JISKA (Jurnal Informatika Sunan Kalijaga)

Perbandingan Kinerja Naïve Bayes dan Random Forest dalam Mendeteksi Berita Palsu

2024-02-01T14:58:40+07:00

Fake news has become a serious problem in today's digital era. The existence of fake news can have various negative impacts, including the spread of misinformation, social unrest, and economic losses. This study compares the performance of Naïve Bayes and Random Forest classification methods in detecting fake news. Both methods were evaluated on a news dataset comprising 44,898 samples. It uses public data from the Kaggle repository. The news samples are represented by four features: title, news content, subject, and news date. This data is then subjected to cleaning, stemming, tokenization, and feature extraction. The results indicate that the Random Forest method outperforms the Naïve Bayes method. The Random Forest method has an accuracy of 99%, while the Naïve Bayes method has an accuracy of 96%. In general, this research demonstrates that the Random Forest method can be a viable alternative for detecting fake news.

Analisis Cluster untuk Pengelompokan Kemampuan Penguasaan ICT Menggunakan K-Means dan Autoencoder

2024-05-14T10:03:09+07:00

Information and Communication Technology (ICT) skills are essential in today's digital age. However, numerous new students possess varying levels of ICT proficiency and may lack the necessary skills expected by universities. ICT training is essential for enhancing students' ICT skills. Nevertheless, delivering the same training to all students proves to be less effective. Therefore, grouping students' ICT skills is crucial to ensure that the training provided aligns with the fundamental abilities of the students. Cluster analysis is a common method for grouping data. This study employs k-Means and autoencoder for cluster analysis, with autoencoder utilized to reduce data dimensions and k-Means to perform the clustering process. The Elbow method is utilized to identify the ideal number of clusters. The optimal number of clusters determined was three clusters. Model evaluation was conducted using the Silhouette coefficient and the Davies-Bouldin Index (DBI). The evaluation results revealed that the combination of k-Means and autoencoder yields superior performance compared to using k-Means alone, as evidenced by a higher Silhouette value and a lower DBI value.

Analisis Sentimen Ulasan Pengguna Aplikasi Alfagift Menggunakan Random Forest

2024-11-05T11:24:04+07:00

Alfagift is a mobile application developed by Alfamart to support online ordering, with features such as promos, transactions, ordering, and delivery from the nearest point according to the consumer's address. User feedback on the Google Play Store shows mixed sentiments, including both positive and negative responses, which can be utilized by developers as material to improve the quality of the application. This study focuses on assessing the sentiment of Alfagift app user reviews through the application of the Random Forest algorithm. A total of 4,379 review data was collected from the Google Play Store and grouped into two categories, namely positive and negative sentiment. The research steps include data collection, data labeling, data preprocessing, word weighting, data division into training and testing sets, Random Forest algorithm implementation, and model evaluation. The test results show that the Random Forest algorithm achieves an accuracy of 97.6% and an AUC of 0.98, which falls into the category of excellent classification. This research is expected to contribute to application developers in understanding user perceptions, so as to improve application quality and increase overall user convenience.

Imputasi Data Temperatur Maksimum Menggunakan Metode Support Vector Regression

2024-05-24T20:30:14+07:00

Temperature is a crucial element affecting various aspects, from agriculture to natural disasters. Temperature data imputation is also important because, in some cases, temperature data is not always complete. This study aims to predict missing temperature data in the East Nusa Tenggara (NTT) region using the Support Vector Regression (SVR) method. The data used comes from six BMKG observation stations in NTT and ERA-5 Reanalysis data. The choice of the SVR method is based on its ability to handle data with complex structures. Modeling is conducted separately for each station using the Radial Basis Function (RBF) kernel. Model evaluation employs the metrics Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R²), presenting the evaluation results with low error. The results show that among several parameter tests, the parameter ranges [C = 1, 5, 10, 15], [ε = 0,1, 0,3, 0,6, 0,9], and [γ = 1, 5, 10, 15] in the SVR method are the best parameter ranges across all stations. The prediction graphs display different temperature fluctuation patterns at each station. This study contributes to enhancing the availability of accurate climate data to support sustainable decision-making in the NTT region.

Perbandingan Random Forest dan Convolutional Neural Network dalam Memprediksi Peralihan Pelanggan

2024-07-22T16:03:04+07:00

The rapid growth of the telecommunications industry has increased competition among companies for customers. As a result, customers often switch services or terminate subscriptions. Retaining customers is very important as it is 10 times cheaper than acquiring new customers. This study compares Random Forest (RF) and Convolutional Neural Network (CNN) algorithms in predicting customer switching, using Correlation-based Feature Selection (CFS) and Recursive Feature Elimination (RFE) for data partitioning. Model evaluation using Confusion Matrix and Area Under Curve (AUC). The evaluation results show that the performance of CNN models with optimization parameters is superior. Using the CFS dataset, the test data evaluation results obtained an accuracy of 98%, AUC 0,96, precision 99%, recall 92%, and f1-score 96%. The best tuning result for CNN is 3 combinations of filter and kernel {[64&7], [32&3], [16&2]} and pool_size=2. The limitation of this research is how to handle the two algorithms being compared. Both use different approaches, namely Supervised Learning and Deep Learning.

