Search

The primary objective of this study is to analyze multivariate time series data by employing the Long Short-Term Memory (LSTM) model. Deep learning models often face issues when dealing with multivariate time series data, which is defined by several variables that have diverse value ranges. These challenges arise owing to the potential biases present in the data. In order to tackle this issue, it is crucial to employ normalization techniques such as min-max and z-score to guarantee that the qualities are standardized and can be compared effectively. This study assesses the effectiveness of the LSTM model by applying two normalizing techniques in five distinct attribute selection scenarios. The aim of this study is to ascertain the normalization strategy that produces the most precise outcomes when employed in the LSTM model for the analysis of multivariate time series. The evaluation measures employed in this study comprise Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and R-Squared (R2). The results suggest that the min-max normalization method regularly yields superior outcomes in comparison to the z-score method. Min-max normalization specifically resulted in a decreased mean absolute percentage error (MAPE) and root mean square error (RMSE), as well as an increased R-squared (R2) value. These improvements indicate enhanced accuracy and performance of the model. This paper makes a significant contribution by doing a thorough comparison analysis of normalizing procedures. It offers vital insights for researchers and practitioners in choosing suitable preprocessing strategies to improve the performance of deep learning models. The study's findings underscore the importance of selecting the appropriate normalization strategy to enhance the precision and dependability of multivariate time series predictions using LSTM models. To summarize, the results indicate that min-max normalization is superior to z-score normalization for this particular use case. This provides a useful suggestion for further studies and practical applications in the field. This study emphasizes the significance of normalization in analyzing multivariate time series and contributes to the larger comprehension of data preprocessing in deep learning models

Energy use is an essential aspect of many human activities, from individual to industrial scale. However, increasing global energy demand and the challenges posed by environmental change make understanding energy use patterns crucial. Accurate predictions of future energy consumption can greatly influence decision-making, supply-demand stability and energy efficiency. Energy use data often exhibits time-series patterns, which creates complexity in forecasting. To address this complexity, this research utilizes Deep Learning (DL), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), and Gated Recurrent Unit (GRU) models. The main objective is to improve the accuracy of energy usage forecasting by optimizing the alpha value in exponential smoothing, thereby improving forecasting accuracy. The results showed that all DL methods experienced improved accuracy when using optimum alpha. LSTM has the most optimal MAPE, RMSE, and R2 values compared to other methods. This research promotes energy management, decision-making, and efficiency by providing an innovative framework for accurate forecasting of energy use, thus contributing to a sustainable and efficient energy system.

The present study explores the domain of Keris classification by employing advanced Convolutional Neural Networks (CNNs) as a potent technique for identifying subtle patterns and cultural characteristics inherent in these renowned Indonesian daggers. The study has presented encouraging findings about the identification of Pamor, Dhapur, and Tangguh categories. However, it is crucial to recognise and confront the inherent constraints associated with this research. The key constraints of the study pertain to the diversity of data, accuracy of labeling, generalizability of the model, and ethical considerations. The acquisition of a comprehensive dataset that effectively encompasses the whole range of Keris patterns offers a significant obstacle. Furthermore, it is crucial to pay careful attention to the accuracy of labeling, since it can be influenced by the subjective character of Keris classification. The important worry lies in guaranteeing the model's capacity to generalise to Keris images that have not been previously encountered, as well as its ability to comprehend and explain its decision-making process. The careful establishment of ethical frameworks is necessary to address ethical problems related to cultural sensitivity and the potential misuse of AI outputs in the realm of cultural heritage. Nevertheless, these constraints offer significant perspectives on potential areas for future investigation and enhancement. Future endeavours may prioritise the augmentation and broadening of the dataset, fostering collaboration with specialists in cultural domains, improving the interpretability of the model, and effectively addressing ethical considerations. The present study not only exhibits potential for expanding artificial intelligence in the domain of cultural preservation, but also contributes to a more profound understanding and recognition of the complex artistry and historical significance encapsulated within the Keris.

