Kinetik: Game Technology, Information System, Computer Network, Computing, Electronics, and Control
Not a member yet
424 research outputs found
Sort by
The Comparison of Imbalanced Data Handling Method in Software Defect Prediction
Get PDFSoftware testing is a crucial process in software development life cycle which will affect the software quality. However, testing is a tedious task and resource consuming. Software testing can be conducted more efficiently by focusing this activitiy to software modules which is prone to defect. Therefore, an automated software defect prediction is needed. This research implemented Extreme Learning Machine (ELM) as classification algorithm because of its simplicity in training process and good generalization performance. Aside classification algorithm, the most important problem need to be addressed is imbalanced data between samples of positive class (prone to defect) and negative class. Such imbalance problem could bias the performance of classifier. Therefore, this research compared some approaches to handle imbalance problem between SMOTE (resampling method) and weighted-ELM (algorithm-level method).The results of experiment using 10-fold cross validation on NASA MDP dataset show that including imbalance problem handling in building software defect prediction model is able to increase the specificity and g-mean of model. When the value of imbalance ratio is not very small, the SMOTE is better than weighted-ELM. Otherwise, weighted-ELM is better than SMOTE in term of sensitivity and g-mean, but worse in term of specificity and accuracy.Software testing is a crucial process in software development life cycle which will affect the software quality. However, testing is a tedious task and resource consuming. Software testing can be conducted more efficiently by focusing this activitiy to software modules which is prone to defect. Therefore, an automated software defect prediction is needed. This research implemented Extreme Learning Machine (ELM) as classification algorithm because of its simplicity in training process and good generalization performance. Aside classification algorithm, the most important problem need to be addressed is imbalanced data between samples of positive class (prone to defect) and negative class. Such imbalance problem could bias the performance of classifier. Therefore, this research compared some approaches to handle imbalance problem between SMOTE (resampling method) and weighted-ELM (algorithm-level method).The results of experiment using 10-fold cross validation on NASA MDP dataset show that including imbalance problem handling in building software defect prediction model is able to increase the specificity and g-mean of model. When the value of imbalance ratio is not very small, the SMOTE is better than weighted-ELM. Otherwise, weighted-ELM is better than SMOTE in term of sensitivity and g-mean, but worse in term of specificity and accuracy
Implementation of K-Means Clustering and Weighted Products in Determining Crime-Prone Locations
Get PDFClustering algorithms can be used to build geographic mapping systems to determine crime-prone locations. This study aims to establish a geographical mapping system to determine crime-prone locations that can help police control certain locations that often occur crime and provide information to people in crime-prone locations. Criminal groups are calculated based on crime data from November 2018 to October 2019 which occurred in 9 districts in Kudus Regency. The crime grouping process uses the k-means method used to classify based on regional vulnerability and uses a weighted product method that functions as a vulnerability ranking that is vulnerable to crime selection. The grouping results obtained from this study are that there are 1 very vulnerable area, 5 areas in the vulnerable category, and 3 safe areas. While the weighted product method produces Melatilor area as a vulnerable area to be defeated by a score of 0.182093. This research provides benefits for the public to see crime-prone areas so that they can be more vigilant, while for the police to analyze crime so as to speed up the process of resolving crime and increase and improve crime prevention measures
Cyber Security Analysis of Academic Services based on Domain Delivery Services and Support using Indonesian E-Government Ratings (PEGI)
