Jurnal Politeknik Negeri Batam (PoliBatam)
Not a member yet
3001 research outputs found
Sort by
Sentiment Analysis Using LSTM Algorithm Regarding Grab Application Services in Indonesia
Get PDFThis study aims to analyze the sentiment of user reviews for the Grab Indonesia application using Long Short-Term Memory (LSTM) algorithms. Two variants of LSTM, namely Stacked LSTM and Bi-Directional LSTM, were compared to determine the most effective model in classifying user review sentiments. Both models were enhanced with Multi-Head Attention mechanisms to capture more complex contextual relationships in sequential data. The data used consists of 2,000 user reviews collected through scraping from the Google Play Store, with sentiment labels of positive and negative. Data preprocessing included labeling, case folding, stopword removal, tokenization, stemming, and the application of the SMOTE technique to address class imbalance. The results show that the Bi-Directional LSTM model achieved the highest validation accuracy of 87%, with an F1-score of 0.90 for the negative class and 0.82 for the positive class, while the Stacked LSTM recorded an accuracy of 84%, with an F1-score of 0.87 for the negative class and 0.78 for the positive class. Overall, the Bi-Directional LSTM demonstrated better performance in identifying both negative and positive sentiments, providing a good balance between precision and recall. This study proves that Bi-Directional LSTM with Multi-Head Attention can improve sentiment analysis performance on user reviews of digital applications, with potential applications in various other platforms.This study aims to analyze the sentiment of user reviews for the Grab Indonesia application using Long Short-Term Memory (LSTM) algorithms. Two variants of LSTM, namely Stacked LSTM and Bi-Directional LSTM, were compared to determine the most effective model in classifying user review sentiments. Both models were enhanced with Multi-Head Attention mechanisms to capture more complex contextual relationships in sequential data. The data used consists of 2,000 user reviews collected through scraping from the Google Play Store, with sentiment labels of positive and negative. Data preprocessing included labeling, case folding, stopword removal, tokenization, stemming, and the application of the SMOTE technique to address class imbalance. The results show that the Bi-Directional LSTM model achieved the highest validation accuracy of 87%, with an F1-score of 0.90 for the negative class and 0.82 for the positive class, while the Stacked LSTM recorded an accuracy of 84%, with an F1-score of 0.87 for the negative class and 0.78 for the positive class. Overall, the Bi-Directional LSTM demonstrated better performance in identifying both negative and positive sentiments, providing a good balance between precision and recall. This study proves that Bi-Directional LSTM with Multi-Head Attention can improve sentiment analysis performance on user reviews of digital applications, with potential applications in various other platforms
Prediction of Corrosion Inhibitor Efficiency Based on Quinoxaline Compounds Using Polynomial Regression
Get PDFCorrosion is a natural process that leads to material degradation due to environmental factors. It significantly impacts financial and safety aspects, including structural weakening and economic losses in various industries such as oil, gas, and nuclear. Corrosion inhibitors, especially organic compounds like quinoxaline, are widely used to reduce corrosion by forming protective layers on metal surfaces. Quinoxaline compounds, characterized by their heterocyclic structure with nitrogen atoms, demonstrate promising inhibition efficiency in corrosive environments. In this study, machine learning (ML) approaches are utilized to predict the corrosion inhibition efficiency of quinoxaline compounds. Algorithms such as Gradient Boosting Regressor (GBR), Extreme Gradient Boosting Regressor (XGBR), and Automatic Relevance Determination (ARD) regression are compared. The implementation of polynomial functions significantly improves the prediction accuracy of these models. Among them, GBR achieved the best value with MSE, RMSE, MAE, MAPE, and R2 values of 0.0000001, 0.0003229, 0.0000029, 0.0002294, and 0.999999998, respectively. These findings highlight the potential of polynomial-enhanced ML models in accurately predicting corrosion inhibition efficiency. Moreover, the study demonstrates the viability of GBR as a reliable tool for analyzing and optimizing corrosion inhibitors