Ardriani, Ni Nengah Dita
Unknown Affiliation

Published : 6 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 6 Documents
Search

Multivariat Predict Sales Data Using the Recurrent Neural Network (RNN) Method Ardriani, Ni Nengah Dita; Yastawil, Jamiin Al Yastawil; Erawati, Kadek Nonik; Yudi Antara, I Gede Made; Santiago, Gede Agus
IJCCS (Indonesian Journal of Computing and Cybernetics Systems) Vol 18, No 1 (2024): January
Publisher : IndoCEISS in colaboration with Universitas Gadjah Mada, Indonesia.

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.22146/ijccs.90165

Abstract

Sales is an activity or business selling a product or service. In this study, I took a case study on Kaggle. Sales problems at the company cause inventory to be very high or vice versa, causing a loss of sales because there are no items to sell. Inventory that is too high results in increased costs due to existing resources being inefficient. In the opposite condition, it will cause a product vacancy in the market. Using the Recurrent Neural Network (RNN) Algorithm, this study predicts sales. The data used is sales data in 2020 with the parameter Number of sales per day in the last four months. The results obtained through testing several training scenarios and testing the implementation of the algorithm, in this case, is the highest accuracy value of 96.92% in the network architecture of three input neuron layers, three hidden layer neurons, one output, division of training, and test data 70: 30, learning value rate of 0.9 and a maximum of 9000000 epochs
Object Detection Based on You Look Only Once Version 8 for Real-Time Applications Santiago, Gede Agus; Sugiartawan, Putu; Ardriani, Ni Nengah Dita
IJCCS (Indonesian Journal of Computing and Cybernetics Systems) Vol 18, No 3 (2024): July
Publisher : IndoCEISS in colaboration with Universitas Gadjah Mada, Indonesia.

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.22146/ijccs.94843

Abstract

This research focus to involves human detection in crowded situations, especially in the lecturer's room. The lecturer's room is very vulnerable because it can be accessed by anyone with only one entry and exit to the lecturer's room, so it would be perfect to place this Yolo camera in front of the lecturer's room so that incoming and outgoing activities can be monitored during work days on campus. The main challenge is how the system can distinguish individuals in dense crowds and identify their relative locations to each other. In this context, it is necessary to find a solution that can overcome the uncertainty of recognizing individuals in a group and accurately understand the location and distance between them. One proposed solution is to use the YOLO algorithm on video recordings to detect human objects in the lecturer's room during working hours. This research introduces the YOLOv8 model, a real-time detection system with high speed and accuracy in detecting and classifying objects in video recordings. YOLOv8 can accurately detect object movement, making it an efficient real-time framework for dealing with complex objects. This research experiment involved using eight different smartphone devices to collect datasets. Using various smartphone devices aims to test object detection performance under various shooting conditions, including variations in image quality, lighting, shooting angle, and camera resolution. The research results show that using multiple smartphone devices in dataset collection can improve the robustness and accuracy of object detection models. By integrating datasets from various sources and shooting conditions, the YOLOv8 model was successfully trained to better recognize objects in different situations, even in campus environments that often have challenges such as weather variations and lighting fluctuations. The test results show an accuracy rate of 93.33% in human object detection
Adaptive Operator and Scaling Factor Selection in Differential Evolution using Parametrized Reinforcement Learning Santiyuda, Kadek Gemilang; Sugiartawan, Putu; Santiago, Gede Agus; Ardriani, Ni Nengah Dita; Kafiyanna, Moch Ilham Nur
Jurnal Sistem Informasi dan Komputer Terapan Indonesia (JSIKTI) Vol 7 No 3 (2025): March
Publisher : INFOTEKS (Information Technology, Computer and Sciences)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33173/jsikti.206

Abstract

Mutation strategy selection along with parameter settings are well known challenges in enhancing the performance of differential evolution (DE). In this paper, we propose to solve these problems as a parametrized action Markov decision process. A multi-pass deep Q-network (MP-DQN) is used as the reinforcement learning method in the parametrized action space. The architecture of MP-DQN comprises an actor network and a Q-network, both trained offline. The networks’ weights are trained based on the samples of states, actions and rewards collected on every DE iterations. We use 99 features to describe a state of DE and experiment on 4 reward definitions. A benchmark study is carried out with functions from CEC2005 to compare the performance of the proposed method to baseline DE methods without any parameter control, with random scaling factor, and to other DEs with adaptive operator selection methods, as well as to the two winners of CEC2005. The results show that DE with MP-DQN parameter control performs better than the baseline DE methods and obtains competitive results compared to the other methods.
Deep Learning Approach for USD to IDR Forecasting with LSTM Ardriani, Ni Nengah Dita; Sugiartawan, Putu; Santiago, Gede Agus; Darma Wandika, I Made Pranadata; Wiwahana Prasetya, I Made Irfan
Jurnal Sistem Informasi dan Komputer Terapan Indonesia (JSIKTI) Vol 8 No 1 (2025): September
Publisher : INFOTEKS (Information Technology, Computer and Sciences)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33173/jsikti.210

