Tata kelola data yang efektif menjadi fondasi utama dalam menjamin kualitas informasi, terutama di sektor kelautan dan perikanan yang memiliki distribusi aktor dan operasi secara geografis luas. Ketersediaan data yang akurat, dapat dilacak, dan relevan sangat penting untuk mendukung pengambilan keputusan berbasis bukti, khususnya dalam kegiatan penyuluhan yang membina lebih dari 48.000 kelompok di seluruh Indonesia. Penelitian ini bertujuan merancang Data Governance Framework penyuluhan di sektor kelautan dan perikanan guna meningkatkan kualitas data penyuluhan. Pendekatan yang digunakan adalah Design and Development Research (DDR), dengan landasan pada prinsip tata kelola data oleh Ladley serta model evaluasi kapabilitas organisasi Capability Maturity Model Integration (CMMI). Framework yang dikembangkan mencakup struktur peran formal (Data Owner, Data Steward, Data Custodian), kebijakan akses berbasis peran, validasi data, manajemen metadata, serta glosarium untuk menjamin konsistensi dan keterlacakan data. Selain itu, penelitian ini menghasilkan Minimum Sustainable Operating Model (MSOM) sebagai strategi implementasi bertahap yang realistis di tengah keterbatasan sumber daya organisasi. Hasil validasi menunjukkan bahwa framework ini tidak hanya konsisten dengan literatur dan prinsip tata kelola data publik, tetapi juga relevan dengan regulasi nasional seperti Undang-Undang Perlindungan Data Pribadi. Framework ini diharapkan dapat meningkatkan akurasi, efisiensi, dan akuntabilitas data penyuluhan, sekaligus memperkuat pengambilan keputusan strategis berbasis data di sektor publik. Abstract Effective data governance is a fundamental foundation for ensuring information quality, particularly in the marine and fisheries sector, which involves geographically dispersed actors and operations. The availability of accurate, traceable, and relevant data is essential to support evidence-based decision-making, especially in extension activities that serve over 48,000 community groups across Indonesia. This study aims to design a Data Governance Framework for extension services in the marine and fisheries sector to improve data quality. The research adopts a Design and Development Research (DDR) approach, grounded in Ladley’s data governance principles and the Capability Maturity Model Integration (CMMI) for evaluating organizational capabilities. The proposed framework includes formal role structures (Data Owner, Data Steward, Data Custodian), role-based access policies, data validation mechanisms, metadata management, and a glossary to ensure data consistency and traceability. Additionally, the study introduces a Minimum Sustainable Operating Model (MSOM) as a phased implementation strategy, suitable for organizations with limited resources. Validation results indicate that the framework aligns with existing literature and public data governance principles, while also complying with national regulations such as the Personal Data Protection Law. This framework is expected to enhance the accuracy, efficiency, and accountability of extension data, while strengthening strategic data-driven decision-making in the public sector.
The identification of oil palm fruit ripeness is an important factor in maintaining harvest quality and improving palm oil productivity. Manual identification processes still have limitations, including subjectivity and inconsistency in assessment results. This study aims to evaluate and compare the performance of a Baseline Convolutional Neural Network (CNN) model with several transfer learning architectures, namely EfficientNetB3, ResNet50, and DenseNet121, for oil palm fruit ripeness classification. The dataset consisted of 302 original images of oil palm fruits categorized into three classes: ripe, unripe, and rotten. To prevent data leakage, the dataset was first divided using a stratified split into training, validation, and testing sets before data augmentation was applied exclusively to the training set. The augmentation techniques included rotation, translation, zooming, brightness adjustment, and horizontal flipping to increase data variability and reduce overfitting. All models were trained using an input size of 224 × 224 pixels, the Adam optimizer, and categorical cross-entropy as the loss function. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics to assess the classification capability of each model. In addition, confusion matrix analysis was conducted to identify classification error patterns across the ripeness categories. The results indicate that transfer learning models outperformed the Baseline CNN model. DenseNet121 achieved the best overall performance, followed by EfficientNetB3 and ResNet50. These findings demonstrate that transfer learning is an effective approach for oil palm fruit ripeness classification, particularly when working with limited datasets. Nevertheless, further studies using larger and more diverse datasets are recommended to improve model generalization capabilities.
