Articles
<![CDATA[Analisis Sentimen Ulasan Video Animasi Menggunakan Metode Latent Semantic Indexing ]]>
<![CDATA[Faraz Dhia Alkadri]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Sutrisno Sutrisno]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 1 (2019): Januari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (415.999 KB)
<![CDATA[Animation videos are growing significantly producing tens even hundreds of titles per year. Certainly not everything were produced was interesting. Some of these videos may not be appealing to some people. To find out whether the animated videos is interesting or not, users can read the reviews given by other user about animation videos. Some sites that are intended to facilitate its users to be able give each other feed back about the animation video they have watched. From those reviews can be seen sentiment whether the review is a review that classified in to positive class sentiment or negative class sentiment. The Latent Semantic Indexing (LSI) method that adopts the Singular Value Decomposition (SVD) matrix reduction process is used to find the relevance between documents. With the LSI method helps us to be able to know the reviews are classified on positive sentiment or negative sentiment. The TF IDF method is used to process textual data into numerical data and cosine similiarity method is used to calculate the similiarity between data which is further classified into positive class sentiment and negative class sentiment. Testing done as much as 19 times by using different k-rank input. Based on the test result, this system produces an optimal accuracy on k-rank =10 that is equal to 86% so we can conclude that latent semantic indexing is good to use for searching relevance between documents.]]>
<![CDATA[Klon Perilaku Menggunakan Jaringan Saraf Tiruan Konvolusional Dalam Game SuperTuxKart ]]>
<![CDATA[Arrizal Amin]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Sigit Adinugroho]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 1 (2019): Januari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (597.803 KB)
<![CDATA[One of the important component of the video game is an artificial intelligence to make the game more competitive. Artificial intelligence used to decide action to reach goal in the game and challenge game player. In the process of developing artificial intelligence, developer needs to program an aritficial intelligence to make a decision for action for each states possible in the game. In this research, artifical neural network will be used as an artificial intelligence inside video game. Neural network will simplify process of developing artificial intelligence because developer does not have to program an algorithm to decide each action for each possible states in the game. Furthermore, neural network can learn or clone gamer's behavior while playing the game. In this research, SuperTuxKart will be used for an example to develop artificial intelligence inside video game. Artificial Intelligence with learning rate 0.0001, momentum 0.3 and epoch 100 reaches accuracy 86.72% for cloning game's behavior while playing video game. So this research concluded that neural network can be used as an artificial intelligence inside game.]]>
<![CDATA[Pengenalan Wajah dengan Pose Unik menggunakan Metode Learning Vector Quantization ]]>
<![CDATA[Achmad Dinda Basofi Sudirman]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Fitri Utaminingrum]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 1 (2019): Januari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (648.489 KB)
<![CDATA[The face is one of the characteristics of human natural physiology that can be used for biometric identification for facial recognition. Face recognition is an alternative to systems such as presence and authentication. Nowadays there are so many companies or researchers to create a system that can recognize people's faces, but there is still a face recognition system that can be tricked by showing people who have been recognized by the system in the system's camera area, even though people who are actually recognized by the system are not in the area that. This research will utilize the LVQ method for classification or facial recognition because it is well proven in face recognition conducted by previous research. Feature extraction is used in the form of skin image taking with HSV color space because HSV color space is better at detecting skin images according to existing research. The unique face image or pose used consists of 3 different eye poses to improve the safety of face recognition. In 10 different test scenarios, the results of this study have an average accuracy of 81.3%. However, the system still cannot distinguish each pose from the existing data.]]>
<![CDATA[Klasifikasi Multilabel Menggunakan Metode Fuzzy Similarity K-Nearest Neighbor Untuk Rekomendasi Pencarian Artikel Online ]]>
<![CDATA[Wahyuni Lubis]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Mochammad Ali Fauzi]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 1 (2019): Januari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (488.852 KB)
