clustering method and adds odd-even by dividing the network into several channels and forming a cluster head for each channel. The results of this research are Channel 1 with a throughput value of 1.88, channel 2 with a throughput value of 21.68, channel 3 with a throughput value of 1.62, and channel 4 with a throughput value of 42.44. The conclusion of this study is that the throughput results are smaller compared to the Multi-Channel Clustering H ierarchy method, because not all nodes are active

Classification of Cash Direct Recipients Using the Naive Bayes with Smoothing

2024-07-12T16:05:14+08:00

Direct Cash Assistance is a social program distributed to residents meeting specific requirements. The village government determines the recipients using a conventional system through village meetings. This approach is greatly influenced by the decision-holders’ subjectivity with non-transparent thinking. This research aims to solve the problem of classifying Direct Cash Assistance recipients by applying probability-based classification. The research method used is smoothed Nave Bayes, which improves Nave Bayes by adding a constant to avoid zero classification. The datasets use variables such as age, type of work, and criteria for receiving assistance. The last variable includes five nominal data, which debilitates Nave Bayes by not obtaining a posterior probability as a prediction class result. We used Direct Cash Assistance data from the Sedati sub-district, Sidoarjo district, East Java. The results of research with original Nave Bayes and smoothed Nave Bayes classification show that smoothed Nave Bayes has good prediction performance with an accuracy of 95.9% with a data split of 60:40. Smoothed- Nave Bayes also solves the problem of 8 data without predictive classes. The prediction results show that Smoothed Nave Bayes performs better than standard Nave Bayes. This research contributes to refining Nave Bayes to complement probability-based classification by adding refinement constants to avoid zero classification.

DynamicWeighted Particle Swarm Optimization - Support Vector Machine Optimization in Recursive Feature Elimination Feature Selection

2024-07-12T16:09:11+08:00

Feature Selection is a crucial step in data preprocessing to enhance machine learning efficiency, reduce computational complexity, and improve classification accuracy. The main challenge in feature selection for classification is identifying the most relevant and informative subset to enhance prediction accuracy. Previous studies often resulted in suboptimal subsets, leading to poor model performance and low accuracy. This research aims to enhance classification accuracy by utilizing Recursive Feature Elimination (RFE) combined with Dynamic Weighted Particle Swarm Optimization (DWPSO) and Support Vector Machine (SVM) algorithms. The research method involves the utilization of 12 datasets from the University of California, Irvine (UCI) repository, where features are selected via RFE and applied to the DWPSO-SVM algorithm. RFE iteratively removes the weakest features, constructing a model with the most relevant features to enhance accuracy. The research findings indicate that DWPSO-SVM with RFE significantly improves classification accuracy. For example, accuracy on the Breast Cancer dataset increased from 58% to 76%, and on the Heart dataset from 80% to 97%. The highest accuracy achieved was 100% on the Iris dataset. The conclusion of these findings that RFE in DWPSO-SVM offers consistent and balanced results in True Positive Rate (TPR) and True Negative Rate (TNR), providing reliable and accurate predictions for various applications.

K-Means Optimization Algorithm to Improve Cluster Quality on Sparse Data

2024-07-12T16:15:58+08:00

The aim of this research is clustering sparse data using various K-Means optimization algorithms. Sparse data used in this research came from Citampi Stories game reviews on Google Play Store. This research method are Density Based Spatial Clustering of Applications with Noise-Kmeans (DB-Kmeans), Particle Swarm Optimization-Kmeans (PSO-Kmeans), and Robust Sparse Kmeans Clustering (RSKC) which are evaluated using the silhouette score. Clustering sparse data presented a challenge as it could complicate the analysis process, leading to suboptimal or non-representative results. To address this challenge, the research employed an approach that involved dividing the data based on the number of terms in three different scenarios to reduce sparsity. The results of this research showed that DB-Kmeans had the potential to enhance clustering quality across most data scenarios. Additionally, this research found that dividing data based on the number of terms could effectively mitigate sparsity, significantly influencing the optimization of topic formation within each cluster. The conclusion of this research is that this approach is effective in enhancing the quality of clustering for sparse data, providing more diverse and easily interpretable information. The results of this research could be valuable for developers seeking to understand user preferences and enhance game quality.