The rapid digital transformation in the public sector necessitates the development of service applications that are not only functionally robust but also highly usable. MyASN, as a digital personnel service platform for Indonesian civil servants (Aparatur Sipil Negara/ASN), is designed to facilitate access to employment-related information and administrative services. However, its implementation has revealed several usability issues, including login difficulties, slow system responsiveness, and feature instability, which negatively affect the overall user experience. This study aims to evaluate the usability level of the MyASN application using the System Usability Scale (SUS) method. Data were collected through the distribution of SUS questionnaires to application users. The results indicate an average SUS score of 54.08, which falls below the standard benchmark, placing the application in the Not Acceptable – Marginal category, with an adjective rating of Poor–OK and a grade of D. Furthermore, the Net Promoter Score (NPS) falls into the Detractor category, indicating low user satisfaction and a limited likelihood of recommendation. These findings suggest that the MyASN application exhibits significant shortcomings in terms of interface design, navigation, feature consistency, and system performance. Therefore, improvements are required, particularly in simplifying the user interface, enhancing system stability, and optimizing overall performance to improve user experience and the quality of digital personnel services.
Necrosis, or body tissue death, occurs when there is insufficient blood flow to the tissue, which can be caused by injury, radiation, or chemicals. One of the main challenges in the automated diagnosis of necrosis is data imbalance in medical datasets, where the number of pathological cases is far less than normal cases. To address this issue, this study implements and evaluates various data sampling techniques, including Random Undersampling (RUS), Random Oversampling (ROS), Combination of Over-Undersampling (COUS), Synthetic Minority Over-sampling Technique (SMOTE), and Tomek Link, then using a Support Vector Machine (SVM) as the classifier. The test results show that the best sampling technique is the Synthetic Minority Over-sampling Technique (SMOTE), which successfully achieved an accuracy of 100% and an Area Under Curve (AUC) of 100%, indicating its significant potential in improving the accuracy of necrosis diagnosis from CT scans.
Recovering deleted data from flash drives has become a crucial requirement in digital forensics and personal data recovery. This study compares file carving and metadata-based file recovery techniques across three applications: Recuva, PhotoRec, and FTK Imager. Fifty files of various types permanently deleted from an 8GB FAT32-formatted flash drive were tested sequentially across the three applications. Evaluation parameters included the number of recovered files, integrity through hash verification, percentage of openable files, ability to preserve original names, and scanning speed. Recuva's deep scan feature excelled in recovering metadata, including file names and folder hierarchies. PhotoRec achieved the highest recovery volume thanks to its aggressive header and footer detection, but was unable to preserve file names or directory structures. FTK Imager achieved the highest file integrity with the most stable recovery, despite requiring a longer scan time. This study confirms that no application excels in all aspects, so tool selection should be tailored to user priorities.
Vibe coding transforms natural language commands into functional code through Artificial Intelligence (AI). This Systematic Literature Review (SLR) evaluates the impacts, challenges, and potential of this approach within education. Adhering to the PRISMA protocol, 17 studies were selected based on inclusion criteria. The distribution encompasses higher education (n=9) and K-12 (n=8), focusing on tools like GitHub Copilot and ChatGPT. Synthesis results from controlled experimental studies among higher education students indicate a significant increase in task completion efficiency of up to 35%. However, a distinction is observed where increased technical productivity does not equate to improved learning achievement; instead, risks concerning diminished conceptual understanding and academic integrity challenges remain. This review is limited by the geographic dominance of studies from North America and Europe and the high heterogeneity of study designs, which necessitated a narrative synthesis. This study recommends a pedagogical shift towards critical human-AI collaboration.