<![CDATA[The article is someone's opinion of the paper that addresses a specific problem that is actual and sometimes controversial to inform, influence, persuade, and entertain the reader. Rapid technological developments led to the large number of articles written online. Each article has a different label online, and allows each article has more than one label. The number of online articles that exist on the internet every day growing which makes the reader's difficulty in finding the desired information. The proper classification can improve the quality of information retrieval. A method of Fuzzy K-Nearest Neighbor Similarity is a method that combines the multilabel classification method of Fuzzy Similarity Measure and MLKNN. Previous research on method FSKNN has better speed in doing computing k nearest neighbors and better performance of the method MLKNN. The steps undertaken in this research is conducting a text preprocessing, document clustering, weighting, classification and search process. On the research of the optimal values obtained this F1 and BEP amounted to 0.933 and 0.937 at k = 1 and alpha = 0.5. On the recommendation of the search articles online using the method FSKNN obtained the highest precision value of 0.5 and 0.8 recall. From the results of F1 and the BEP obtained, indicating that the method FSKNN was kind enough to do a multilabel classification articles online.]]>
<![CDATA[Implementasi Algoritme Genetika dan Analytical Hierarchy Process untuk Penerimaan Siswa Baru pada Sekolah Menengah Kejuruan ]]>
<![CDATA[Mala Nurhidayati]]>;
<![CDATA[Dian Eka Ratnawati]]>;
<![CDATA[Yuita Arum Sari]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 1 (2019): Januari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (453.148 KB)
<![CDATA[Vocational high schools aim to produce qualified graduates in their expertise and skills in each majors. In the process of getting qualified graduates, it takes an initial step, namely selecting new students. The selection process is carried out by considering the value of the Indonesian National Examination, the value of the English National Examination, the value of the National Mathematics Examination, and the National Examination Science score. The methods used to solving this problem are Analytical Hierarchy Process (AHP) and Genetic Algorithms. Individuals in the Genetic Algorithm method are 6 genes. In the initial individuals formed the reproduction process namely crossover and mutation. The last process in this reasearch is the calculation of fitness value. This process resolved by calculating the results of accuracy. The results of the fitness value obtained the largest to the smallest fitness value. From research conducted the output is passing or not passing from prospective students in a predetermined majors. The average accuracy of the combination of the AHP method and the genetic algorithm is 86.85% and the accuracy is obtained using the AHP method is only 76.52%. Based on these results it can be concluded that the merger of AHP methods and genetic algorithms can solve the problem of new student admission at vocational high schools.]]>
<![CDATA[Penentuan Jumlah Kendaraan Menggunakan Blob Detection dan Background Subtraction ]]>
<![CDATA[San Sayidul Akdam Augusta]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Putra Pandu Adikara]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 1 (2019): Januari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (425.013 KB)
<![CDATA[Traffic is an important problem because traffic is medium to move from one place to another. When there is a traffic problem or become stuck, then people's mobility will have problem too. Traffic density data is an important role to understand traffic condition. Currently, in order to obtain traffic density still do it in conventional way, that is with some people to count each vehicles passing by at certain time. The purpose of this research is to apply algorithm Background Subtraction and Blob Detection to determine total vehicles and test the result of the vehicle counter system. Background Subtraction is used to process segmentation to separate an object with the background by counting difference between each pixel and use a threshold to make two dominant group of pixel. The method used to determine object position and total vehicles by Blob Detection and Background Subtraction. Testing done with twenty image by taking smallest error value as a best evaluation. The performance of precision is 93.44%, recall is 77.03% and accuracy is 73.75%. The value of precision, recall and accuracy needs to be increased again by adding test parameters and multiplying datasets with different conditions. The results show that the Blob Detection and Background Subtraction methods can give pretty good results when blob between vehicles is spaced. This method does not provide good results when used in heavy traffic conditions with vehicle bodies sticking together.]]>
<![CDATA[Rekomendasi Resep Masakan Berdasarkan Ketersediaan Bahan Masakan Menggunakan Metode N-Gram dan Cosine Similarity ]]>
<![CDATA[Ratna Tri Utami]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Indriati Indriati]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 2 (2019): Februari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (415.522 KB)
<![CDATA[Cooking recipes are the guidelines of a housewife in making a dish. Many recipes that there are easy for housewives to cook. But the reality, there are still a lot of housewife who doesn't understood the compatibility between the composition of cooking materials with dishes to be made. So it takes innovation to facilitate the search for a recipe in accordance with the composition of the available ingredients. It can be included in a form of information retrieval system. N-gram and cosine similarity methods can be used to match the available ingredients with the recommended recipes. Excess cosine similarity method didnt affect by the short length of a text document, because it just calculated only the term value of each document. The N-gram method consists of 3 types of processes: unigram, bigram, trigram which are serves for word processing. In this research, a model for recommendation of relevant recipes using N-gram method and cosine similarity was developed. The tests performed were the measurement of similarity and threshold determination. The results obtained that the system succeeded in calculating the similarity with the value of cosine 0.9. The greater of the value so it closer to the recommendation of the recipe in accordance with the query. From the third results of the best N-gram process is unigram with a threshold value is greater than or equal to 90% and a recall value of 1 and precision 0,2. It can be concluded that unigram is the best N-gram method process to recommend the recipes based on the ingredient.]]>
