Jurnal Online Informatika
Not a member yet
276 research outputs found
Sort by
Development of a Digital Platform Prototype, to Facilitate Inclusive Learning for Children with Special Needs
Get PDFPersons with disabilities have the same rights and responsibilities as citizens. Based on the 1945 Constitution Republic of Indonesia, article 31 paragraph 1 and Law Number 20 of 2003 concerning the National Education System, it can be concluded that the state provides full guarantees for Children with Special Needs to obtain quality education services. Many of the problems of inclusive learning that occurred during the Covid-19 pandemic, ranging from the unpreparedness of the school to various problems with environmental factors so that innovation was needed to overcome these problems. In this article, the author develops a prototype of a digital-based learning platform as a solution to facilitate inclusive learning for children with special needs
PSO based Hyperparameter tuning of CNN Multivariate Time- Series Analysis
Get PDFConvolutional Neural Network (CNN) is an effective Deep Learning (DL) algorithm that solves various image identification problems. The use of CNN for time-series data analysis is emerging. CNN learns filters, representations of repeated patterns in the series, and uses them to forecast future values. The network performance may depend on hyperparameter settings. This study optimizes the CNN architecture based on hyperparameter tuning using Particle Swarm Optimization (PSO), PSO-CNN. The proposed method was evaluated using multivariate time-series data of electronic journal visitor datasets. The CNN equation in image and time-series problems is the input given to the model for processing numbers. The proposed method generated the lowest RMSE (1.386) with 178 neurons in the fully connected and 2 hidden layers. The experimental results show that the PSO-CNN generates an architecture with better performance than ordinary CNN
Sentiment Analysis for the Brazilian Anesthesiologist Using Multi-Layer Perceptron Classifier and Random Forest Methods
Get PDFSexual harassment is defined as giving sexual attention both verbally, either in speech or writing, and physically to victims who are predominantly women, On July 13, 2022, there was a tweet featuring a video of sexual harassment that made it trend in various countries. The video irritated Twitter users and made various comments resulting in various sentiments that can be analyzed using sentiment analysis. The purpose of this study is to see what the public thinks about the sexual harassment case of Brazilian anesthesiologist. Besides the sentiment analysis, another aim of this study is to see how objective are those sentiments based on their polarity. This study uses a comparison of two methods in sentiment analysis, namely Multi-Layer Perceptron Classifier and Random Forest, and labeling automatically using TextBlob. This results in 94.44% accuracy, 94.44% precision, 92% recall and 93% f1_score. For MLP Classifier and accuracy 96.42%, precision 94.44%, recall 96.66% and f1_score 95.56% for Random Forest. Sentiment polarity score from the TextBlob is -0.5 and subjectivity is 0.4 which indicates that most statements are negative and subjective score is 0.4, which means those sentiments are subjective in nature
Anti-Corruption Disclosure Prediction Using Deep Learning
Get PDFCorruption gives major problem to many countries. It gives negative impact to a nation economy. People also realized that corruption comes from two sides, demand from the authority and supply from corporate. On that regard, corporates may have their part in fight against corruption in the form of anti- corruption disclosure (ACD). This study proposes new method of ACD prediction in corporate using deep learning. The data in this study are taken from every companies listed in Indonesia Stock Exchange (IDX) from the year 2017 to 2019. The companies can be categorized in 9 categories and the data set has 8 features. The overall data has 1826 items in which 1032 items are ACD and the other 794 items are non-ACD. In this study, the deep neural network or deep learning is composed from input layer, output layer and 3 hidden layers. The deep neural network uses Adam optimizer with learning rate 0.0010, batch size 16 and epochs 500. The drop out is set to 0.05. The accuracy result from deep learning in predicting ACD is considered good with the average training accuracy is 74.76% and average testing accuracy is 76.37%. However, the loss result isn’t good with average training loss and testing loss are respectively 51.76% and 50.96%. Since the aim of the study to find the possibility of deep learning as alternative of logistic regression in ACD prediction, accuracy comparison from deep learning and logistic regression is held. Deep learning has average prediction accuracy of 76.37% is better than logistic regression with average accuracy of 67.15%. Deep learning also has higher minimum accuracy and maximum accuracy compared to logistic regression. This study concludes that deep learning may give alternatives in ACD prediction compared the more common method of logistic regression
Performance Analysis of Cache Replacement Algorithm using Virtual Named Data Network Nodes
Get PDFAs a future internet candidate, named Data Network (NDN) provides more efficient communication than TCP/IP network. Unlike TCP/IP, consumer requests in NDN are sent based on content, not the address. The previous study evaluated the NDN performance using a simulator. In this research, we modeled the system using virtual NDN nodes, making the model more relevant to the real NDN. As an essential component in every NDN router, the content store (CS) has a function to keep the data. We use First In First Out (FIFO) and Least Recetly Used (LRU) in our nodes as cache replacement algorithms. The in-depth exploration is done using various scenarios. The result shows that the cache hit ratio (CHR) increases if the size of the CS, the number of interests, and the number of consumers increases. CHR decreases as the number of producers and the number of prefixes increase. As CHR increases, round trip time (RTT) decreases. LRU provides better performance for all cases: higher CHR of 5-15% and lower RTT of 1-10% than FIFO