Deteksi kecacatan perangkat lunak berperan penting dalam mengidentifikasi komponen yang berpotensi bermasalah sebelum kegagalan terjadi. Meskipun prediksi kecacatan dapat mengoptimalkan waktu pengembangan, menurunkan biaya, dan meningkatkan kepuasan pelanggan, efektivitasnya sangat bergantung pada pemilihan fitur yang relevan untuk model klasifikasi. Penelitian ini mengevaluasi penerapan Grey Wolf Optimizer (GWO) untuk seleksi fitur pada deteksi kecacatan perangkat lunak menggunakan Support Vector Machine (SVM), serta membandingkan kinerjanya dengan Genetic Algorithm (GA), Particle Swarm Optimization (PSO), dan Firefly Algorithm (FFA). Evaluasi dilakukan pada 12 dataset dari NASA Metrics Data Program (NASA MDP) dengan penerapan Synthetic Minority Oversampling Technique (SMOTE) untuk menangani ketidakseimbangan kelas. Hasil menunjukkan bahwa seleksi fitur berbasis GWO secara konsisten meningkatkan performa SVM dibandingkan penggunaan semua fitur, serta secara signifikan mengungguli metode berbasis GA, PSO, dan FFA pada seluruh dataset. Abstract Software defect detection plays a crucial role in identifying potentially faulty components before failures occur. While defect prediction can optimize development time, reduce costs, and enhance customer satisfaction, its effectiveness largely depends on selecting the most relevant features for classification models. This study evaluates the application of the Grey Wolf Optimizer (GWO) for feature selection in software defect detection using a Support Vector Machine (SVM) and compares its performance with Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Firefly Algorithm (FFA). The evaluation, conducted on 12 datasets from the NASA Metrics Data Program (NASA MDP), incorporates the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance. Results show that GWO-based feature selection consistently improves SVM performance over using all features and significantly outperforms GA-, PSO-, and FFA-based approaches across all datasets.
This study aims to analyze and compare the level of usability between two prominent digital travel booking platforms in Indonesia, namely Tiket.com and Traveloka. The background of this research lies in the criticality of superior user experience in determining competitive advantage within the fierce digital service landscape. The main objective is to quantitatively measure and analyze the difference in usability scores between the two platforms. The methodology utilizes a quantitative approach by implementing the standard System Usability Scale (SUS) questionnaire. The questionnaire consists of 10 alternating positive and negative items, presented using a 5-point Likert scale. These items are repeated for each application to obtain individual usability scores. The survey will be distributed to 30 respondents active users. The results are expected to specifically identify the relative strengths and weaknesses of both applications, providing strategic input for system developers to enhance user interface quality and achieve optimal customer satisfaction.
The East Kalimantan SAMSAT (Sistem Administrasi Manunggal Satu Atap) website serves as a crucial digital public service for motor vehicle tax payments and related services. However, suboptimal user experience (UX) and user interface (UI) often hinder accessibility and public satisfaction. This study primarily aims to analyze and identify UI/UX design constraints on the East Kalimantan SAMSAT platform, and to provide recommendations for improvements based on user perspectives. The applied method is Design Thinking, which includes five stages: Empathize, Define, Ideate, Prototype, and Test. The initial stage focused on data collection through observation, in-depth user interviews, and heuristic evaluation to uncover user needs and pain points. The analysis findings revealed issues such as complex navigation, inconsistencies in visual design, and lack of explicit information, all of which significantly increased user cognitive load. Based on these results, the study proposes specific design recommendations in the form of wireframe and mockup improvements. These recommendations focus on simplifying the user flow, increasing the consistency of UI elements, and developing a more intuitive search feature. This design proposal is expected to boost the effectiveness, efficiency, and user satisfaction of online SAMSAT services. These design recommendations were then validated during the testing phase through usability testing. Test results showed significant improvements in key metrics: the average task success rate increased to 90% and the task completion time decreased by 40%. These findings demonstrate that the systematic application of human-centered design is effective in bridging the gap between user needs and the quality of digital public services at the East Kalimantan SAMSAT.