Get PDFSafe academic services are the most important part of universities. The security of academic services is very important to maintain information optimally and safely. Along with the development of technology, academic information services are often misused by some irresponsible parties that can cause threats. To prevent these things from happening, it is necessary to know the extent of governance of higher education academic information system security by evaluating. So the research was conducted to determine the maturity of the security of Higher Education academic information service security by using the COBIT 5 framework in the DSS05 domain. The DSS05 domain in COBIT 5 is a good framework for use in implementing and evaluating the security of academic information services. Meanwhile, to determine the achievement of the evaluation of the security level of academic information systems, the Indonesian e-government ranking (PEGI) method is required. The combination of the COBIT 5 framework in the DSS05 domain using the PEGI method in academic information security service is able to provide a level of achievement in the form of Customer Value. The results of the COBIT 5 framework analysis of the DSS05 domain using the PEGI method get a score of 3.50 so that the quality of academic information service security evaluation achievement is at a very good level. At this level, universities are increasingly open to technological development. Higher education has applied the concept of quantification in every process, and has always been monitored and controlled for its performance in the security of academic information systems
Application of Early Diagnosis of Diabetes Mellitus (DM) Equipped with Calorie Needs for DM Sufferers using the Fuzzy Mamdani Method
Get PDFDiabetes Mellitus (DM) is one of the deadliest degenerative diseases in the world. The prevalence of DM in Indonesia from year to year shows asignificant increase. The high number of these causes the need for appropriate action and anticipation for health workers, DM families and DM people themselves. In this study, a system application model was created by using informatics techniques in health for early diagnosis of DM and what calorie needs needed for DM sufferers. This system was created using a GUI application and the Mamdani fuzzy method. The purpose of creating this system is to help in making an initial decision for DM diagnosis. The results obtained, first a DM diagnosis system with 6 input variables, 3 output variables, and 155 rules with MAPE achieved 29.48%. The second is the calorie requirements system with 2 input variables, 2 output variables namely BMI with MAPE 10.57% BMR with MAPE 9.7% and 9 rules with the results achieved by 99%
Towards an Effective Tuberculosis Surveillance in Indonesia through Google Trends
Get PDFBackground. The search digital footprint, such as in Google Trend (GT), forms a large dataset that is suitable to be used as surveillance data and supports early warning systems. These advantages become great opportunities for disease surveillance agencies in Indonesia to get rapid early disease monitoring. Objective. Due to limited research in this area and the increasing level of internet penetration in Indonesia, a further study is needed in disease monitoring by utilizing Google Trends. In this research, we explore, analyze and create a set of the best search terms to be used in utilizing GT for disease surveillance in Indonesia, especially Tuberculosis. Method. We use correlation as the technique to define the relatedness between the real case data and GT results. We collect data from the Ministry of Health of Indonesia. From the data, we design a set of new search terms to take GT trend data. The collected data is analyzed using the Pearson correlation. Result. The analysis shows that the studied search terms give strong positive relationships between GT trend data and Tuberculosis cases number in Indonesia. From the correlation analysis, we get a set of proposed effective search terms with the highest score equals to 0.907. Conclusion. Finally, it is possible to monitor and make quick surveillance in tuberculosis in Indonesia through Google Trend and we have created a novel set of search terms that can be used as the basis in monitoring other diseases in Indonesi
Patient Queue Systems in Hospital Using Patient Treatment Time Prediction Algorithm
Get PDFPatient Treatment Time Prediction Algorithm was very important to build an outpatient queue system at the hospital. This study aims to build a system of outpatient queues to predict the waiting time of outpatients in the eye clinic at one of Cirebon hospitals. Patient Treatment Time Prediction algorithm was calculated based on historical data or medical records of patients in the hospital with 120 patient data. The Patient Treatment Time Prediction algorithm was trained by improved Random Forest algorithm for each service and a waiting time for each service. Prediction of waiting time for each patient service was obtained by calculating the consumption of patient care time based on patient characteristics. The waiting time for each service predicted by the trained Patient Treatment Time Prediction algorithm is the total waiting time of patients in the queue for each service. This research resulted in a system that can show the time taken by patients in every service available in the eye clinic. Patient time consumption in each service produced varies according to the patient's condition, in this case based on the patient's gender and age. This research provides benefits in terms of improving performance in each department involved, optimizing human resources, and increasing patient satisfaction. This research can be developed for each department in the hospital