Abstract

This Research explores the use of Long Short-Term Memory (LSTM) networks for forecasting the USD to IDR exchange rate, with the goal of improving prediction accuracy in the volatile foreign exchange market. By leveraging historical data, including daily exchange rates and trading volume, the LSTM model captures long-term dependencies and patterns within the time series data. The results show that the LSTM model effectively predicts general trends and medium-term fluctuations, demonstrating its capacity to follow market dynamics. However, the model struggles with extreme volatility and sudden market shifts, particularly during unforeseen geopolitical or economic events. This limitation highlights the need for further enhancement through the incorporation of additional features, such as macroeconomic indicators, sentiment analysis, and real-time news data. Furthermore, the study suggests the potential benefits of combining LSTM with other machine learning techniques to create hybrid models that can better handle short-term fluctuations and extreme events. In conclusion, while LSTM shows promise for exchange rate forecasting, its performance can be improved by refining model parameters, incorporating diverse data sources, and exploring hybrid approaches. This research provides valuable insights for traders, investors, and policymakers seeking to make more informed decisions in the foreign exchange market.
LSTM Neural Network for Predicting Tourist Arrivals to Bali Erawati, Kadek Nonik; Sugiartawan, Putu; Ardriani, Ni Nengah Dita; Hartama, I Dewa Agung Bayu Mega; Frasetya, I Gusti Ngurah Hendra; Mahendra, I Gede Orka
Jurnal Sistem Informasi dan Komputer Terapan Indonesia (JSIKTI) Vol 8 No 1 (2025): September
Publisher : INFOTEKS (Information Technology, Computer and Sciences)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33173/jsikti.211

Abstract

Tourism is a key pillar of Bali’s economy, contributing significantly to employment, cultural preservation, and income generation. Accurate forecasting of tourist arrivals is crucial for sustainable growth and resource optimization. This study applies Long Short-Term Memory (LSTM) neural networks to predict tourist arrivals in Bali, leveraging historical data and external factors such as global economic indicators, flight frequencies, cultural events, and environmental conditions. LSTM’s ability to model complex temporal dependencies and non-linear relationships offers significant advantages over traditional methods like ARIMA, especially in handling seasonal patterns and irregularities. The model was trained on a robust dataset, preprocessed to address missing values, outliers, and variability. Performance evaluation metrics, including RMSE, demonstrate high predictive accuracy during stable periods but highlight limitations in handling anomalies such as the COVID-19 pandemic. To address these challenges, recommendations include integrating additional external variables, employing hybrid models, and conducting scenario-based sensitivity analyses to enhance adaptability and robustness. The results highlight the practical utility of AI-driven forecasting tools in tourism management, providing actionable insights for policymakers and stakeholders to optimize planning, mitigate risks, and support sustainable development. This research contributes to the growing field of AI applications in tourism, promoting resilience and competitiveness in an increasingly dynamic global market.
Optimizing Chili Price Prediction Using Machine Learning Classification Antara, I Gede Made Yudi; Sugiartawan, Putu; Ardriani, Ni Nengah Dita; Dewa, Hari Putra Maha; Widya Dharma, I Gusti Ngurah Adi; Satya, I Putu Adnya
Jurnal Sistem Informasi dan Komputer Terapan Indonesia (JSIKTI) Vol 8 No 1 (2025): September
Publisher : INFOTEKS (Information Technology, Computer and Sciences)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33173/jsikti.214

Abstract

Optimizing chili price prediction is critical for agricultural stakeholders, enabling better decision-making in supply chain management, market strategies, and farming practices. This research focuses on leveraging machine learning classification models to improve the accuracy and reliability of chili price predictions. The research addresses the challenges of class imbalance, which often occurs due to the uneven representation of price fluctuations in datasets. Resampling techniques, including oversampling the minority class with Synthetic Minority Oversampling Technique (SMOTE) and undersampling the majority class, were employed to balance the dataset and enhance the model's sensitivity to less frequent price drops. Key predictive features such as weather conditions, market demand, transportation costs, and economic indicators were integrated into the models. Advanced classification algorithms like Random Forests and Gradient Boosted Trees were utilized, demonstrating their effectiveness in handling non-linear relationships and class imbalance. Regularization techniques and k-fold cross-validation were applied to prevent overfitting and ensure robust model performance across different data subsets.The results show significant improvements in precision, recall, and overall model accuracy, making the approach suitable for real-world applications. By optimizing machine learning models, this research provides actionable insights for stakeholders to manage price volatility effectively, supporting sustainable agricultural practices and market stability.