The rapid growth of digital transactions in Indonesia has increased the demand for efficient, secure, and stable information systems. DANA, as one of the largest digital wallets, plays a crucial role in providing fast payment services. This study aims to analyze the implementation of information systems in the DANA application, focusing on system architecture and data processing that support transaction efficiency. A descriptive-qualitative case study method was employed through literature review, technical documentation analysis, and application feature observation. Results show that DANA's microservices architecture, real-time data processing, and distributed cloud infrastructure significantly accelerate transactions (average 5 seconds) and achieve 99% uptime. In conclusion, structured and adaptive information system implementation greatly contributes to the speed, security, and operational efficiency of DANA as a digital transaction platform.
The industrial world and information systems are undergoing a major shift in management practices. Growing and diverse user demands have driven information systems to evolve from passive data management into intelligent systems that assist organizational management in decision-making processes. This research aims to analyze the evolution of information systems, from conventional models to the implementation of Artificial Intelligence (AI) based systems within organizations. The findings indicate that the transformation of information systems is not merely a software update, but a paradigm shift: from passive systems reliant on manual input (Conventional Phase), to systems capable of integrated workflows (Automated Phase), and finally to systems that can learn and provide independent recommendations (Intelligent System Phase). The primary finding of this study is that the transition to intelligent systems significantly enhances operational efficiency. However, this must be balanced with high data quality as a reliable information source and supported by the readiness of human resources. In conclusion, information systems in the digital era have transformed from simple administrative tools into essential strategic partners that help organizations navigate data complexity and improve decision-making.
The Student Academic Information System (SIAM) is an information system for managing academic data at Mulia University Balikpapan (UM). Although this system has helped students access academic services, there are still issues related to ease of use between systems. Therefore, this study was conducted to assess the usability level of the SIAM interface using the System Usability Scale (SUS) method. This study used a quantitative approach with a SUS questionnaire distributed to active students of Mulia Balikpapan University who had experience using SIAM. The data obtained was analyzed to determine the usability category of the system and identify the obstacles perceived by users. Based on data calculations, an average score of 60.69 was obtained, reflecting the system's position in the Marginal High category. Technically, this score refers to a D grade scale with an OK adjective rating. Although the system is functionally capable of accommodating user needs, the data indicates that optimization of design elements is still necessary. These findings confirm that SIAM Universitas Mulia has reached a threshold of acceptable acceptance, but still requires some adjustments to achieve optimal usability performance. The results of this study are expected to provide useful input for the development and refinement of the SIAM interface so that it can improve user comfort, efficiency, and satisfaction in supporting academic activities.
Facial emotion recognition is an important research area in computer vision and artificial intelligence, with applications in human–computer interaction, affective computing, and intelligent systems. This study aims to evaluate the performance of a Convolutional Neural Network (CNN) for facial emotion recognition using the FER2013 dataset. The FER2013 dataset consists of grayscale facial images with a resolution of 48×48 pixels and includes seven emotion classes: angry, disgust, fear, happy, neutral, sad, and surprise. Due to its low image resolution and imbalanced class distribution, FER2013 presents significant challenges for emotion classification tasks. An experimental research approach was employed by implementing a baseline CNN architecture composed of convolutional, pooling, and fully connected layers. Image normalization and batch-based data generation were applied during preprocessing. The model was trained using the Adam optimizer with categorical cross-entropy loss, and an early stopping mechanism was utilized to prevent overfitting. Model performance was evaluated using accuracy, precision, recall, F1-score, and confusion matrix analysis. The experimental results show that the proposed CNN model achieved an overall test accuracy of 55.50%. Emotions with distinctive facial features, such as happy and surprise, obtained higher F1-scores, while minority and visually subtle classes, particularly disgust and fear, exhibited lower performance. These findings indicate that a simple CNN architecture can provide reasonable performance on challenging facial emotion datasets while highlighting the impact of class imbalance and limited image resolution. The proposed model can serve as a baseline for further improvements in facial emotion recognition systems.
