Garuda - Garba Rujukan Digital

Deep Learning Menggunakan Algoritma Xception dan Augmentasi Flip Pada Klasifikasi Kematangan Sawit Masaugi, Fathan Fanrita; Yanto, Febi; Budianita, Elvia; Sanjaya, Suwanto; Syafria, Fadhilah
KLIK: Kajian Ilmiah Informatika dan Komputer Vol. 4 No. 6 (2024): Juni 2024
Publisher : STMIK Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/klik.v4i6.1938

Palm oil is an important commodity in Indonesia, especially as Indonesia is the highest palm oil exporting country in the world. Ripe palm fruit is marked by a change in color of the fruit from black to reddish yellow. Apart from that, immature palm fruit has a negative and significant effect on CPO production. The data collection process was carried out by directly taking pictures of palm fruit on oil palm plantations and data obtained from Kaggle. The total amount of data is 1000 images and 1000 data resulting from flip augmentation. The Xception algorithm is an algorithm in deep learning which stands for Extreme version of Inception. This combination was then proven to provide better accuracy in classifying images from a dataset. The optimizer used is the optimizer in TensorFlow, namely Adam (Adaptive Moment Estimation) using learning rate and dropout values. Images of mature and immature palm oil were classified using the Xception algorithm with augmented and without augmented data. In addition, experiments were carried out by changing the parameter values ??of learning rate to 0.1, 0.01, 0.001 and dropout to 0.1, 0.01, 0.001. It was found that the data division was (90;10) with the best accuracy reaching 95%. Test parameters carried out by trialling were proven to increase accuracy when compared to without using parameters and flip augmentation. The best accuracy of the Xception model is 95% on augmented data with a learning rate of 0.001 and a dropout of 0.1.

Implementasi VGG 16 dan Augmentasi Zoom Untuk Klasifikasi Kematangan Sawit Mazdavilaya, T Kaisyarendika; Yanto, Febi; Budianita, Elvia; Sanjaya, Suwanto; Syafria, Fadhilah
KLIK: Kajian Ilmiah Informatika dan Komputer Vol. 4 No. 6 (2024): Juni 2024
Publisher : STMIK Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/klik.v4i6.1940

Indonesia is a country that has very abundant palm oil plantations and makes palm oil one of the largest export commodities in Indonesia. Fruit maturity on oil palms has a significant influence on palm oil and kernel production. The level of ripeness in palm oil fruit can affect several contents in it, such as tocopherol content, yield and FFA. The classification will be divided into 2 classes, namely between ripe and immature fruit with data on 500 images of ripe fruit and 500 images of immature fruit, data taken from the Kaggle site and private gardens taken using a cellphone camera. The data that has been obtained is augmented which is useful for enriching the data to make it more abundant. Data augmentation uses zoom augmentation and makes the original 1000 data increase to 2000 data. The model used is VGG 16 which is part of deep learning. The existing dataset is then preprocessed, resized and rescaled, then divides the data into 3, namely train, test and valid data. After dividing the data, then carry out the classification process with VGG 16 and set the hyperparameters after that the model will learn with 20 epochs. The model will learn with 57 schemes to compare and find highest accuracy. After the model has finished learning, it is evaluated using a confusion matrix. The results obtained were that the 90:10 data division using data augmentation with a learning rate of 0.01 and a dropout of 0.001 obtained the best accuracy, reaching 93.8%.

Optimalisasi Convolutional Neural Network Menggunakan Augmentasi dan Hyperparameter untuk Klasifikasi Daging Sapi dan Daging Babi Jasril, Jasril; Sanjaya, Suwanto
JUSTIN (Jurnal Sistem dan Teknologi Informasi) Vol 12, No 4 (2024)
Publisher : Jurusan Informatika Universitas Tanjungpura

