Claim Missing Document
Check
Articles

Found 6 Documents
Search

PREDIKSI POTENSIAL GEMPA BUMI INDONESIA MENGGUNAKAN METODE RANDOM FOREST DAN FEATURE SELECTION Tantyoko, Henri; Sari, Dian Kartika; Wijaya, Andreas Rony
IDEALIS : InDonEsiA journaL Information System Vol 6 No 2 (2023): Jurnal IDEALIS Juli 2023
Publisher : Universitas Budi Luhur

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.36080/idealis.v6i2.3036

Abstract

Gempa bumi adalah suatu peristiwa alamiah yang terjadi saat terjadi pelepasan energi secara tiba-tiba dalam kerak bumi, mengakibatkan getaran dan guncangan pada permukaan bumi. Gempa bumi merupakan salah satu bencana alam yang dapat menyebabkan kerusakan fisik yang besar, dampak ekonomi yang signifikan, dan hilangnya nyawa manusia. Beberapa penyebab gempa bumi antara lain aktivitas tektonik lempeng bumi, pergerakan lempeng tektonik, dan deformasi kerak bumi. Untuk mengurangi jumlah korban jiwa, perlu dilakukan prediksi kapan gempa bumi akan terjadi di suatu wilayah. Salah satu cara untuk memprediksi ialah dengan menggunakan metode Machine Learning yaitu Random Forest (RF), metode ini memanfaatkan beberapa pohon keputusan yang selanjutnya dilakukan voting untuk menentukan keputusan akhir prediksi . Model yang baik adalah model yang menghasilkan kesalahan seminimal mungkin. Oleh karena itu, penulis melakukan skema seleksi fitur untuk mengolah fitur-fitur yang memiliki korelasi yang kuat. Prediksi menggunakan RF dengan seleksi fitur menghasilkan F1 score sebesar 92.23%, yang lebih baik 5.02% dibandingkan tanpa menggunakan seleksi fitur. Metode RF + Seleksi Fitur ini juga jauh lebih baik jika dibandingkan metode machine learning tradisional lainnya seperti SVM, Naïve Bayes, dan Decision Tree.
Hybrid Algoritma Vgg16-Net Dengan Support Vector Machine Untuk Klasifikasi Jenis Buah dan sayuran Dwi Putro, Aditya; Tantyoko, Henri
JTIM : Jurnal Teknologi Informasi dan Multimedia Vol 5 No 2 (2023): August
Publisher : Puslitbang Sekawan Institute Nusa Tenggara

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35746/jtim.v5i2.335

Abstract

Arsitektur deep learning VGG16 terbukti efektif dalam hal melakukan klasifikasi citra pada dataset ImageNet, akan tetapi memiliki keterbatasan dalam jumlah parameter sangat banyak dan potensi overfitting pada dataset kecil. SVM memiliki kelebihan dalam hal menangani masalah overfitting pada dataset yang relatif kecil, sementara VGG16 memiliki keunggulan dalam mengekstraksi fitur yang berkualitas dari citra dengan performa yang sangat baik. SVM juga dapat membantu memperbaiki kinerja klasifikasi pada VGG16 dengan meminimalkan risiko overfitting dan meningkatkan akurasi klasifikasi pada dataset yang relatif kecil. Oleh karena itu, penulis memilih untuk hybrid algoritma VGG16Net Dengan Support Vector Machine Untuk Klasifikasi Jenis buah dan sayuran, yang nantinya arsitektur VGG16 digunakan untuk ekstraksi fitur dari citra dan fitur-fitur tersebut dijadikan input untuk SVM. Keputusan menggunakan VGG16 digabungkan dengan SVM adalah untuk meningkatkan akurasi klasifikasi dataset citra buah dan sayuran, Namun, penggunaan SVM membutuhkan pemilihan parameter yang tepat dan teknik prapemrosesan data yang tepat untuk mencapai hasil yang baik. Dan dalam penelitian ini penulis berhasil mengklasifikasikan citra buah dan sayuran, akurasi sebelum hybrid svm mendapatkan 94.52% training accuracy dan testing (validation) accuracy sebesar 87.85%. dan hasil loss mendapat training loss sebesar 0.58 dan testing loss accuracy sebesar 12.5%. Setelah dilakukan hybrid vgg16 dengan svm didapatkan training accuracy sebesar 99.87 % dan testing (validation) accuracy sebesar 91.76 %. Untuk hasil loss mendapat training loss sebesar 0.13 dan testing loss accuracy sebesar 8.24%. Oleh karena itu, arsitektur CNN VGG-16Net digabungkan dengan SVM dapat menghasilkan model klasifikasi yang baik, terutama pada dataset yang relatif kecil dan dapat menjadi pilihan yang sesuai dalam klasifikasi citra.
Comparison of the Word2vec Skipgram Model Method Linkaja Application Review using Bidirectional LSTM Algorithm and Support Vector Machine Ayuningtyas, Puji; Tantyoko, Henri
JUSTIN (Jurnal Sistem dan Teknologi Informasi) Vol 12, No 1 (2024)
Publisher : Jurusan Informatika Universitas Tanjungpura