Technological advancements have encouraged Micro, Small, and Medium Enterprises (MSMEs) to adopt digital marketing strategies to enhance competitiveness. This study aims to analyze the role of marketing digitalization, particularly visual-based digital campaigns and influencer promotion, in improving the competitiveness of local MSMEs. The research employed a quantitative descriptive approach using an online questionnaire distributed to consumers who interacted with MSME digital promotions. The collected data were analyzed descriptively. The results indicate that digital campaigns significantly influence consumer purchase decisions, with 90% of respondents rating digital campaigns as effective to very effective. Furthermore, 100% of respondents stated that digital marketing had a positive impact on MSME development. These findings confirm that marketing digitalization plays a strategic role in strengthening MSME competitiveness in the digital era.
Employee performance evaluation is a crucial aspect in supporting objective and measurable managerial decision-making. However, in practice, performance assessment is often still subjective and not fully supported by systematic analytical methods. This study aims to analyze and compare the combined Analytical Hierarchy Process and Technique for Order Preference by Similarity to Ideal Solution (AHP–TOPSIS) method with the Multi-Objective Optimization on the Basis of Ratio Analysis (MOORA) method in a Decision Support System for employee performance evaluation at PT Classics Prima Sejahtera. This research applies a quantitative approach using a case study method, where data are collected through observation, interviews, and performance assessment questionnaires. AHP is employed to determine the weights of evaluation criteria based on their level of importance, while TOPSIS and MOORA are used to rank employee performance. The results indicate that both methods are capable of producing objective employee rankings; however, differences in ranking order occur due to the distinct computational characteristics of each method. The AHP–TOPSIS method demonstrates advantages in terms of accuracy in criteria weighting, whereas MOORA offers a simpler and more efficient calculation process. This study concludes that both methods can be effectively applied as decision support tools, with the selection of methods depending on the needs and complexity of decision-making within the organization.
Stunting is a public health problem that impacts child growth and development, particularly during the First 1,000 Days of Life (1000 HPK). This study aims to apply the K-Means Clustering algorithm to classify stunting risk levels based on 1000 HPK data in Nagari Aia Gadang Timur. The data used include maternal and child health indicators, such as maternal nutritional status, prenatal check-up history, birth weight, exclusive breastfeeding, and child growth measurements. The K-Means algorithm was used to group the data into several clusters. The results showed that the formed clusters were able to clearly distinguish low, medium, and high stunting risk groups. The application of K-Means Clustering can facilitate early identification of stunting risk and support data-driven decision-making in planning stunting prevention and intervention programs at the nagari level.
Penelitian ini bertujuan untuk menerapkan algoritma enkripsi Sparkle Schwaemm128-128, yang merupakan algoritma enkripsi ringan yang berdasarkan pada permutasi Sparkle, untuk proses enkripsi pengiriman data pada sensor GPS berbasis IoT dengan protokol modul NRF24L01. Metode penelitian yang digunakan adalah eksperimen dengan menggunakan perangkat keras NodeMCU ESP8266, sensor GPS NEO-6Mv2, modul komunikasi NRF24L01, dan platform Arduino IDE. Hasil penelitian menunjukkan bahwa algoritma Sparkle Schwaemm128-128 dapat mengenkripsi dan mendekripsi data sensor GPS dengan kecepatan 8012 mikrodetik untuk waktu enkripsi dan 13108 untuk waktu dekripsi. Sisa memori penggunaan pada mikrokontroler NodeMCU ESP8266 berkisar 51765 byte yaitu hanya 36,75% memori yang terpakai pada node sensor dan 52174 byte yaitu hanya 36,36% memeri yang terpakai pada node pusat. Penelitian ini diharapkan dapat memberikan solusi keamanan yang efisien dan efektif untuk sistem IoT yang menggunakan sensor GPS dan modul komunikasi NRF24L016.