Penggunaan Teknik Transfer Learning pada Metode CNN untuk Pengenalan Tanaman Bunga

2024-08-09T11:24:53+07:00

This study explores the impact of using the transfer learning method to improve flower recognition performance using Convolutional Neural Network (CNN) models. The dataset used consists of 4242 flower images divided into five classes: daisy, tulip, rose, sunflower, and dandelion. This research implements three models: basic CNN, VGG16, and EfficientNetB3, to test the effectiveness of transfer learning in flower classification. The basic CNN model achieved a training accuracy of 73.38% and validation accuracy of 71.76%, but it is limited in generalizing to new data. The VGG16 model achieved perfect training accuracy but experienced overfitting, with validation accuracy stabilizing around 85-90%. Meanwhile, the EfficientNetB3 model with transfer learning reached a training accuracy of 98.50% and a validation accuracy of 94.00%, demonstrating strong generalization without significant overfitting. The experiment was conducted using data augmentation techniques, and performance evaluation was carried out using accuracy, precision, and recall metrics. The results show that transfer learning with the EfficientNetB3 model provides the best performance in flower classification compared to the basic CNN and VGG16 models. For future research, further development can be done by expanding the types of flower datasets and applying additional optimization techniques to improve accuracy in more complex models.

Perbandingan Sensitivitas Metode SAW, MAUT dan WSM pada Anugerah Mutu Non-Akademik Universitas

2025-01-13T08:10:21+07:00

Decision Support System works best with suitable method. Unfortunately, not all methods are equally used. Two rarely used methods are MAUT and WSM methods. To find out whether a method is more suitable for a case than another method is to do a sensitivity test. By doing sensitivity tests between the two methods and other methods such as the commonly used SAW in the same case, it’s possible to find out the comparison of sensitivity percentages between the three. One case that can be helped by a Decision Support System using the three methods is ANOMIK assessment at universities. The three methods produced the same best alternative, namely Faculty 9. After doing a sensitivity test, the results showed that WSM method was the most sensitive with a value of 4.954%, followed by SAW method with a value of 4.901% and finally MAUT method with 3.844%.

Algoritma Random Forest dan Synthetic Minority Oversampling Technique (SMOTE) untuk Deteksi Diabetes

2025-02-12T08:21:37+07:00

Diabetes is one of the challenges in global health. Indonesia ranks 5th in the world with the highest rate of diabetes. This research uses the Random Forest algorithm for diabetes detection. The purpose of the study is to detect diabetes with the Random Forest algorithm that provides accurate and efficient results in the early diagnosis of diabetic patients. The data used is secondary data "Diabetes Dataset" which consists of 952 data and has 17 features. The test scenario in this study divides the data consisting of 3 parts, namely scenario 1 90%:10% ratio, scenario 2 70%:30% ratio, scenario 3 50%:50% ratio. In each scenario, a comparison between using SMOTE and not using SMOTE is applied. The best performance results are obtained in scenario 1 which uses SMOTE, which produces 97% accuracy, 100% precision, 94% recall and the last is F1-Score which produces 97%.

Klasifikasi Penyakit pada Tanaman Berdasarkan Citra Daun Menggunakan Metode Convolutional Neural Network

2024-08-09T11:28:16+07:00

The agricultural sector is a vital part of the economy, providing food, raw materials, and employment opportunities. In Indonesia, this sector faces significant challenges, such as low interest from younger generations and plant disease issues. Plant disease identification typically requires experienced experts, but this process is time-consuming and costly. This research aims to develop a plant disease classification model using Convolutional Neural Network (CNN) to assist farmers in identifying diseases in rice, corn, tomato, and potato plants based on leaf images. Testing was conducted with data splitting ratios of 70:30, 80:20, and 90:10, using both single-stage and multi-stage classification methods. The best results were achieved with an 80:20 data ratio using single-stage classification, with an average accuracy of 80%, precision of 80%, recall of 81%, and F1-score of 79%. This study demonstrates that the CNN method is effective in plant disease classification, with optimal performance at an 80:20 data ratio and single-stage classification. It is hoped that this research can help farmers quickly and accurately identify and manage plant diseases, as well as encourage innovation in the agricultural sector. The implementation of CNN in plant disease classification shows great potential in enhancing the efficiency and accuracy of disease detection, ultimately supporting the sustainability and development of the agricultural sector.

Prediksi Kualitas Udara Menggunakan Metode CatBoost

2024-11-21T17:19:27+07:00

Air is important for life, but industrial activities, forest burning, cigarette smoke and transportation increase air pollution. AirVisual AQI 2024 data places Jakarta in 11th place in the world with the highest level of pollution, reaching 127 which is unhealthy for sensitive groups, and poses a risk of causing serious illnesses such as skin and respiratory diseases. This research uses the CatBoost method to predict the air quality index using Jakarta SPKU data taken from Kaggle. The data is processed through pre-processing and divided into four models with different comparisons of training and testing data. Each model was tested with the parameters iteration, depth, learning_rate, and l2_leaf_reg, using GridSearchCV to find the best combination. The results show that the model with 90% training data and 10% testing data provides the best accuracy of 97%, due to the larger proportion of training data. This research shows that the CatBoost method can provide accurate air quality predictions, which is important to support efforts to reduce the impact of pollution and improve public health.