for industrial applications.Corrosion is a natural process that leads to material degradation due to environmental factors. It significantly impacts financial and safety aspects, including structural weakening and economic losses in various industries such as oil, gas, and nuclear. Corrosion inhibitors, especially organic compounds like quinoxaline, are widely used to reduce corrosion by forming protective layers on metal surfaces. Quinoxaline compounds, characterized by their heterocyclic structure with nitrogen atoms, demonstrate promising inhibition efficiency in corrosive environments. In this study, machine learning (ML) approaches are utilized to predict the corrosion inhibition efficiency of quinoxaline compounds. Algorithms such as Gradient Boosting Regressor (GBR), Extreme Gradient Boosting Regressor (XGBR), and Automatic Relevance Determination (ARD) regression are compared. The implementation of polynomial functions significantly improves the prediction accuracy of these models. Among them, GBR achieved the best value with MSE, RMSE, MAE, MAPE, and R2 values of 0.0000001, 0.0003229, 0.0000029, 0.0002294, and 0.999999998, respectively. These findings highlight the potential of polynomial-enhanced ML models in accurately predicting corrosion inhibition efficiency. Moreover, the study demonstrates the viability of GBR as a reliable tool for analyzing and optimizing corrosion inhibitors for industrial applications
Development of an IoT-Based Mobile Plastic Shredder for Optimized Waste Management in Batam
Get PDFPlastic waste management has become a critical environmental issue, with its improper handling leading to severe ecological and health impacts. This research addresses the challenge by designing and developing an IoT-based mobile plastic shredding machine aimed at improving waste management efficiency, particularly in Batam City, Indonesia. Utilizing Borg and Gall’s R&D framework, this study integrates IoT technology to enhance the machine’s functionality, enabling real-time data collection and remote monitoring through mobile applications. The machine comprises three functional levels: a storage area for raw plastic bottles, a shredding unit with proximity sensors, and a post-shredding storage compartment. Key innovations include weight sensors for automatic material handling and real-time data transmission via the Blynk IoT platform, controlled by an Arduino microcontroller. The modular design ensures portability, easy maintenance, and adaptability for use in various locations, including coastal areas. Prototyping involved integrating proximity sensors, load cells, relays, and motor control systems to ensure smooth operation. The machine demonstrated consistent performance during testing, with its IoT features enabling remote control and monitoring via smartphones. This facilitates optimized waste collection and contributes to reducing environmental pollution caused by plastic waste. The IoT-based mobile plastic shredding machine not only enhances waste management efficiency but also supports sustainability goals. Its portability and environmentally friendly design make it a practical solution for managing plastic waste in underserved areas. This innovation provides a significant step toward addressing the global plastic waste crisis, aligning with technological advancements to promote sustainable waste management practices
Development Strategy of MSMEs Convection Based on Islamic Economic Perspective : Case Study of Kembar Seragam Sekolah Convection Business
Get PDFThis study aims to analyze the development strategy of Micro, Small, and Medium Enterprises (MSMEs) in convection based on the perspective of Islamic economics. This perspective emphasizes public welfare as a strategic foundation for creating sustainable growth. This research method uses a qualitative method with a case study approach. Data were obtained through in-depth interviews with MSME actors and related parties, direct observation, and document analysis. Data analysis was carried out using the SWOT (Strengths, Weaknesses, Opportunities, Threats) framework to evaluate internal and external factors that affect MSMEs in convection. The results of the study indicate that although MSMEs face various challenges, such as limited capital and marketing, there are great opportunities to grow through the use of digital technology and innovative marketing strategies. Based on the SWOT analysis, the recommended strategies include strengthening product competitiveness, improving managerial skills, and expanding the