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26418/justin.v12i4.80337

Tujuan penelitian ini adalah untuk menentukan model optimal pada klasifikasi sapi dan babi dengan menerapkan augmentasi data serta hyperparameter pada Convolutional Neural Network (CNN) arsitektur EfficientNet-B0. Data citra daging sapi dan daging babi yang diambil langsung dari beberapa pasar yang ada di kota Pekanbaru. Data diambil menggunakan kamera DSLR dan kamera smartphone dengan jarak antara 10cm sampai 15cm dan pencahayaan menyesuaikan dengan kondisi cahaya pada lingkungan pasar. Proses pelatihan dan pengujian model klasifikasi menggunakan beberapa skenario yaitu kombinasi pembagian data, jenis dataset, optimizer, fungsi aktivasi, dan learning rate. Berdasarkan hasil pengujian, model klasifikasi yang memiliki nilai akurasi tertinggi adalah 0,93 yaitu model dengan skenario jenis dataset gabungan (dataset original ditambah dengan dataset hasil augmentasi) dengan pembagian data 90% data latih dan 10% data uji. Hasil pengujian akurasi tertinggi menunjukkan model tidak overfitting, tetapi masih ada beberapa data citra daging sapi yang diklasifikasikan menjadi daging babi ataupun oplosan, sehingga perlu dilakukan penelitian lebih lanjut untuk meminimalkan masalah tersebut karena sebagai seorang muslim harus memastikan daging sapi yang dimakan adalah benar daging sapi.

Klasifikasi Sentimen pada Dataset Terbatas Menggunakan Random Forest dan Word2Vec Fitri, Dina Deswara; Agustian, Surya; Pizaini, Pizaini; Sanjaya, Suwanto
Journal of Computer System and Informatics (JoSYC) Vol 6 No 1 (2024): November 2024
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/josyc.v6i1.6246

Sentiment measurement of public opinion on social media is essential for understanding societal views on various issues, including public figures and political events. This research explores the effectiveness of the Random Forest algorithm with Word2Vec-based word representation for sentiment classification on a limited dataset. The case study involves tweets regarding Kaesang Pangarep as the Chairman of PSI, supplemented by external data related to Covid-19 and general topics. The dataset was processed using cleaning techniques, case folding, stopword removal, stemming, and tokenization. Words in the dataset were represented using the Word2Vec model with a Continuous Bag of Words (CBOW) architecture and a vector dimension of 500. Random Forest was employed to classify sentiment into positive, negative, or neutral categories. In the initial phase, the model was trained using 300 samples per label; however, the results showed unsatisfactory performance with an F1-Score of 49.00% and an accuracy of 50.00%. To improve performance, the dataset was expanded by adding 900 samples from Kaesang and 1,080 samples from external topics. The final results indicated an improvement with an F1-Score of 49.89%, an accuracy of 58.29%, precision of 49.16%, and recall of 56.47%. This research confirms that the use of Random Forest with word representation from Word2Vec can enhance sentiment classification performance, even with a limited dataset, and contributes to the development of sentiment analysis techniques in the field of machine learning.

Klasifikasi Kelayakan Air Minum dengan Backpropagation Neural Network Berbasis Penanganan Missing Value dan Normalisasi Kurniawan, Saifur Yusuf; Sanjaya, Suwanto; Vitriani, Yelfi; Afrianty, Iis
Journal of Information System Research (JOSH) Vol 6 No 1 (2024): Oktober 2024
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/josh.v6i1.5871

The issue of drinking water quality and its suitability for human consumption represents a significant concern in contemporary society, particularly in the context of maintaining public health. The existing research on the classification of drinking water eligibility has yet to yield conclusive results. The objective of this research is to utilize the backpropagation neural network method to categorize drinking water feasibility data, thereby ensuring that the water consumed meets established safety standards. The data utilized in this study were obtained from an open repository and encompass a total of 3,276 data points. The data set comprises nine water quality parameter attributes, namely pH, hardness, solids, chloramines, sulfate, conductivity, organic carbon, trihalomethanes, and turbidity. The data underwent a series of pre-processing steps, including the removal of missing values, the replacement of missing values with the average value of the attribute, and normalization using the MinMax Scaler and Z-score methods. The artificial neural network architecture comprises three principal components: input, hidden, and output neurons. The optimal architecture scenario is [9; 17; 15; 10; 1], comprising nine input neurons, 17 neurons in the initial hidden layer, 15 neurons in the second hidden layer, 10 neurons in the third hidden layer, and a single output neuron. The evaluation results demonstrate that this model effectively classifies drinking water eligibility data with an accuracy rate of 0.6579. However, the results indicate that the accuracy achieved requires further improvement for more reliable applications. These findings illustrate the promising potential of the BPNN method in classifying drinking water quality data.