<![CDATA[Pengenalan Sidik Jari Balita Menggunakan Metode Zone Based Linear Binary Pattern dan Extreme Learning Machine ]]>
<![CDATA[Dea Valentina]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Rendi Cahya Wihandika]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 2 (2019): Februari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (1240.476 KB)
<![CDATA[Fingerprint recognition is one of the technological developments that is feasible since childhood. Along with the increasing number of toddlers, an introduction system is needed that is able to uniquely identify toddlers with the biometric patterns they have. Toddler fingerprint patterns have a low contrast between ridges and valleys and have a size (distance between ridges) that is smaller than adult fingerprints making it difficult to design accurate algorithms that are able to extract important features and match them in a strong way. In this study, the process begins with pre-processing and then uses the Zone Based Linear Pattern method to extract features on toddler fingerprints and the Extreme Learning Machine (ELM) classification method to recognize the identity of the fingerprint owner. The test results using a combined binary pattern for the Zone Based Linear Binary Pattern extraction method, the gaussian filtering, opening and adaptive thresholding technique for pre-processing images with dimensions of 200x200 in the image, the z-score method for normalizing data and the number of hidden neurons by 50 with binary sigmoid activation function for ELM classification produces the best accuracy of 72.33%. Based on these results it can be concluded that the Zone Based Linear Binary Pattern and Extreme Learning Machine methods can be used to recognize toddler fingerprints.]]>
<![CDATA[Seleksi Fitur Information Gain Pada Klasifikasi Citra Makanan Menggunakan Hue Saturation Value dan Gray Level Co-Occurrence Matrix ]]>
<![CDATA[Frisma Yessy Nabella]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Randy Cahya Wihandika]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 2 (2019): Februari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (590.664 KB)
<![CDATA[Choosing food has become a challenge for those who are presented with new food choices. Classification is important for those who have a strict diet regarding food that they consume. Food selection is essential for those who are visually impaired to identify food items. The classification process in this research is initiated with the pre-processing of the image, resulting in a segmented image which is then continued with feature extraction where Hue Saturation Value (HSV) for color extraction and Gray Level Co-Occurence Matrix (GLCM) for texture features. Based on features that have been extracted the next step is to gather relevant features using Information gain to reduce the workload. The last process is classification, using K-Nearest Neighbor. Accuracy results are 95,24% at best using only HSV with k=1 for feature selection. A combination of HSV and GLCM using Information gain results in a accuracy from 57,14% to 87,61%. This also applies to only using GLCM with information gain that raises the accuracy from 57,14% to. 74,28%. With the previous statement taken into consideration, Information Gain as a feature selection method increases accuracy with a significant amount and is able to obtain relevant feature if the list of features is abundant. If there are only a few features used, the accuracy increment is not that significant but it decreases the workload of the system.]]>
<![CDATA[Seleksi Fitur Information Gain pada Klasifikasi Citra Makanan Menggunakan Ekstraksi Fitur Haralick dan YUV Color Moment ]]>
<![CDATA[Devinta Setyaningtyas Atmaja]]>;
<![CDATA[Yuita Arum Sari]]>;
<![CDATA[Randy Cahya Wihandika]]>
<![CDATA[ Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer Vol 3 No 2 (2019): Februari 2019 ]]>
Publisher : <![CDATA[Fakultas Ilmu Komputer (FILKOM), Universitas Brawijaya ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
Full PDF (383.913 KB)
<![CDATA[Food classification is a classic problem that still becomes an interest for many researchers. Several studies have been conducted using only one type of food, which is fruit as the object of the classification. This research was conducted to improve the previous ones. This study uses five types of single food as its object. The method used is color feature extraction using YUV Color Moment, texture feature extraction using Haralick, and feature selection using Information Gain. The classification algorithm is K-Nearest Neighbor (KNN). The highest accuracy obtained is 94.26% obtained from the combined features of the two selected feature extraction methods. From these results, it can be concluded that the application of a combination of feature extraction methods, namely color and texture, and feature selection method greatly influence the food image classification process.]]>