market through digital platforms. By integrating the principles of Islamic economics, this strategy is not only oriented towards economic growth but also towards the welfare of the wider community. Based on the results of the IFAS and EFAS analysis, the position of the convection MSMEs is in Quadrant I (Strengths-Opportunities) in the SWOT matrix. This quadrant shows that the convection MSMEs has internal strengths that can be utilized to take external opportunities. This position provides direction for implementing the SO Strategy (Maxi-Maxi Strategy). The results of the analysis show that convection MSMEs have great potential to develop, especially through digital technology-based strategies and collaboration with local communities
Fraud Detection in Government in the Last Ten Years
Get PDFOver the past three years, Indonesian government state losses have exhibited a noticeable upward trend. The fraud triangle theory offers valuable insights into the relationship between fraud (as manifested in state losses) and three key factors: pressure, opportunity, and rationalization. These elements are reflected in instances of revenue shortfalls, potential losses, and non-compliance with regulations. Our study analyzed data from various government entities, including the central government, local governments, state-owned enterprises (SOEs), regional-owned enterprises (ROEs), public service agencies, regional public service institutions, and other government-related agencies over the past decade (2013 to 2023). Our findings reveal that pressure stemming from revenue shortfalls, opportunities associated with potential losses, and rationalization arising from non-compliance with regulations significantly contribute to fraudulent activities within the government sector. Based on our research, the fraud triangle theory, with its focus on revenue shortfalls, potential losses, and non-compliance with regulations, provides a robust framework for identifying fraudulent practices within the government sector
Financial Distress, Transfer Pricing, and Inventory Intensity: Their Effects on Tax Avoidance in Mining Companies
Get PDFThis research aims to examine the effect of financial distress, transfer pricing, and inventory intensity on tax avoidance in mining companies listed on the Indonesia Stock Exchange for the 2020–2023 period. The research sample consists of 84 companies selected using the purposive sampling method with secondary data obtained from the IDX and the official websites of the respective companies. Data analysis was conducted using multiple linear regression with SPSS version 27. The results show that financial distress and transfer pricing have no effect on tax avoidance due to strict government oversight and the high risk of being audited, while inventory intensity affects tax avoidance because inventory related costs can reduce taxable income
Analisis Pengaruh Temperatur Ambient Terhadap Kinerja Cooling Tower Unit 2 Berdasarkan Evaluasi Range dan Approach pada PLTGU PT Mitra Energi Batam
No full textPLTGU PT. Mitra Energi Batam has observed that Cooling Tower Unit 2 tends to produce hot steam in larger quantities compared to Unit 1. The greater production of hot steam in Cooling Tower Unit 2 certainly cannot be separated from the influence of ambient temperature (surrounding air temperature). Where the temperature around the Cooling Tower Unit 2 environment affects the performance of the Cooling Tower. The cooler the environment, the better the cooling tower performance. This research aims to determine the extent of influence of ambient temperature on the performance of Cooling Tower Unit 2 in terms of Range and Approach calculations and to evaluate the difference between "Range" and "Approach" calculations in measuring the performance of Cooling Tower Unit 2. Data collection was carried out for 8 hours in 1 days out of 10 days of data collection in November 2023. The Cooling Tower performance calculation method used is Range and Approach Evaluation, where Range is a calculation of incoming water temperature with outgoing water temperature and Approach is a calculation of outgoing water temperature with Wet Bulb Temperature (ambient air temperature Cooling Tower). Based on the Range Value Evaluation, the calculation results were found to be in the range of 9.52°C – 9.97°C and 8,391°C – 10,331°C. If the Range Value is smaller then the Cooling Tower performance can be said to be good, for the Approach Value the calculation results are found to be 16,515°C – 18,205°C and 17,625°C – 