Pengembangan Aplikasi Pendeteksi Daging Sapi dan Babi Menggunakan Deep Learning Arsitektur EfficientNet-B6 Berbasis Android Pangestu, Yoga; Sanjaya, Suwanto; Jasril; Agustian, Surya; Safaat, Nazruddin
Jurnal Informatika Ekonomi Bisnis Vol. 7, No. 2 (June 2025)
Publisher : SAFE-Network

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37034/infeb.v7i2.1195

The advancement of digital technology has generated a demand for applications that assist the public in ensuring the halal status of food products, particularly in distinguishing between beef and pork. This study aims to develop an Android-based application for detecting beef and pork using Deep Learning methods with the EfficientNet-B6 architecture, employing the eXtreme Programming software development approach. The image classification model utilizes a Convolutional Neural Network architecture integrated into a Python-based server, while the user interface is developed with Java in Android Studio. System testing was conducted using black-box methods on several Android devices, with varying room conditions and meat types. The results show that the application can classify meat with an accuracy of 66.7%, considering room conditions such as light and dark environments, and meat types including fatty and non-fatty. This application provides fast response times and a user-friendly interface. This application is expected to enable users to independently and efficiently verify the halal status of meat, thereby supporting the needs of Muslim consumers in the digital era.

Penggunaan Convolutional Neural Network NASNetLarge Dalam Klasifikasi Citra Daging Babi dan Sapi Aqilah, M Alfandri; Jasril; Sanjaya, Suwanto; Insani, Fitri
Bulletin of Computer Science Research Vol. 5 No. 4 (2025): June 2025
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bulletincsr.v5i4.666

The adulteration of beef with pork is a serious issue in Indonesia, particularly for Muslim consumers who are required to consume halal products. According to a Kompas (2020) report, a case of meat adulteration involving 100 kilograms of mixed meat sold as beef was discovered in Tangerang City. This practice not only violates religious laws but also poses threats to public health and consumer trust. To address this challenge, this study adopts a deep learning approach using NASNetLarge for the classification of pork, beef, and mixed meat images. Unlike previous research that utilized EfficientNet-B2 and achieved an accuracy of 98.23%, this study’s NASNetLarge approach produced a comparably competitive accuracy of 98.03%. The dataset used consists of 1,932 images sourced from the Kaggle platform, which were processed through preprocessing and augmentation stages. The data were then split into two distribution scenarios: the entire dataset and a balanced class dataset with 90:10 and 80:20 ratios. Evaluation results show that the best parameter combination was achieved in the first scenario with a 90:10 ratio using augmented images, a learning rate of 0.001, 128 dense units, and the Adam optimizer. The model achieved the highest accuracy of 98.03%, with a precision of 98.63%, recall of 98.40%, and an F1-score of 98.50%. These results indicate that NASNetLarge is effective in accurately and consistently classifying meat images. Image augmentation significantly improved model performance, and the 90:10 data ratio yielded more optimal results compared to 80:20. These findings have the potential to support food surveillance efforts by enabling rapid and accurate detection of meat adulteration.

Optimasi Hyperparameter Deep Learning untuk Deteksi X-Ray Paru-Paru Menggunakan Bayesian Optimization Shahira, Fayza; Negara, Benny Sukma; Yanto, Febi; Sanjaya, Suwanto
JIEET (Journal of Information Engineering and Educational Technology) Vol. 9 No. 1 (2025)
Publisher : Universitas Negeri Surabaya