The ensemble method is considered an advanced method in both prediction and classification. The application of this method is estimated to have a more optimal output than the previous classification method. This article aims to determine the ensemble's performance to classify journal quartiles. The subject of agriculture was chosen because Indonesia is an agricultural country, and the interest of researchers in this field shows a positive response. The data is downloaded through the Scimago Journal and Country Rank with the accumulation in 2020. Labels have four classes: Q1, Q2, Q3, and Q4. The ensemble applied is Boosting and Bagging with Decision Tree (DT) and Gaussian Naïve Bayes (GNB) algorithms compiled from 2144 instances. The Boosting meta-ensembles used are Adaboost and XGBoost. From this study, the Bagging Decision Tree has the highest accuracy score at 71.36, followed by XGBoost Decision Tree with 69.51. The third is XGBoost Gaussian Naïve Bayes with 68.82, Adaboost Decision Tree with 60.42, Adaboost Gaussian Naïve Bayes with 58.2, and Bagging Gaussian Naïve Bayes with 56.12 results. This paper shows that the Bagging Decision Tree is the ensemble method that works optimally in this subject classification. This result suggests that the ensemble method can still fail to produce an ideal outcome that approaches the SJR system.

Convolutional Neural Network (CNN) is an effective Deep Learning (DL) algorithm that solves various image identification problems. The use of CNN for time-series data analysis is emerging. CNN learns filters, representations of repeated patterns in the series, and uses them to forecast future values. The network performance may depend on hyperparameter settings. This study optimizes the CNN architecture based on hyperparameter tuning using Particle Swarm Optimization (PSO), PSO-CNN. The proposed method was evaluated using multivariate time-series data of electronic journal visitor datasets. The CNN equation in image and time-series problems is the input given to the model for processing numbers. The proposed method generated the lowest RMSE (1.386) with 178 neurons in the fully connected and 2 hidden layers. The experimental results show that the PSO-CNN generates an architecture with better performance than ordinary CNN.

The tembang macapat can be classified using its cultural concepts of guru lagu, guru wilangan, and guru gatra. People may face difficulties recognizing certain songs based on the established rules. This study aims to build classification models of tembang macapat using a simple yet powerful Naïve Bayes classifier. The Naive Bayes can generate high-accuracy values from sparse data. This study modifies the concept of Guru Lagu by retrieving the last vowel of each line. At the same time, guru wilangan’s guidelines are amended by counting the number of all characters (Model 2) rather than calculating the number of syllables (Model 1). The data source is serat wulangreh with 11 types of tembang macapat, namely maskumambang, mijil, sinom, durma, asmaradana, kinanthi, pucung, gambuh, pangkur, dandhanggula, and megatruh. The k-fold cross-validation is used to evaluate the performance of 88 data. The result shows that the proposed Model 1 performs better than Model 2 in macapat classification. This promising method opens the potential of using a data mining classification engine as cultural teaching and preservation media.

Nowadays, food is one of the things that has been globalized, and everyone from different parts of the world has been able to cook food from other countries through existing online recipes. Based on that, this study developed a translation formula using a neural machine translation (NMT). NMT is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder–decoders. Our experiment led to novel insights and practical advice for building and extending NMT with the applied long short-term memory (LSTM) method to 47 bilingual food recipes between Spanish-English and English-Spanish. LSTM is one of the best machine learning methods for translating languages because it can retain memories for an extended period concurrently, grasp complicated connections between data, and provides highly useful information in deciding translation outcomes. The evaluation for this neural machine translation is to use BLEU. The comparing results show that the translation of recipes from Spanish-English has a better BLEU value of 0.998426 than English-Spanish with a data-sharing of 70%:30% during epoch 1000. Researchers can convert the country's popular cuisine recipes into another language for further research, allowing it to become more widely recognized abroad.

Deep learning is a machine learning approach that produces excellent performance in various applications, including natural language processing, image identification, and forecasting. Deep learning network performance depends on the hyperparameter settings. This research attempts to optimize the deep learning architecture of Long short term memory (LSTM), Convolutional neural network (CNN), and Multilayer perceptron (MLP) for forecasting tasks using Particle swarm optimization (PSO), a swarm intelligence-based metaheuristic optimization methodology: Proposed M-1 (PSO-LSTM), M-2 (PSO-CNN), and M-3 (PSO-MLP). Beijing PM2.5 datasets was analyzed to measure the performance of the proposed models. PM2.5 as a target variable was affected by dew point, pressure, temperature, cumulated wind speed, hours of snow, and hours of rain. The deep learning network inputs consist of three different scenarios: daily, weekly, and monthly. The results show that the proposed M-1 with three hidden layers produces the best results of RMSE and MAPE compared to the proposed M-2, M-3, and all the baselines. A recommendation for air pollution management could be generated by using these optimized models.

