JOIV : International Journal on Informatics Visualization
Not a member yet
786 research outputs found
Sort by
Energy-Efficient Rainfall Prediction Using Support Vector Machine on Edge Ai Platforms
Get PDFThe integration of AI into various sectors, including agriculture, has been advancing significantly. Implementing AI in the context of IoT and edge AI presents challenges due to resource limitations. Current climate changes affect planting strategies, pest management, and harvest timing. This study explores an SVM-based machine-learning model with multiple kernels to classify weather conditions as rainy or clear. The research includes two phases: model training on a PC-based system and model deployment on an edge AI device. The training phase includes preprocessing with PCA and fine-tuning of parameters, such as kernel types (linear, polynomial, sigmoid, and RBF), C and gamma. The development phase involves deploying the model on an ESP32, where execution time and power consumption are evaluated. The results show that the SVM model with an RBF kernel, C of 0.1, and gamma of 1 achieves a precision of 79.37%. Inference on the ESP32 yields an average execution time of 35.5 ms and a power consumption of 66 mA, showing a 202-fold reduction in power usage compared to the PC-based system and a 59-fold increase in execution time. This reduced power consumption supports the feasibility of edge AI for climate-based agricultural applications, enabling effective rainfall prediction. The findings contribute to the development of precision agriculture by providing insights into climate prediction, which can inform planting decisions, pest management, and harvest timing, thereby advancing the application of edge AI in response to global climate change
Addressing Class Imbalance of Health Data: A Systematic Literature Review on Modified Synthetic Minority Oversampling Technique (SMOTE) Strategies
Get PDFThe Synthetic Minority Oversampling Technique (SMOTE) method is the baseline for solving unbalanced data problems. The working concept of the SMOTE method is to generate new synthetic data patterns by performing linear interpolation between minority class samples based on k-nearest neighbors. However, the SMOTE method has weaknesses, namely the problem of overgeneralization due to excessive sampling of sample noise and increased overlapping between classes in the decision boundary area, which has the potential for noise data. Based on the weaknesses of the Smote method, the purpose of this research is to conduct a systematic literature review on the Smote method modification approach in solving unbalanced data. This systematic literature review method comprises keyword identification, article search process, determination of selection criteria, and selection results based on criteria. The results of this study showed that the SMOTE modification approach was based on filtering, clustering, and distance modification to reduce the resulting noise data. The filtering approach removed the noise data before SMOTE, positively impacting resolving unbalanced data. Meanwhile, the use of a clustering approach in SMOTE can minimize the overlapping artificial minority data that has noise potential. The most used datasets are Pima 60% and Haberman 50%. The most used performance evaluation on unbalanced data is f1-measure 57%, accuracy 55%, recall 43%, and AUC 27%. The implication of the results of this literature review is to provide opportunities for further research in modifying SMOTE in addressing health data imbalances, especially handling noise and overlapping data. The thoroughness of our literature review should instill confidence in the research community
Unveiling Gold Membership Classification Using Machine Learning
Get PDFThe main challenge in loyalty programs is selecting customers with limited funding. To address it, we explore various machine learning-based classification models. This study aims to enhance the effectiveness of a marketing strategy that promotes gold membership to customers with prior transaction history. Previously, much research applied decision trees, random forests, and logistic regression for classification, but gradient boosting is still unpopular. However, in this study, the Gradient Boost algorithm exhibits the best performance among these models, achieving an impressive accuracy of around 88%. This result underscores the model's capability to classify customers, thereby suggesting its potential to significantly enhance the marketing strategy's effectiveness. The analysis identifies crucial features that influence the model's predictive capabilities. Notably, the recency of the last visit, the number of transactions involving wine and meat, marital status, and the number of offline store transactions are identified as influential factors. Leveraging machine learning techniques enables the automation of the customer selection process, facilitating the attraction of a more extensive customer base. By targeting those customers most likely to respond positively to the gold membership offer, efficient resource allocation can be achieved. This research provides valuable insights and practical recommendations for implementing an effective marketing strategy under resource constraints. Combining machine learning algorithms and feature identification enables efficient targeting of potential customers, maximizing the impact of the gold membership offering. Implementing the findings of this study could lead to increased customer acquisition and improved overall business performance
A Multi-tier Model and Filtering Approach to Detect Fake News Using Machine Learning Algorithms