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26740/jieet.v9n1.p53-63

Penyakit paru-paru, seperti pneumonia dan COVID-19, merupakan ancaman serius terhadap kesehatan masyarakat, terutama jika diagnosisnya mengalami keterlambatan. Pendekatan deteksi dini melalui citra X-ray dada banyak digunakan, namun akurasinya sangat bergantung pada kemampuan sistem klasifikasi. Penelitian ini bertujuan untuk meningkatkan performa klasifikasi citra X-ray paru-paru dengan mengimplementasikan metode deep learning menggunakan arsitektur ResNet-101 yang dioptimasi menggunakan teknik Bayesian Optimization. Dataset yang digunakan dalam penelitian ini terdiri dari tiga kelas yaitu Normal, Pneumonia, dan COVID-19, masing-masing sejumlah 1.000 citra. Kinerja model hasil optimasi dibandingkan dengan model baseline pada tiga skenario split data yaitu 90:10, 80:20, 70:30. Hasil penelitian mengindikasikan bahwa model yang telah dioptimasi mampu meningkatkan performa pada seluruh metrik evaluasi mencakup akurasi, presisi, recall, spesifisitas, dan F1-score. Akurasi tertinggi tercatat sebesar 93,83% pada skenario 80:20, melampau akurasi baseline yang sebesar 91,83. Selain itu, kurva akurasi dan loss menunjukkan proses training yang stabil dan konvergen secara cepat tanpa indikasi overfitting yang signifikan. Penerapan Bayesian Optimization terbukti efektif dalam menemukan konfigurasi hyperparameter optimal yang berdampak pada peningkatan dalam tiap metrik evaluasi

Pengembangan Aplikasi Pendeteksi Daging Sapi dan Babi Menggunakan Deep Learning Arsitektur EfficientNet-B6 Berbasis Android Pangestu, Yoga; Sanjaya, Suwanto; Jasril; Agustian, Surya; Safaat, Nazruddin
Jurnal Informatika Ekonomi Bisnis Vol. 7, No. 2 (June 2025)
Publisher : SAFE-Network

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37034/infeb.v7i2.1195

The advancement of digital technology has generated a demand for applications that assist the public in ensuring the halal status of food products, particularly in distinguishing between beef and pork. This study aims to develop an Android-based application for detecting beef and pork using Deep Learning methods with the EfficientNet-B6 architecture, employing the eXtreme Programming software development approach. The image classification model utilizes a Convolutional Neural Network architecture integrated into a Python-based server, while the user interface is developed with Java in Android Studio. System testing was conducted using black-box methods on several Android devices, with varying room conditions and meat types. The results show that the application can classify meat with an accuracy of 66.7%, considering room conditions such as light and dark environments, and meat types including fatty and non-fatty. This application provides fast response times and a user-friendly interface. This application is expected to enable users to independently and efficiently verify the halal status of meat, thereby supporting the needs of Muslim consumers in the digital era.

PENGELOMPOKAN DATA KONDISI MESIN SCREW PRESS MENGGUNAKAN ALGORITMA FUZZY C-MEANS Jasril, Jasril; Al Fiqri, M. Faiz; Sanjaya, Suwanto; Handayani, Lestari; Insani, Fitri
Information System Journal Vol. 8 No. 01 (2025): Information System Journal (INFOS)
Publisher : Universitas Amikom Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24076/infosjournal.2025v8i01.2133

Kinerja mesin screw press sangat memengaruhi efisiensi dan kualitas produksi minyak kelapa sawit. Salah satu komponen penting dalam sistem ini adalah Back Pressure Vessel (BPV) yang menyalurkan uap ke berbagai stasiun proses. Penelitian ini bertujuan untuk mengelompokkan kondisi mesin berdasarkan temperatur dan tekanan menggunakan algoritma Fuzzy C-Means (FCM). Data yang dianalisis berasal dari mesin BPV PT. XYZ periode April–Mei 2024 sebanyak 23.002 entri. Tahapan penelitian meliputi seleksi data, pra-pemrosesan, normalisasi Min-Max Scaler, klasterisasi FCM, dan evaluasi menggunakan metode Elbow dan Davies-Bouldin Index (DBI). Hasil awal menunjukkan tiga klaster dengan distribusi kondisi mesin dari stabil hingga memerlukan perawatan. Metode Elbow menunjukkan jumlah klaster optimal sebanyak empat, sedangkan DBI menunjukkan dua klaster dengan nilai terbaik 0,389. Hasil ini menunjukkan bahwa FCM mampu mengelompokkan kondisi mesin secara efektif dan dapat digunakan sebagai dasar dalam pengambilan keputusan perawatan. Penelitian ini disarankan untuk dikembangkan dengan atribut tambahan.

Title

Found 24 Documents
Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title Search

Found 24 Documents Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title

Found 24 Documents
Search