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26418/justin.v12i1.72530

Abstract

Word embedding is a phase in word processing that seeks to convert each word into a vector representation. Word2Vec is a sort of word embedding that is frequently utilized in natural language processing research. Choosing the proper algorithm can help increase the performance of the word embedding method while doing text data categorization tasks. This research uses the Bidirectional LSTM deep learning algorithm and the Support Vector Machine (SVM) machine learning algorithm. The crawling approach was used to obtain data by accessing the LinkAja Application ID on the Google Play Store. The total number of rows in the dataset was 35560. Labeling data involves categorizing it into two target classes: positive (score 1) and negative (score 0). This study employs the Word2Vec approach with skipgram architecture during the vectorization stage. Vector size, window, min count, and sg are the four parameters employed. The bidirectional LSTM architecture employs a sequential model that consists of three neural network layers: embedding, bidirectional, and dense. In the meanwhile, the SVM architecture employs the Radial Basis Function (RBF) kernel parameters. For the final stage of algorithm testing, the accuracy of the bidirectional LSTM (BiLSTM) algorithm was 0.9505, which means it was higher than the support vector machine (SVM) algorithm with an accuracy value of 0.93.
Classification of Real and Fake Images Using Error Level Analysis Technique and MobileNetV2 Architecture Baihaqi, Muhamad Nur; Sugiharto, Aris; Tantyoko, Henri
Jurnal Masyarakat Informatika Vol 16, No 1 (2025): May 2025
Publisher : Department of Informatics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/jmasif.16.1.73283

Abstract

Advancements in technology have made image forgery increasingly difficult to detect, raising the risk of misinformation on social media. To address this issue, Error Level Analysis (ELA) feature extraction can be utilized to detect error level variations in lossy-formatted images such as JPEG. This study evaluates the contribution of ELA features in classifying authentic and forged images using the MobileNetV2 model. Two scenarios were tested using the CASIA 2.0 dataset: without ELA and with ELA. Fine-tuning was performed to adapt the model to the new problem. Experimental results show that incorporating ELA improves model accuracy up to 93.1%, compared to only 76.41% in the scenario without ELA. Validation using k-fold cross-validation yielded a high average f1-score of 96.83%, confirming the effectiveness of ELA in enhancing the classification performance of authentic and forged images.
A Comparative Study of Machine Learning Models for Short-Term Load Forecasting Vianita, Etna; Tantyoko, Henri
Jurnal Masyarakat Informatika Vol 16, No 1 (2025): May 2025
Publisher : Department of Informatics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/jmasif.16.1.73130

Abstract

Short-Term Load Forecasting (STLF) was a critical task in power system operations, enabling efficient energy management and planning. This study presented a comparative analysis of five machine learning models namely XGBoost, Random Forest, Multi-Layer Perceptron (MLP), Support Vector Regression (SVR), and LightGBM using real-world electricity demand data collected over a four-month period. Two modeling approaches were explored: one using only time-based features (hour, day of the week, month), and another incorporating historical lag features (lag_1, lag_2, lag_3) to capture temporal patterns. The results showed that MLP with lag features achieved the best performance (RMSE: 57.63, MAE: 34.54, MAPE: 0.22), highlighting its ability to model nonlinear and sequential dependencies. In contrast, SVR and LightGBM experienced performance degradation when lag features were added, suggesting sensitivity to feature dimensionality and data volume. These findings emphasized the importance of model-feature alignment and temporal context in improving forecasting accuracy. Future work could explore the integration of external variables such as weather and holidays, as well as the application of advanced deep learning architectures like LSTM or hybrid models to further enhance robustness and generalizability.
An Efficient Bidirectional Gated Recurrent Unit Approach for Student Study Duration Modeling and Timely Graduation Forecasting Purnama, Satriawan Rasyid; Tantyoko, Henri; Vianita, Etna
Jurnal Masyarakat Informatika Vol 16, No 2 (2025): Issue in Progress
Publisher : Department of Informatics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/jmasif.16.2.73275

Abstract

Delays in student graduation remain a persistent challenge in higher education, with approximately 28% of students requiring more than four years to complete their studies, exceeding the standard duration. This study addresses the issue by proposing a predictive model to estimate students’ graduation year using a Bidirectional Gated Recurrent Unit (BiGRU) neural network. The model is trained on a combination of academic and financial indicators, including Grade Point (GP) scores from the first to the fifth semester, cumulative Grade Point Average (GPA), and the single tuition fee tier (UKT). The integration of these features allows the model to learn temporal patterns in students’ academic progression and financial capacity. Empirical analysis reveals that students in the UKT 8 group consistently demonstrate superior academic performance, as evidenced by their higher average GPA across semesters, compared to students in lower UKT groups. The BiGRU model achieves a Mean Absolute Percentage Error (MAPE) of 9.5%, indicating high predictive accuracy. These findings highlight the potential of deep learning models, particularly BiGRU, in forecasting academic outcomes. Furthermore, the insights generated from this model can serve as a valuable tool for universities in formulating targeted academic interventions and policies aimed at promoting timely graduation and reducing dropout rates.