Manual classification of sleep stage is a time-consuming but necessary step in the diagnosis and treatment of sleep disorders, and its automation has been an area of active study. The previous works have shown that low dimensional fast Fourier transform (FFT) features and many machine learning algorithms have been applied. In this paper, we demonstrate utilization of features extracted from EEG signals via FFT to improve the performance of automated sleep stage classification through machine learning methods. Unlike previous works using FFT, we incorporated thousands of FFT features in order to classify the sleep stages into 2–6 classes. Using the expanded version of Sleep-EDF dataset with 61 recordings, our method outperformed other state-of-the art methods. This result indicates that high dimensional FFT features in combination with a simple feature selection is effective for the improvement of automated sleep stage classification.

The eyewitness message on twitter as a social network sensor aims to determine the classification process's performance. In the classification of flood disaster messages, preprocessing data is required before the classification process is carried out. Preprocessing affects the resulting level of accuracy in the classification process. Stopword removal is part of preprocessing so that the effect of stopword removal on classification performance will be examined. Support Vector Machine (SVM) is used to classify by weighting words using Term Frequency-Inverse Document Frequency (TF-IDF). The data taken from Twitter is 3000 data with 1000 data each for each label. The effect of the stopword on accuracy performance can be seen in several experiments that have been carried out. We had already conducted three different experiments, and the highest level of accuracy was 76.6%, 76.87%, and 77.87%. Based on the experiments that have been carried out, stopword is very influential on the accuracy generated by the classification of flood disaster messages on Twitter.

This study examined the tar characteristic of wood pellet using a downdraft gasifier system. The wood pellet was used as a feedstock with a variation of moisture content between 2 to 6 wt. %. Tar sample was taken using the JP 2009-40885 method. Gas chromatography mass spectrometry and gravimetric analysis were used to identify and to analyze the tar sample, respectively. The results show that an increase in moisture content slightly decreased the total concentration of tar formation on the nine tar compounds of the biomass. Additionally, the lowest moisture content produced the highest amount of light tar on tar classification. The amount of light tar indicates the ease of ignition in the gasifier. Therefore, this condition will minimize the failure-of-equipment problem because of no clogs in the gasifier equipment. The lowest moisture content had the lowest tar density, and this minimized the deposit of tar in the gasifier. Therefore, the lowest moisture content had low tar production in the biomass gasification process using wood pellets.

Indonesia is an ASEAN country with the most motorcycle users, where one-third of its population own motorcycles. The automatic motorcycle is the most widely used type due to its agility and fuel-efficient abilities. Sudden motorcycle damage could hamper the users' activities. However, most of these users do not know the reason for the damage. This paper presents the development of expert system for diagnosing the damage of motorcycle using forward chaining method. This system was implemented in mobile application. Through a mobile application, a solution for these users can be obtained. This application immediately discovers the damaged location and repair process. Furthermore, it acts as the first solution before motorcycle repair is carried out in a shop. In this study, the forward chaining method was implemented. It is based on a pattern-matching algorithm whose primary objective is to match facts (input data) with appropriate rules from the rule base. Various test results showed that the diagnostic application used for automatic motorcycle damages 100% worked.

Musibah bencana alam banjir merupakan salah satu musibah alam yang sangat umum kejadian di Indonesia baik di pulau Jawa ataupun di luar Pulau Jawa. Sumber informasi dari media social Twitter dapat digunakan sebagai Social Network Sensor (SNS) untuk bencana banjir. Penelitian ini bertujuan untuk melakukan klasifikasi pesan twitter khususnya tentang bencana banjir sebagai hasil dari algoritma machine learning. Pada klasifikasi pesan bencana banjir dilakukan serangkaian proses data preprocessing dan dilanjutkan dengan ekstraksi fitur dan pembobotan kata dari data twitter. Berbagai teknik yang dilakukan pada proses data preprocessing berpengaruh terhadap tingkat akurasi hasil klasifikasi yang dihasilkan. Selanjutnya, penelitian ini menyampaikan bahwa algoritma Support Vector Machine (SVM) adalah sebuah algoritma dapat dimanfaatkan untuk klasifikasi data ke dalam tiga (3) kelas kelas saksi mata, kelas non saksi mata dan kelas tidak diketahui. Total jumlah data yang telah diolah sebanyak 3000 data untuk 3 kelas label data. Pendekatan metode One Versus One (OVO) untuk dua kelas label data dan One Versus All (OVA) untuk lebih dari dua kelas label pada algoritma SVM. Pada penelitian ini telah diterapkan algoritma Support Vector Machine untuk multiclass dengan metode OVA. Hasil eksperimen telah menunjukkan pendekatan OVA pada algoritma SVM dengan kernel RBF menghasilkan evaluasi performansi yang paling tinggi. Hasil evaluasi tersebut adalah nilai akurasi hingga 87.03%. Sedangkan hasil riset sebelumnya dengan algoritma SVM tetapi pendekatan yang berbeda hanya mencapai nilai akurasi hingga 77.87%.

The inspection of packaging defects is a crucial aspect of maintaining the quality of industrial production, especially in the case of boxed products. This study introduces a novel approach for detecting physical defects in product packaging boxes by integrating image processing with deep learning, specifically transfer learning with two images as an input. The proposed method utilizes both top view and side view images of the packaging to determine its condition, a significant departure from the conventional single image input. Our approach incorporates 16 pre-trained model variants from EfficientNetV2, MobileNetV3, and ResNetV2 for transfer learning as feature extractors. The experimental findings demonstrate that the best model that leverages EfficientNetV2 variant achieves 100% accuracy and F1 score in terms of classification performance. However, the most optimal model in terms of classification performance and inference speed was the one that leveraged ResNetV2 variant. This model scored 95% accuracy and 95.24% F1 score, with an inference speed of 91 ms per prediction.