An application of ANFIS for Lung Diseases Early Detection System
Get PDFIndonesian Basic Health Research in 2018 showed the prevalence of pneumonia, pulmonary tuberculosis (TB) and lung cancer in Indonesia 4.0% 0.4% and 0.18%, respectively. However, the number of lung specialists is small. According to the Indonesian Lung Specialist Association webpage, the number of doctors joined in the association up to 2008 were 452. This amount is very less when compared with existing lung disease cases. Thus, the handling of lung disease will be too late. The use of ANFIS for early detection of lung disease is growing. However, the systems designed are need preprocessing data to be executed and still applied for one type of disease. This paper will design a desktop application based on ANFIS expert system to detect lung disease early, i.e. for pneumonia, pulmonary TB and lung cancer. The system will work based on simple symptoms expressed by the patient. Subtractive clustering is used for clustering process. The results of the training showed that the models were able to give better performance compared to the model built using conventional clustering methods. The test results show that those three models have comparable performance compared to their counterpart. Software validation shows that the it gives 94.00% succeed for training data and up to 100% for testing data. This application is not intended to replace the role of a doctor, but to help diagnose the patient's condition earlier
Classification of Lexile Level Reading Load Using the K-Means Clustering and Random Forest Method
Get PDFThere are various ways to improve the quality of someone's education, one of them is reading. By reading, insight and knowledge of various kinds of things can increase. But, the ability and someone's understanding of reading is different. This can be a problem for readers if the reading material exceeds his comprehension ability. Therefore, it is necessary to determine the load of reading material using Lexile Levels. Lexile Levels are a value that gives a size the complexity of reading material and someone's reading ability. Thus, the reading material will be classified based a value on the Lexile Levels. Lexile Levels will cluster the reading material into 2 clusters which is easy, and difficult. The clustering process will use the k-means method. After the clustering process, reading material will be classified using the reading load Random Forest method. The k-means method was chosen because of the method has a simple computing process and fast also. Random Forest algorithm is a method that can build decision tree and it’s able to build several decision trees then choose the best tree. The results of this experiment indicate that the experiment scenario uses 2 cluster and SMOTE and GIFS preprocessing are carried out shows good results with an accuracy of 76.03%, precision of 81.85% and recall of 76.05%
Comparative Study of Decision Tree, K-Nearest Neighbor, and Modified K-Nearest Neighbor on Jatropha Curcas Plant Disease Identification
Get PDFJatropha Curcas is a very useful plant that can be used as a bio fuel for diesel engines replacing the coal. In Indonesia, there are few plantation that plant Jatropha Curcas. But there is so limited farmers that understand in detail about the disease of Jatropha Curcas and it may cause a big loss during harvesting when the disease occured with no further action. An expert system can help the farmers to identify the lant diseases of Jatropha Curcas. The objective of this research is to compare several identification and classification methods, such as Decision Tree, K-Nearest Neighbor and its modification. The comparison is based on the accuracy. Modified K-Nearest Neighbor method given the best accuracy result that is 67.74%
Optimization Fuzzy Inference System based Particle Swarm Optimization for Onset Prediction of the Rainy Season
Get PDFRainfall which is occurred in an area explain the Onset Rainy Season (ORS). ORS is a characteristic of the rainy season which is important to know, but the characteristics of the rain itself is very difficult to predict. We use the method of Fuzzy Inference System (FIS) to predict ORS. Unfortunately, FIS is weak to determine parameters so that influences the working FIS method. In this study, we use PSO to optimize parameter of the FIS method to increase perform of the FIS method for onset prediction of the rainy season with the predictor Sea Surface Temperature Nino 3.4 and Index Ocean Dipole. We used coefficient correlation to determine the relationship between two variables as predictors and RMSE as evaluate to all methods. The experiment result has shown that the work of FIS-PSO after optimizing produced the good work with the coefficient correlation = 0.57 and RMSE = 2.96 that is the smallest value that is better performance if compared with other methods. It can be concluded that the method proposed can increase the onset prediction of the rainy season