Data Analysis of Social Media\u27s Impact on COVID19 Pandemic Users\u27 Mental Health
Get PDFSocial media has a significant impact on people\u27s daily lives and spread widely. Unrestrained usage of social media could have worsening consequences on mental health. The majority of COVID-19 users who were exposed to social media learned numerous facts, which made their anxiety and depression-related mental health disorders worse. This study aims to determine how social media usage affects users\u27 mental health during the COVID19 pandemic. Through surveys and expert interviews, this study collects both quantitative and qualitative data. The total number of respondents involved was 106 with the average age group of 18-41-year-old. Using reliability testing (Cronbach alpha test) and inferential statistic (Pearson Correlation and Chi-Square), results show that during the COVID19 pandemic, there is a significant link between social media use and mental health. Anxiety and depression brought on by social media are more common among young adults, predominantly female, between the ages of 18 and 24 than in men. Additionally, correlation plot analysis with a variety of queries reveals the mental health issues and activities on social media
Internet of Things (IoT) for Soil Moisture Detection Using Time Series Model
Get PDFTechnology in agriculture has been widely and massively applied. One of them is automation technology and the use of big data through the Internet of Things (IoT). The use of IoT allows a process to run automatically without human intervention. Extreme weather changes and narrow land use are one of the main problems in agriculture. The development of IoT devices has been widely developed regarding this subject. One of them is a soil moisture detection system. This study aims to build an IoT soil moisture detection system. The system will use a sensor as input which is then processed in a microcontroller device and the prediction results are sent to the IoT cloud platform. Prediction results are obtained using a time series model and then its performance is evaluated using RMSE. This model was chosen because the structure of the observed soil moisture data is based on time. The results of this study indicate that the soil moisture IoT system can work well. This is supported by the results of the prediction evaluation value of the RMSE = 1.175682x10-5 model which is very small
Diabetes Risk Prediction Using Extreme Gradient Boosting (XGBoost)
Get PDFOne of the uses of medical data from diabetes patients is to produce models that can be used by medical personnel to predict and identify diabetes in patients. Various techniques are used to be able to provide a diabetes model as early as possible based on the symptoms experienced by diabetic patients, including using machine learning. The machine learning technique used to predict diabetes in this study is extreme gradient boosting (XGBoost). XGBoost is an advanced implementation of gradient boosting along with multiple regularization factors to accurately predict target variables by combining simpler and weaker model set estimations. Errors made by the previous model are tried to be corrected by the next model by adding some weight to the model. The diabetes prediction model using XGBoost is shown in the form of a tree, with the accuracy of the model produced in this study of 98.71
Random Forest Method Approach to Customer Classification Based on Non-Performing Loan in Micro Business
Get PDFThis study aims to classify potential customers’ characteristics based on non- performing loans through the random forest method. This research uses data obtained from Syariah Mandiri Bank branch in Jambi, which includes data on micro-financing customers in years 2016–2020. The random forest method is used for analysis. The novelty of this work is that, unlike existing researches that used other soft-computing methods, we employ Random Forest method, specifically using an imbalanced class sampling technique. The obtained results show that credit risk can be estimated by taking into account factors such as age, monthly installments, margin, price of insurance, loan principal, occupation, and long installments. The research results indicate that the sensitivity, precision, and G-mean value increase compared to using the original data. Random forest with oversampling technique has the high Area Under the ROC Curve score that is equal to 66.69%
Multi Rule-based and Corpus-based for Sundanese Stemmer
Get PDFThe purpose of this study is to develop a stemming method by involved several methods including morphological (with affix and pro-lexeme removal), syllable (canonical) pattern, and corpus data as a comparison of the final results of stemming. The algorithm checks a number of the string first and removes affixes, then check the syllable pattern according to the stripping result, then compares to the corpus data which determines the final stemming process. In this study, the corpus data was taken from Sundanese dictionary consists of a single word used for the root word and the extracted dataset from the online Sundanese magazine. The results showed that the stripping of affix and pro-lexeme can remove the corresponding affixes and pro-lexeme then compares words that have a syllable pattern then executes the basic words quickly and the use of corpus can improve accuracy and reduce the over-stemming problems that occur in the stemming process