Get PDFFake news trends have overgrown in our societies over the years through social media platforms. The goal of spreading fake news can easily mislead and manipulate the public’s opinion. Many previous researchers have proposed this domain using classification algorithms or deep learning techniques. However, machine learning algorithms still suffer from high margin error, which makes them unreliable as every algorithm uses a different way of prediction. Deep learning requires high computation power and a large dataset to operate the classification model. A filtering model with a consensus layer in a multi-tier model is introduced in this research paper. The multi-tier model filters the news label correctly predicted by the first two-tier layer. The consensus layer acts as the final decision when collision results occur in the first two-tier layer. The proposed model is applied to the WEKA software tool to test and evaluate the model from both datasets. Two sequences of classification models are used in this research paper: LR_DT_RF and LR_NB_AdaBoost. The best performance of sequence for both datasets is LR_DT_RF which yields 0.9892 F1-Score, 0.9895 Accuracy, and 0.9790 Matthews Correlation Coefficient (MCC) for ISOT Fake News Dataset, and 0.9913 F1-Score, 0.9853 Accuracy, and 0.9455 MCC for CHECKED Dataset. This research could give researchers an approach for fake news detection on different social platforms and feature-base
Comparison of Adam Optimization and RMS prop in Minangkabau-Indonesian Bidirectional Translation with Neural Machine Translation
Get PDFLanguage is a tool humans use to establish communication. Still, the language used is one language and between regions or nations with their languages. Indonesia is a country that has a diversity of second languages and is the fourth most populous country in the world. It is recorded that Indonesia has nearly 800 regional languages, but research activities in natural language processing are still lacking. Minangkabau is an endangered language spoken by the Minangkabau people in Indonesia's West Sumatra province. According to UNESCO, the Minangkabau language is listed as a language that is "definitely endangered," with only around 5 million speakers worldwide. This study uses neural machine translation (NMT) to create a formula based on this information. Neural machine translation, in contrast to conventional statistical machine translation, intends to build a single neural network that can be built up to achieve the best performance. Because it can simultaneously hold memory for a long time, comprehend complicated relationships in data, and provide information that is very important in determining the outcome of translation, LSTM is one of the most powerful machine-learning techniques for translating languages. The BLUE score is utilized in the NMT evaluation. The test results use 520 Minangkabau sentences, conducting tests based on the number of epochs ranging from 100-1000, resulting in optimization using Adam being better than optimization RMSprop. This is evidenced by the results of the best BLUE-1 score of 0.997816 using 1000 epochs
Classification of Lombok Songket and Sasambo Batik Motifs Using the Convolution Neural Network (CNN) Algorithm
Get PDFSasambo batik is a traditional batik from the West Nusa Tenggara province. Sasambo itself is an abbreviation of three tribes, namely the Sasak (sa) in the Lombok Islands, the Samawa (sam), and the Mbojo (bo) tribes in Sumbawa Island. Classification of batik motifs can use image processing technology, one of which is the Convolution Neural Network (CNN) algorithm. Before entering the classification process, the batik image first undergoes image resizing. After that, proceed with the operation of the convolution, pooling, and fully connected layers. The sample image of Lombok songket motifs and Sasambo batik consists of 20 songket fabric data with the same motif and color and 14 songket data with the same motif but different colors. In addition, there are 10 data points on songket fabrics with other motifs and colors. In addition, there are 5 data points on Sasambo batik fabrics with the same motif and color and 5 data points on Sasambo batik fabrics with the same motif but different colors. The training data rotates the image by 150 as many as 20 photos. Testing with motifs with the same color shows that the system's success rate is 83.85%. The highest average recognition for Sasambo batik cloth is in testing motifs with the same color for data in the database at 93.66%. The CNN modeling classification results indicate that the Sasambo batik cloth can be a reference for developing songket categorization using a website platform or the Android system
Leveraging Various Feature Selection Methods for Churn Prediction Using Various Machine Learning Algorithms
Get PDFThis study aims to examine the effect of customer experience on customer retention at DQLab Telco, using machine learning techniques to predict customer churn. The study uses a dataset of 6590 customers of DQLab Telco, which contains various features related to their service usage and satisfaction. The data includes various features such as gender, tenure, phone service, internet service, monthly charges, and total charges. These features represent the demographic and service usage information of the customers. The study applies several feature selection methods, such as ANOVA, Recursive Feature Elimination, Feature Importance, and Pearson Correlation, to select the most relevant features for churn prediction. The study also compares three machine learning algorithms, namely Logistic Regression, Random Forest, and Gradient Boosting, to build and evaluate the prediction models. This study finds that Logistic Regression without feature selection achieves the highest accuracy of 79.47%, while Random Forest with Feature Importance and Gradient Boosting with Recursive Feature Elimination achieve accuracy of 77.60% and 79.86%, respectively. The study also identifies the features influencing customer churn most, such as monthly charges, tenure, partner, senior citizen, internet service, paperless billing, and TV streaming. The study provides valuable insights for DQLab Telco in developing customer churn reduction strategies based on predictive models and influential features. The study also suggests that feature selection and machine learning algorithms play a vital role in improving the accuracy of churn prediction and should be customized according to the data context