Evaluasi dan monitoring pada pasien penderita tuberkulosis paru membutuhkan pengobatan yang tepat. Dalam beberapa tahun terakhir, pada tahap monitoring telah ditemukan penanda biologis yang disebut suPAR (soluble urokinase plasminogen activator receptor) berpotensi sebagai biomarker untuk mendiagnosis, prognosis, dan evaluasi penyakit paru. Penelitian ini bertujuan untuk menyelidiki faktor-faktor yang berhubungan dengan pasien tuberkulosis paru dan menentukan faktor yang paling signifikan berdasarkan waktu pengamatan, indeks massa tubuh, dan laju endapan darah. Sampel sebanyak 60 pasien tuberkulosis paru di Malang dievaluasi secara longitudinal setiap dua minggu sekali selama 13 periode. Dalam penelitian ini, kami menggunakan analisis jalur dengan generalized linear mixed model (GLMM) dengan metode Weighted Least Square (WLS) untuk menyelidiki hubungan antarvariabel terhadap kadar monosit dan kadar suPAR pada pasien tuberkulosis paru. dan membandingkan hasilnya dengan Ordinary Least Square (OLS). Hasil penelitian menunjukkan bahwa model terbaik adalah GLMM dengan struktur kovarian unstructured dengan AIC terkecil dan R2 terbesar. Selain itu, indeks massa tubuh memiliki pengaruh yang paling signifikan terhadap kadar monosit dan suPAR pada pasien TB paru. Oleh karena itu, sangat penting untuk mempertimbangkan indeks massa tubuh pasien dalam evaluasi dan monitoring pasien tuberkulosis paru. Abstract Evaluating and monitoring patients with pulmonary tuberculosis needs an appropriate treatment. In recent years, a biological marker called suPAR (soluble urokinase plasminogen activator receptor) has been identified in the monitoring stage and has the potential as a biomarker for diagnosing, prognosing, and monitoring disease. This study aims to investigate the factors associated with patients with pulmonary tuberculosis and determine the most significant factor based on observation time, BMI, and ESR. A total of 60 patients diagnosed with pulmonary tuberculosis in Malang were included and evaluated longitudinally every two weeks over 13 periods. In this study, we use path analysis with the generalized linear mixed model (GLMM) using the Weighted Least Square (WLS) method to investigate the relationship between the variables on monocyte and suPAR levels in pulmonary tuberculosis patients and compare the results those obtained using the Ordinary Least Square (OLS) method. The results demonstrate that the optimal model is the GLMM with an unstructured covariance structure, exhibiting the smallest AIC and the largest R2. Additionally, body mass index exerts the most significant effect on monocyte and suPAR levels in patients with pulmonary tuberculosis. Consequently, considering patient’s BMI when evaluating and monitoring patients with pulmonary tuberculosis is imperative.