21,892°C. If the Approach value is smaller, then the Cooling Tower performance can be said to be good because the outlet water temperature can approach the value of the wet bulb temperature. For the Efficiency Value, the calculation results obtained were at values of 34.98°C – 37.26°C and 28.44% - 35.68%.PLTGU PT. Mitra Energi Batam telah diamati bahwa Cooling Tower Unit 2 cenderung menghasilkan uap panas dalam jumlah yang lebih besar dibandingkan dengan Unit 1. Produksi uap panas yang lebih besar pada Cooling Tower Unit 2 tentunya tidak lepas dari pengaruh temperatur ambient (suhu udara sekitar). Dimana suhu yang ada pada sekitar lingkungan Cooling Tower Unit 2 mempengaruhi kinerja Cooling Tower. Semakin dingin lingkungan semakin baik kinerja Cooling Tower. Penelitian ini bertujuan untuk mengetahui sejauh mana pengaruh temperatur ambient terhadap kinerja Cooling Tower Unit 2 dalam hal perhitungan Range dan Approach serta mengevaluasi perbedaan antara perhitungan "Range" dan "Approach" dalam mengukur kinerja Cooling Tower Unit 2. Pengambilan data dilakukan selama 8 jam dalam 1 hari dari 10 hari pengambilan data pada bulan November 2023. Metode perhitungan kinerja Cooling Tower yang digunakan berupa Evaluasi Range dan Approach, dimana Range merupakan perhitungan temperatur air masuk dengan temperatur air keluar dan Approach merupakan perhitungan Temperatur air keluar dengan Temperatur Wet Bulb (suhu udara sekitar Cooling Tower). Berdasarkan Evaluasi Nilai Range didapatkan hasil perhitungan berada pada kisaran 9.52°C – 9.97°C dan 8.391°C – 10.331°C. Jika Nilai Range semakin kecil maka kinerja Cooling Tower dapat dikatakan baik, untuk Nilai Approach didapatkan hasil perhitungan berada pada nilai 16.515°C – 18.205°C dan 17.625°C – 21.892°C. Jika Nilai Approach semakin kecil, maka kinerja Cooling Tower dapat dikatakan baik karena temperatur air keluar dapat mendekati nilai dari temperatur wet bulb. Untuk Nilai Efisiensi didapatkan hasil perhitungan berada pada nilai sebesar 34.98°C – 37.26°C dan 28.44% - 35.68%
ESP32-Based Smart Storage System with Fingerprint and HMI Integration for Workspaces
No full textManagement of laboratory equipment in higher education institutions often faces challenges such as lack of supervision, difficulties in inventory tracking, and an unstructured system. To address these issues, this research designs a Laboratory Component Storage Management System that integrates fingerprint technology, proximity sensors, HMI, and a database. The proximity sensor shows a success rate of 75% in detecting objects up to a distance of 10 cm. The stepper motor system on the X, Y, and Z axes has an average error of less than 1%, supporting movement precision. The VB.NET-based interface has proven successful through black box testing, with an Admin borrowing time of 47–50 seconds and a User borrowing time of 55–70 seconds, as well as a return time of 42–45 seconds and 52–62 seconds, respectively. This system enhances security, efficiency, and user experience in the practical process and can be an alternative solution to the weakness of the manual inventory system in the laboratory
Sentiment Analysis on Public Perception of the Nusantara Capital on Social Media X Using Support Vector Machine (SVM) and K-Nearest Neighbor (K-NN) Methods
No full textThe relocation and development of the National Capital City (IKN) as the center of government activities has become a hot topic, sparking diverse opinions among the public. The proposal to move the capital from DKI Jakarta to East Kalimantan has drawn significant attention from online communities, particularly on social media platform X (Twitter). This study aims to explore public sentiment regarding the development of IKN by applying artificial intelligence-based classification algorithms, namely Support Vector Machine (SVM) and K-Nearest Neighbor (K-NN). Sentiments are categorized as positive or negative to provide deeper insights into public perceptions. Through web crawling techniques, a total of 4,000 data points were collected. After the preprocessing stage, 3,608 data points remained, which were then translated into English to facilitate labeling using the Vader Sentiment method. The analysis results indicate that negative sentiment (1,873) is more dominant than positive sentiment (1,735). The data was then split into two sets: 80% for training (2,886 data points) and 20% for testing (722 data points). Based on the evaluation results, SVM and K-NN proved to be effective for sentiment analysis. SVM achieved an accuracy of 76%, precision