Supeno Mardi Susiki Nugroho

Features extraction to improve performance of clustering process on student achievement

In clustering data, there are two popular methods which are usually used: k-Means and Fuzzy C Means (FCM). Clustering process by these two methods, however, are sometimes influenced by the data suitable being used. This may affect the performance, for example: execution time, accuracy level. In order to overcome this problem, especially in a student evaluation system, we propose a feature extraction stage, which is implemented in the data preprocessing before being used by FCM. This extraction itself is performed based on the category and the Blooms Taxonomy by collecting student data in a serious game. The experimental results show that these proposed methods are able to increase the accuracy level and to reduce the execution time. In terms of accuracy, our method is, on average, 2.3–4.7% higher than that of the original FCM. In terms of the execution time, the proposed FCM is, on average, 2.2–2.7 second faster than the original.

Simulation multi behavior NPCs in fire evacuation using emotional behavior tree

Evacuation procedure in building fire has several points, and the main destination for occupants are going through the exit door of the building safely. In this research we learn how occupants behave with the emotion influence in a building fire. Emotions have important role to affect decision making because it can increase or decrease the rational value and aim at being realistic and naturally. The possibility to make NPCs behave naturally are implement multi behavior and handle all behavior using artificial intelligence (AI) technique. AI architecture have to support variant behavior and easily to reuse for complex character behaviors. Behavior trees (BTs) is the one of many AI techniques that more readable and scalable for action selection mechanism. So, we propose to implement Emotional Behavior Trees (EmoBTs) to handle dynamic behavior scenario that influenced by emotions. We already made the scenario for NPCs, that have emotion to decide which act would be selected. The experiment result compare how each NPCs are going to act by the emotions influence.

Urban distribution CCTV for smart city using decision tree methods

Supervision is important in the application of Smart City. Closed-Circuit Television (CCTV) is one of the main tools of Smart City surveillance. Some CCTVs connected with Information and Communication Technology (ICT) form Smart Monitoring. In order to support Smart Mobility, CCTV is installed to monitor road conditions. The problem is the installation of CCTV which is not always appropriate with the location conditions. Consequently situations such as road density, accidents, crime can not be monitored optimally. This research aims to find the relation of type CCTV camera with location placement. Classification technology is used to construct predictive models. Using the Decision Tree algorithm obtained an accuracy of the prediction model is 87.96%.

Predicting daily consumer price index using support vector regression method based cloud computing

Severe inflation can cause a countrys economic downturn. Therefore, inflation needs to be controlled. One of inflation control conducted by the government is predicting and calculating inflation using CPI indicators on a monthly. Prediction with monthly frequency, could be too late, because inflation has been a few days and it is not known quickly. With the development of internet technology today, various data sources related to inflation easily obtained in real-time. This data can be used for daily CPI prediction. Daily predictions allow policy makers to make better policies. CPI prediction using daily data will face challenges. The growing variants and data volumes need good computing systems. Cloud computing can be used to solve the problem. This is a preliminary research in developing daily CPI prediction model using big data and cloud computing. Here we focus on developing a daily CPI prediction model using the Support Vector Regression (SVR) method in a cloud computing. For better accuracy, we compared the kernel functions of SVR and tuning SVR parameters using the grid search and Random Search method. In addition, we compared SVR with the Random Forest method. These daily CPI predictions are simulated into cloud computing environments. From this simulation we show computation time and accuration comparisons needed if run on personal computers with cloud computing. The results showed that SVR using RBF kernel has less mse value 0.3454 in monthly prediction and 0.0095 in daily predictions. And Random Forest result is slightly different than SVR — RBF, mse value 0.0171 in daily prediction. Experiment show that running CPI prediction have less time, for 1644 data need takes 522s than PC takes 837s.

Live migration based on cloud computing to increase load balancing

Nowadays, the availability of Internet services is very important. One of these internet services is information system service. When the information system can not be accessed, the user cannot access the information. Therefore, the server used for that application should remain online nonstop. However, in reality the services failure often happens due to server error or repairing process. One of the solution is cloud computing so that the running application can be moved to another server known as migration. Due to regular migration process takes time to move the data, so the services cannot be access. It is necessary to migrate directly to the destination server without moving the data. One way to implement that live migration is replicating the existing data on the origin server to the destination server at any time. Direct migration can be done by replicating data from the origin server to the destination server every time. In this paper we analyze performance of the live migration on cloud computing to improve the availability of information systems. The goal is for users to access information services continuously.

Mt. Kelud haze removal using color attenuation prior

Kelud crater observation using closed-circuit television (CCTV) has not been used as the main guide in the world of volcanology. This is caused by observations manually by volcanologist who is not certain and depends on their ability and experience. In practice, there is still obstacles haze in the image taken from CCTV record. This paper present color attenuation prior method to eliminate haze on the digital image. The results obtained showed that the selected method is capable of eliminating sparse haze and moderate haze but not dense haze.

Outer contour extraction of skull from CT scan images

Extraction of the outer contour of the skull is an important step in craniofacial reconstruction. The outer contour is required for surface reconstruction of the skull. In this paper, we propose a method to extract the outer contour of the skull. The extraction process consists of four stages: defining the region of interest, segmentation of the bone, noise removal and extraction of the outer contour based on scanning from the four sides of the image. The proposed method successfully extracts the outermost contour of the skull and avoids redundant data.

Crowd navigation using leader-follower algorithm based Reciprocal Velocity Obstacles

This research employs Reciprocal Velocity Obstacles (RVO) to create oscillation- and collision-free motion with multiagent navigation. By combining RVO with the steering behaviour i.e. leader following behaviour thus creating leader-follower relationship between agents. This relationship leads to the navigation to their shared destination. Each agent i.e. follower navigates to follow their leader to reach their destination together. Experimental result demonstrates the navigation of hundreds of agents in densely populated environments which showing 18% faster compared with relation-less navigation to achieve the crowd goal.