Catur diperkirakan memiliki sekitar 1043 kemungkinan posisi. Angka tersebut jauh melampaui kemampuan komputasi komputer yang ada saat ini, sehingga mengembangkan sebuah mesin catur dengan mempertimbangkan seluruh kemungkinan posisi dianggap tidak memungkinkan. Saat ini, penggunaan Neural Network pada pengembangan mesin catur sedang mengalami peningkatan dan telah membawa hasil yang menjanjikan sejak pertama kali diperkenalkan oleh AlphaZero milik Google DeepMind pada tahun 2017. Penelitian ini bertujuan untuk membawa potensi pendekatan baru pada ranah pengembangan mesin catur berbasis Neural Network dengan memperkenalkan mesin catur Deeplefish yang melakukan gerakan berdasarkan keluaran model Long Short Term Memory (LSTM). Menggunakan lebih dari 57.000 pertandingan yang terbagi menjadi 1.200.000 posisi, model dilatih untuk memprediksi langkah berikutnya oleh putih untuk sebuah rangkaian gerakan yang diberikan. Model meraih loss sebesar 3,01 dan Average Centipawn Loss (ACPL) sebesar 219 pada data uji. Deeplefish meraih 2 kemenangan, 72 kekalahan, dan 10 hasil seri pada tahap pengujian. Hasil yang tidak memuaskan ini dapat disebabkan oleh subjektivitas data terhadap cara berpikir pemain, menghasilkan kurangnya pola gerakan yang signifikan untuk dipelajari oleh model. Abstract Chess has been estimated to have around 1043 possible positions. This number surpasses the computing ability of any computer available, therefore, building a chess engine that considers every possible position is deemed impractical. Currently, the use of neural network in chess engine development is on the rise and has been delivering promising results since the introduction of Google DeepMind’s AlphaZero in 2017. This research aims to bring a new potential approach to the field of neural network based chess engine development by introducing Deeplefish chess engine that uses a Long Short Term Memory (LSTM) model as move generator. Trained on more than 57.000 games broken down into more than 1.200.000 positions, the model is trained to predict the next move played by white for a given sequence of moves. The model achieved a loss of 3.01 and an Average Centipawn Loss (ACPL) of 219 on the validation set. Deeplefish achieved 2 wins, 72 losses, and 10 draws on the testing, showing a lack of board and contextual awareness. This unsatisfactory results are likely due to the subjectivity of the data to the player’s way of thinking, resulting in lack of significant move pattern to be learned by the model.
Permasalahan dalam sektor pertanian biofarmaka di Indonesia saat ini mencakup beberapa aspek penting yang menghambat perkembangan dan produktivitasnya. Salah satu masalah utama adalah kurangnya infrastruktur dan akses terhadap teknologi pertanian modern. Banyak petani masih menggunakan metode tradisional karena keterbatasan teknologi. Untuk mengatasi hal ini, dikembangkanlah FocketFarm, sebuah platform edukasi berbasis web untuk pertanian greenhouse. Tujuannya adalah memberikan pengetahuan dan keterampilan praktis dalam pertanian greenhouse kepada pengguna. Pengguna FocketFarm dapat memperoleh informasi tentang teknik bertani, perawatan tanaman, pengelolaan sumber daya, dan praktik-praktik pertanian yang berkelanjutan. Untuk mengembangkan website FocketFarm, yaitu dengan menggunakan metode Framework for the Application of System Thinking (FAST) tujuan nya adalah untuk menyusun desain sistem secara lebih efisien dan sesuai dengan tujuan yang ditetapkan. Webste FocketFarm dapat memperoleh informasi tentang teknik bertani di dalam greenhouse, perawatan tanaman, pengelolaan sumber daya, dan praktik-praktik pertanian yang berkelanjutan. Dengan antarmuka yang mudah digunakan, FocketFarm memungkinkan akses yang intuitif bagi pengguna dari latar belakang yang beragam. Dengan demikian, FocketFarm berfungsi sebagai alat efektif dalam meningkatkan pemahaman dan keterampilan dalam pertanian greenhouse melalui platform web yang mudah diakses Abstract Chess has been estimated to have around 1043 possible positions. This number surpasses the computing ability of any computer available, therefore, building a chess engine that considers every possible position is deemed impractical. Currently, the use of neural network in chess engine development is on the rise and has been delivering promising results since the introduction of Google DeepMind’s AlphaZero in 2017. This research aims to bring a new potential approach to the field of neural network based chess engine development by introducing Deeplefish chess engine that uses a Long Short Term Memory (LSTM) model as move generator. Trained on more than 57.000 games broken down into more than 1.200.000 positions, the model is trained to predict the next move played by white for a given sequence of moves. The model achieved a loss of 3.01 and an Average Centipawn Loss (ACPL) of 219 on the validation set. Deeplefish achieved 2 wins, 72 losses, and 10 draws on the testing, showing a lack of board and contextual awareness. This unsatisfactory results are likely due to the subjectivity of the data to the player’s way of thinking, resulting in lack of significant move pattern to be learned by the model.