Optimizing Educational Assessment: The Practicality of Computer Adaptive Testing (CAT) with an Item Response Theory (IRT) Approach
Get PDFThis research aims to develop a Computer Adaptive Test (CAT) system using the Items Response Theory (IRT) approach. This study is part of developing a web-based system using the Research and Development (R&D) method, employing the Four-D (4-D) model. At its core, this system is similar to a Computer-Based Test (CBT). Still, the critical difference lies in its ability to randomize and provide questions that align with the test-taker's skill levels using the Items Response Theory (IRT) algorithm. The system employs the 3-PL model from the Items Response Theory, considering the difficulty level of questions, the discriminative power of questions, and the likelihood of guessing or interference in the questions. The examination system randomly assigns questions to students based on their responses to previous questions, ensuring that each test-taker receives a unique question sequence. The exam concludes when a test-taker accurately estimates their ability, i.e., SE <= 0.01, or when all questions have been answered. The outcome of this research is a Computer Adaptive Test (CAT) system based on the Items Response Theory (IRT), which can be used to assess students' learning outcomes. This research was implemented in the Multimedia Department of SMK Negeri 1 Gunung Talang, with 90 students as the research sample. The evaluation of the practicality of this system received very high scores, indicating that the Computer Adaptive Test (CAT) system based on the Items Response Theory (IRT) is considered highly practical and effective in achieving the established measurement goals
Offline Handwriting Writer Identification using Depth-wise Separable Convolution with Siamese Network
Get PDFOffline handwriting writer identification has significant implications for forensic investigations and biometric authentication. Handwriting, as a distinctive biometric trait, provides insights into individual identity. Despite advancements in handcrafted algorithms and deep learning techniques, the persistent challenges related to intra-variability and inter-writer similarity continue to drive research efforts. In this study, we build on well-separated convolution architectures like the Xception architecture, which has proven to be robust in our previous research comparing various deep learning architectures such as MobileNet, EfficientNet, ResNet50, and VGG16, where Xception demonstrated minimal training-validation disparities for writer identification. Expanding on this, we use a model based on similarity or dissimilarity approaches to identify offline writers' handwriting, known as the Siamese Network, that incorporates the Xception architecture. Similarity or dissimilarity measurements are based on the Manhattan or L1 distance between representation vectors of each input pair. We train publicly available IAM and CVL datasets; our approach achieves accuracy rates of 99.81% for IAM and 99.88% for CVL. The model was evaluated using evaluation metrics, which revealed only two error predictions in the IAM dataset, resulting in 99.75% accuracy, and five error predictions for CVL, resulting in 99.57% accuracy. These findings modestly surpass existing achievements, highlighting the potential inherent in our methodology to enhance writer identification accuracy. This study underscores the effectiveness of integrating the Siamese Network with depth-wise separable convolution, emphasizing the practical implications for supporting writer identification in real-world applications
Implementation of Word Trends Using a Machine Learning Approach with TF-IDF and Latent Dirichlet Allocation
Get PDFIn today's technological age, the prevalence of social media has become ubiquitous, facilitating the easy dissemination of information and communication. This has led to the uploading of various content, including opinions on mental health, particularly in Indonesia. Mental health refers to an individual's emotional, psychological, and social well-being, commonly affecting individuals from adolescence to adulthood. This research utilized Twitter data on mental health issues gathered from October to November 2022, employing TF-IDF and Latent Dirichlet Allocation (LDA) to conduct topic modeling for word trend analysis based on user-generated content. The sentiment analysis concept was used to label text as either negative or positive sentiment. Subsequently, TF-IDF weighed the word frequency in the documents/tweets, categorizing the data based on the resulting sentiments. Manual labeling ensured accuracy, avoiding potential errors from libraries provided in the Indonesian language. Employing these two topic modeling techniques, conclusions were drawn for each concept, aiming to identify word trends, mainly focusing on mental health discourse within Twitter user-generated content. Results indicated the synchronicity of the keyword 'mental health' with word trends generated by LDA. At the same time, TF-IDF produced word trends based on positive and negative labels, revealing commonly used terms by Twitter users to express these concerns. Furthermore, subsequent research can be experimented by comparing topic modeling techniques using Latent Semantic Allocation (LSA), Probabilistic Latent Semantic Analysis (pLSA), and Hierarchical Dirichlet Process (HDP), where LSA and pLSA present approaches closely aligned with LDA