Abstract CNN originates from image processing and is not commonly known as a forecasting technique in time-series analysis which depends on the quality of input data. One of the methods to improve the quality is by smoothing the data. This study introduces a novel hybrid exponential smoothing using CNN called Smoothed-CNN (S-CNN). The method of combining tactics outperforms the majority of individual solutions in forecasting. The S-CNN was compared with the original CNN method and other forecasting methods such as Multilayer Perceptron (MLP) and Long Short-Term Memory (LSTM). The dataset is a year time-series of daily website visitors. Since there are no special rules for using the number of hidden layers, the Lucas number was used. The results show that S-CNN is better than MLP and LSTM, with the best MSE of 0.012147693 using 76 hidden layers at 80%:20% data composition.

The need for accurate time-series results is badly demanding. LSTM has been applied for forecasting time series, which is generated when variables are observed at discrete and equal time intervals. Nevertheless, the problem of determining hyperparameters with a relatively high random rate will reduce the accuracy of the prediction results. This paper aims to promote LSTM with tuned-PSO and Bifold-Attention mechanism. PSO optimizes LSTM hyperparameters, and Bifold-attention mechanism selects the optimal input for LSTM. An accurate, adaptive, and robust time-series forecasting model is the main contribution, compared with ARIMA, MLP, LSTM, PSO-LSTM, A-LSTM, and PSO-A-LSTM. The model comparison is based on the accuracy of each model in forecasting Beijing PM2.5, Beijing Multi-Site, Air Quality, Appliances Energy, Wind Speed, and Traffic Flow. The Proposed model, LSTM with tuned-PSO and Bifold-Attention mechanism, has lower MAPE and RMSE than baselines. In other words, the model outperformed all LSTM base models in this study. The proposed model’s accuracy is adaptable in daily, weekly, and monthly multivariate time-series datasets. This ground-breaking innovation is valuable for time-series analysis research, particularly the implementation of deep learning for time-series forecasting.

Computational methods have been used extensively to solve problems in the education sector. This paper aims to explore the computational method's recent implementation in solving global Vocational education and training (VET) problems. The study used a systematic literature review to answer specific research questions by identifying, assessing, and interpreting all available research shreds of evidence. The result shows that researchers use the computational method to predict various cases in VET. The most popular methods are ANN and Naïve Bayes. It has significant potential to develop because VET has a very complex problem of (a) curriculum, (b) apprenticeship, (c) matching labor market, and (d) attracting enrollment. In the future, academics may have broad overviews of the use of the computational method in VET. A computer scientist may use this study to find more efficient and intelligent solutions for VET issues.

Peramalan session website journal dilakukan untuk pendukung pengambilan keputusan dalam rangka meningkatkan kualitas dan nilai akreditasi pada website jurnal. Data sessions dianalisis berdasarkan pergerakan pola data time series menggunakan metode multilayer perceptron. Karakteristik yang dimiliki oleh multilayer perceptron yaitu keunggulan dalam penentuan nilai bobot yang lebih baik daripada metode lain, multilayer perceptron dapat digunakan tanpa pengetahuan sebelumnya dan algoritma dapat diimplementasikan dengan mudah serta mampu menyelesaikan masalah linear dan nonlinear sehingga nilai peramalan menjadi lebih baik. Penelitian menggunakan berbagai persentase data train dan test. Perbandingan data train dan test yang memiliki nilai terbaik adalah 80% data train dan 20% data test dengan learning rate 0.4 dan arsitektur 2-1-1. Hasil evaluasi model diperoleh nilai MSE dan RMSE, 0.015357 dan 0.123999 untuk training set serta, 0.018996 dan 0.137826 untuk MSE dan RMSE dari test set. Waktu eksekusi yang dibutuhkan untuk melakukan peramalan adalah 580.0651 second atau 9.667751 menit.