of 78%, recall of 81%, and an f1-score of 79%, while K-NN attained an accuracy of 65%, precision of 62%, recall of 98%, and an f1-score of 76%. With superior performance, SVM emerges as a more reliable method for classifying public sentiment regarding the IKN development policy.The relocation and development of the National Capital City (IKN) as the center of government activities has become a hot topic, sparking diverse opinions among the public. The proposal to move the capital from DKI Jakarta to East Kalimantan has drawn significant attention from online communities, particularly on social media platform X (Twitter). This study aims to explore public sentiment regarding the development of IKN by applying artificial intelligence-based classification algorithms, namely Support Vector Machine (SVM) and K-Nearest Neighbor (K-NN). Sentiments are categorized as positive or negative to provide deeper insights into public perceptions. Through web crawling techniques, a total of 4,000 data points were collected. After the preprocessing stage, 3,608 data points remained, which were then translated into English to facilitate labeling using the Vader Sentiment method. The analysis results indicate that negative sentiment (1,873) is more dominant than positive sentiment (1,735). The data was then split into two sets: 80% for training (2,886 data points) and 20% for testing (722 data points). Based on the evaluation results, SVM and K-NN proved to be effective for sentiment analysis. SVM achieved an accuracy of 76%, precision of 78%, recall of 81%, and an f1-score of 79%, while K-NN attained an accuracy of 65%, precision of 62%, recall of 98%, and an f1-score of 76%. With superior performance, SVM emerges as a more reliable method for classifying public sentiment regarding the IKN development policy
Comparison of Data Normalization Techniques on KNN Classification Performance for Pima Indians Diabetes Dataset
No full textThis study analyzes the comparison of data normalization techniques in the K-Nearest Neighbors (KNN) model for diabetes classification using the Pima Indians Diabetes dataset. The three normalization techniques evaluated are Min-Max Scaling, Z-Score Scaling, and Decimal Scaling. After preprocessing, such as handling missing values and removing duplicates, as well as feature selection using the Random Forest method, the features removed include SkinThickness, Insulin, Pregnancies, and BloodPressure. The evaluation was carried out using accuracy, precision, recall, F1-Score, specificity, and ROC AUC metrics. The results show that Min-Max Scaling provides a significant improvement in all metrics, with the highest accuracy of 0.8117 and ROC AUC of 0.8050. Z-Score Scaling provides good results, but not as good as Min-Max Scaling. Decimal Scaling shows the lowest performance. Statistical tests using Paired T-Test show significant differences between Min-Max Scaling and without normalization on all metrics (P-Value <0.05), while Z-Score Scaling and Decimal Scaling are only significant on some metrics, with P-Values of 0.08363 and 0.43839 respectively for accuracy and ROC AUC. Overall, Min-Max Scaling proved to be the best normalization method for improving KNN performance in diabetes classification.This study analyzes the comparison of data normalization techniques in the K-Nearest Neighbors (KNN) model for diabetes classification using the Pima Indians Diabetes dataset. The three normalization techniques evaluated are Min-Max Scaling, Z-Score Scaling, and Decimal Scaling. After preprocessing, such as handling missing values and removing duplicates, as well as feature selection using the Random Forest method, the features removed include SkinThickness, Insulin, Pregnancies, and BloodPressure. The evaluation was carried out using accuracy, precision, recall, F1-Score, specificity, and ROC AUC metrics. The results show that Min-Max Scaling provides a significant improvement in all metrics, with the highest accuracy of 0.8117 and ROC AUC of 0.8050. Z-Score Scaling provides good results, but not as good as Min-Max Scaling. Decimal Scaling shows the lowest performance. Statistical tests using Paired T-Test show significant differences between Min-Max Scaling and without normalization on all metrics (P-Value <0.05), while Z-Score Scaling and Decimal Scaling are only significant on some metrics, with P-Values of 0.08363 and 0.43839 respectively for accuracy and ROC AUC. Overall, Min-Max Scaling proved to be the best normalization method for improving KNN performance in diabetes classification