Improved Image Segmentation using Adaptive Threshold Morphology on CT-Scan Images for Brain Tumor Detection

2024-07-12T16:20:11+08:00

Diagnosing disease by playing the role of image processing is one form of current medical technology development. The results of image processing performance have been able to provide accurate diagnoses to be used as material for decision-making. This research aims to carry out the process of detecting brain tumor objects in Computed Tomography (CT-Scan) images by developing a segmentation technique using the Adaptive Threshold Morphology (ATM) algorithm. The performance of the ATM algorithm in the segmentation process involves the Extended Adaptive Global Treshold (eAGT) function to produce an optimal threshold value. This research method involves several stages of the process in detecting tumor objects. The preprocessing stage is carried out using the cropping and filtering process which is optimized using the eAGT function. The next stage is the morphological segmentation process involving erosion and dilation operations. The final stage of the segmentation process using the ATM algorithm is labeling objects that have been detected. The research dataset used 187 Computed Tomography-Scan images from 10 brain tumor patients. The results of this study show that the accuracy rate for detecting brain tumor objects in Computed Tomography-Scan images is 93.47%. These results can provide an automatic and effective detection process based on the optimal threshold value that has been generated. Overall, this research contributes to the development of segmentation algorithms in image processing and can be used as an alternative solution in the treatment of brain tumor patients.

Multiclass Text Classification of Indonesian Short Message Service (SMS) Spam using Deep Learning Method and Easy Data Augmentation

2024-07-12T16:35:49+08:00

The ease of using Short Message Service (SMS) has brought the issue of SMS spam, characterized by unsolicited and unwanted. Many studies have been conducted utilizing machine learning methods to build models capable of classifying SMS Spam to overcome this problem. However, most of these studies still rely on traditional methods, with limited exploration of deep learning-based approaches. Whereas traditional methods have a limitation compared to deep learning, which performs manual feature extraction. Moreover, many of these studies only focus on binary classification rather than multiclass SMS classification which can provide more detailed classification results. The aim of this research is to analyze deep learning model for multiclass Indonesian SMS spam classification with six categories and to assess the effectiveness of the text augmentation method in addressing data imbalace issues arising from the increased number of SMS categories. The research method used were Indonesian version of Bidirectional Encoder Representations from Transformers (IndoBERT) model and exploratory data analysis (EDA) augmentation technique to address imbalance dataset issue. The evaluation is conducted by comparing the performance of the IndoBERT model on the dataset and applying EDA techniques to enhance the representation of minority classes. The result of this research shows that IndoBERT achieves 91% accuracy rate in classifying SMS spam. Furthermore, the use of EDA technique results in significant improvement in f1-score, with an average 12% increase in minority classes. Overall model accuracy also improves to 93% after EDA implementation. This research concludes that IndoBERT is effective for multiclass SMS spam classification, and the EDA is beneficial in handling imbalanced data, contributing to the enhancement of model performances.

Color Feature Extraction for Grape Variety Identification: Naïve Bayes Approach

2024-07-23T09:36:20+08:00

The problem addressed in this research is the lack of an efficient and accurate method for automatically identifying grape varieties. Accurate identification is crucial for quality control in the agricultural and food industries, impacting product labeling, pricing, and consumer trust. The aim of this research is to develop an automated system to classify green, black, and red grapes using digital image processing technology. This research method employs Naïve Bayes classification combined with color feature extraction. Testing was conducted under two scenarios: a database scenario with predefined grape image datasets and an out-of-database scenario with images resembling grape colors. Image processing includes resizing images to 200x200 pixels, Gamma Correction, Gaussian filtering, conversion to Lab* color space, K-Means Clustering for segmentation, followed by feature extraction and Naïve Bayes classification. The results of this research are that in the database scenario, the system achieved accuracies of 98.33% with an 80:20 data split and 98.89% with a 70:30 split. In the out-of-database scenario, accuracies were 96.67% with an 80:20 split and 97.78% with a 70:30 split. The conclusion of this research is the proposed method provides a reliable and efficient solution for automatic grape variety identification, benefiting quality control in agriculture and food industries.