Journals play a pivotal role in disseminating scientific knowledge, housing a multitude of valuable research articles. In this digital age, the evaluation of journals and their quality is essential. The SCImago Journal Rank (SJR) stands as one of the prominent platforms for ranking journals, categorizing them into five index classes: Q1, Q2, Q3, Q4, and NQ. Determining these index classes often relies on classification methodologies. This research, drawing inspiration from the Cross-Industry Standard Process for Data Mining (CRISP-DM), seeks to employ the Random Forest method to classify journals, thus contributing to the refinement of journal ranking processes. Random Forest stands out as a robust choice due to its remarkable ability to mitigate overfitting, a common challenge in machine learning classification tasks. In the context of approximating SJR index classes, Random Forest, when utilizing the Gini index, exhibits promise, albeit with an initial accuracy rate of 62.12%. The Gini index, an impurity measure, enables Random Forest to make informed decisions while classifying journals into their respective SJR index classes. However, it is worth noting that this accuracy rate represents a starting point, and further refinement and feature engineering may enhance the model's performance. This research underscores the significance of machine learning techniques in the domain of journal classification and journal-ranking systems. By harnessing the power of Random Forest, this study aims to facilitate more accurate and efficient categorization of journals, thereby aiding researchers, academics, and institutions in identifying and accessing high-quality scientific literature. [ABSTRACT FROM AUTHOR]

This paper aims to explore the relationship between deep learning and forecasting within the context of the Sustainable Development Goals (SDGs). The primary objective is to systematically review 38 articles published between 2019 and 2023, following PRISMA guidelines, to understand the current landscape of deep learning forecasting for SDGs. Using data from 2019-2023 allows capturing the latest developments in deep learning forecasting for Sustainable Development Goals (SDGs), while excluding data before 2019 and after 2023 is based on the desire to avoid including potentially less relevant or unpublished research and to maintain focus on the most current and contextually relevant literature. The methodological approach involves analyzing the application of deep learning methods for forecasting within various SDG fields and identifying trends, challenges, and opportunities. The literature review results reveal the popularity of LSTM models, challenges related to data availability, and the interconnected nature of SDGs. Additionally, the study demonstrates that deep learning models enhance forecast accuracy and computational performance, as measured by Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and R-squared (R2). The findings underscore the importance of advanced data preparation techniques and the integration of deep learning with SDGs to improve forecasting outcomes. The novelty of this research lies in its comprehensive overview of the current landscape and its valuable insights for researchers, policymakers, and stakeholders interested in advancing sustainable development goals through deep learning forecasting. Finally, the paper suggests future research directions, including exploring the potential of hybrid forecasting models and investigating the impact of emerging technologies on SDG forecasting methodologies. Innovative methods for imputing missing values in deep learning forecasting models could be further explored to enhance predictive accuracy and robustness. [ABSTRACT FROM AUTHOR]

This research explores the vital role of ontology in learning forecasting in electrical engineering and informatics. As formally defined models of knowledge, ontologies are critical in organizing concepts for predictive learning. More than just an inquiry, our research reveals complex interconnections centered on Internet of Things (IoT) design, the semantic web, and knowledge modeling. Applications demonstrate the practical significance of ontologies in fostering intelligent connections, advancing information production, and improving interactions between computers, devices, and humans. This research introduces a comprehensive forecasting learning ontology to highlight the importance of ontologies in education, scientific inquiry, and developing systems for predictive analysis. Ontologies provide a structured framework for understanding the essence of predictive learning, encompassing key elements such as ideas, terminology, methodology, algorithms, data preprocessing, assessment, validation, data sources, application environments, interactions with technology, and learning processes. Emphasizing ontologies as indispensable instruments that drive technological development, our work underscores structured representation, semantic interoperability, and knowledge integration. In summary, this research improves the understanding of ontologies in forecasting by explaining practical applications and revealing new perspectives. Its unique contribution lies in its specific applications and natural consequences, laying the foundation for the future progress of ontology and learning forecasting, especially in educational contexts. [ABSTRACT FROM AUTHOR]