Information of 3D protein structure plays an important role in various fields, including health, biotechnology, biomaterials, and so on. Knowledge of the 3D structure of proteins will help in understanding the interactions that can take place with other molecules such as drug molecules and other effector molecules. To get the 3D protein structure, it is generally carried out experimentally using x-ray diffraction and NMR (Nuclear Magnetic Resonance) methods. The experimental method requires a relatively long time and expertise to handle the completion of the structure. Until now, not all protein structures can be determined because the level of complexity varies from one protein to another. One approach is to use machine learning that leverages evolution information and deep learning method. The results given can improve the accuracy of the prediction compared to the conventional approach. However, the level of accuracy is still influenced by the number of homologous proteins in the database. Therefore, this study propose to replace the process of extracting the evolutionary information from Multiple Sequence Analysis (MSA) into Transformer-based self-supervised pre-trained. To test these changes, an experiment was carried out on a 3D protein structure prediction model based on Long Short-Term Memory (LSTM) and Universal Transformer. The proposed method results show a decrease in the value of distance Root-Mean Square Deviation (dRMSD) of 0.561 Angstrom and Root Mean Square Error (RMSE) torsion angle of 0.11 degree on the Universal Transformer predictor. The results of the T-Test show that the decrease in the two indicators shows a significant result. Therefore, pre-trained data can be used as an evolutionary information only for the Universal Transformer predictor.

Not all topics within a knowledge graph are represented at the same depth, which might lead to biased conclusions derived from the graph. Tools have been created in an effort to highlight knowledge imbalances, one of which is ProWD, built to analyze imbalances in the Wikidata knowledge graph. However, as often found, the usability aspect of Semantic Web tools is commonly overlooked, resulting in their limited acceptance. In this paper, we utilize the standard approach to improve usability, i.e., the User-Centered Design (UCD), for ProWD. We employ the full range of steps of UCD to improve the user experience of ProWD despite the complex nature of knowledge graph concepts underlying ProWD. The result of the ProWD redesign is then evaluated using the System Usability Scale (SUS) and User Experience Questionnaire (UEQ) scores, as well as the task success rate and completion time, suggesting that the overall usability of ProWD has successfully improved.

To capture cultural heritage is to capture the experience of people who are directly involved in creating, witnessing, and maintaining cultural heritage objects. Ideally, the people accessing digital representations of cultural heritage objects are able to understand the significance underlying the objects. The question is how to capture (the experience of) cultural objects in digital form. Various modalities exist for representing cultural heritage: unstructured textual data, possibly including images or videos, as well as structured data. To illustrate our approaches, we pick one cultural heritage representative: rendang, one of the five national dishes of Indonesia, believed to have existed as early as the 15th century. From textual sources, we may learn about rendang and its history. According to Nurmufida et al.,8 rendang is a traditional cuisine originating from West Sumatra, with beef and coconut milk as its main ingredients. From image sources, we may see what rendang looks like. For example, Wikimedia Commons contains images of rendang that show that, despite being similar to curry, rendang is actually drier. Next, we may wonder how to cook rendang. A simple YouTube search provides a wide selection of videos showing how to cook rendang, with the chefs ranging from local Indonesians to international chefs. In fact, the word rendang originates from how it is cooked; that is, slowly (merandang, in the Minangnese language).

An earthquake is a sudden, rapid shaking of the ground caused by the shifting of the Earth's tectonic plates. Earthquakes pose serious threats that cause economic losses and casualties. To mitigate such risks, it is crucial to better understand earthquakes through data-driven analysis. In this paper, we propose an approach to time series analysis over earthquake data, consisting of two steps: exploration and prediction. The exploration step relies on exploratory data analysis (EDA) comprising descriptive statistics and data visualization, whereas the prediction step focuses on how to predict the number of earthquakes for the following years. We perform our time series analysis using various machine learning techniques over a global earthquake dataset from 1965-2016 and report insights as well as lessons learned from the study.

Recommendation systems have become a staple feature for any e-commerce sites. The ability to predict whether a customer likes an unseen product forms the very foundation of a recommendation system. In this paper, we concern the issue of explicit rating prediction over cosmetic products. Given a dataset of cosmetic product ratings, we analyze the characteristics of the dataset and implement a wide range of algorithms, such as KNN and matrix factorization, to predict such ratings. We evaluate the performance of these algorithms using MAE and RMSE measures, and discuss factors that may contribute to their performance results. Our experiments have shown that the SVD++ technique performs the best among all with an MAE of 0.7699 and an RMSE of 0.9696. We hope that our paper can shed new light on the selection of explicit rating prediction algorithms not only in the domain of beauty products, but also in wider scenarios.

The increasing number of Commuter Line passengers calls for innovation as to how the crowd density across the carriages of Commuter Line trains can be better distributed. We develop an IoT system to detect the crowd density of Commuter Line trains so that (incoming) passengers can be better informed regarding which carriage is best to get in, hence ameliorating the train density distribution. We investigate two different approaches for density detection: CNN and YOLO+KNN. Moreover, we also analyze the impact of different single-board computers, that is, Raspberry Pi 3B and NVIDIA Jetson Nano, and that of different camera angle settings. In total, there are 20 different scenario combinations. We comparatively evaluate the density detection performance as well as the business value for each scenario.

Fall is the second leading cause of accidental injury and death worldwide. This event often occurs in the elderly and the frequency is increasing every year. Reliable fall activity detection system can reduce the risk of injuries suffered. Since falls are unwanted events or occur suddenly, it is difficult to collect actual fall data. It is also difficult because of the similarity to some activities such as squatting and picking up objects from the floor. In addition, in recent years the publicly available fall datasets are limited. Therefore, in 2019, some researchers tried to create a comprehensive fall dataset that simulates the actual events using camera and sensor devices. The experiment produced a multimodal dataset UP-Fall. Using this dataset, this work tries to detect falling activity using Convolutional Neural Network and Long Short-Term Memory approaches. CNN is used to detect spatial information from image data, while LSTM is used to exploit temporal information from signal data. Then, the results of the two models are combined with the majority voting strategy. Based on the evaluation results, CNN obtained an accuracy of 98.49% and LSTM 98.88%. Both models contribute to the performance of the majority voting strategy with the result that the accuracy (98.31%) exceeds baseline accuracy (96.4%). Other evaluation metrics also improved such as precision goes up to 11%, recall 14%, and F1-score 12% in comparison with baseline.