Article Per Year (5 Year)
p-Index From 2021 - 2026
0.702
P-Index
This Author published in this journals
All Journal Jurnal Infra
Alvin Nathaniel Tjondrowiguno
Program Studi Informatika
Computer Science & IT

Published : 19 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 19 Documents
Search

 1   2 
Menghasilkan Background Game Music dengan Menggunakan Deep Convolutional Generative Adversarial Network Daniel Widjojo; Henry Novianus Palit; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 8, No 2 (2020)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

One of many important factor in producing a good quality game is the existence of a good quality Background Game Music (BGM). But the difficulty of getting game music assets makes game developers have to pay extra money, waste more time and effort to get a good quality game music. This can hinder the process of making a game. For that reason, if there is a program that is able to produce BGM with good quality, it will greatly help the work of the game developer.The method used in this study is the Deep Convolutional Generative Adversarial Network (DCGAN). The data that being used is Musical Instrument Digital Interface (MIDI) file format which will then be converted into a pianoroll format. The pianoroll will then be converted into a matrix and entered into the DCGAN model. Before conducting training process, it is necessary to make a preprocessing, postprocessing and a model DCGAN. Testing in this study is done by finding the best parameters and architecture of DCGAN to produce Background Game Music with good quality.The test results show that DCGAN is a very sensitive model in terms of architecture and hyperparameter, therefore it needs extra attention in tuning architecture and hyperparameter for DCGAN. Besides that, music that is converted into pianoroll format lacks the ability to highlight its features and making it difficult to learn by DCGAN. From the end results, DCGAN is able to produce Background Game Music but with poor quality.
Aplikasi Analisa Sentimen Pada Komentar Berbahasa Indonesia Dalam Objek Video di Website YouTube Menggunakan Metode Naïve Bayes Classifier Maximillian Christianto; Justinus Andjarwirawan; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 8, No 1 (2020)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

From the increasing interest of the people in Indonesia in the use of YouTube, this has triggered the emergence of content creators who choose YouTube as a medium for work. So that the content creators are competing to produce video works that can be enjoyed by YouTube users. Various methods are used by content creators to improve the quality of the video produced.The process carried out in this thesis is the process of processing raw data that has been collected from YouTube, before the training and classification process. The process of data preprocessing needs to be done to overcome the raw data that is varied and inconsistent so that it can affect the training process and the classification process. Data preprocessing conducted in this thesis includes Tokenization, Stopwords Removal, Stemming. The classification process is the process by which the classification algorithm is run on comments that are used as input data for comments.Applications used in scientific papers have succeeded in producing smoothing values, where the value shows that the comments belong to the classifications of positive sentiments, negative sentiments or non-Indonesian data.
Evaluasi Kinerja Penggabungan Knowledge Graph Embedded-Based Question Answering dan TransP pada Data Freebase Fransisco Remon Liemena; Henry Novianus Palit; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 8, No 2 (2020)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

In the past few years, data storage and analysis using graph keep increasing. One of the implementation of this is knowledge graph. There are many methods proposed on information extraction from knowledge graph, one of them is natural language question answering. However, all of the researches around question answering use direct query to find the answer. Knowledge Graph Embedding-based Question Answering (KEQA) is the latest method that implements deep learning and embedding to answer questions. Experiments demonstrate that KEQA outperforms other question answering methods. Despite having high accuracy, KEQA still uses simple and outdated embedding method.Knowledge graph embedding is one of the method for knowledge graph representation where the entities and relations are represented in vector (embedding) using deep learning. Many proposed embedding methods do not really consider the depth of a knowledge graph. TransP is a proposed method that consider the indirect relationship to represent a knowledge graph. Experimental results show that TransP outperforms other embedding methods in the task given. Based on this, KEQA will be built using TransP with the expectation that the accuracy of KEQA will increase.Based on the result of the experiment, TransP achieves Mean Rank of 5.390,25 and HIT10 of 28,5%. After that, KEQA with embedding can achieve up to 88,89% accuracy, and KEQA without embedding can achieve up to 88,89% accuracy. Experiment also shows that scoring parameters value with affect KEQA with embedding. In conclusion, TransP can increase the accuracy of KEQA.
Penerapan Recurrent Neural Network untuk Pembuatan Ringkasan Ekstraktif Otomatis pada Berita Berbahasa Indonesia Kristian Halim; Henry Novianus Palit; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 8, No 1 (2020)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Technology advancement in modern world allow huge amount of information to flow everyday and news became one of the source to get that information. Because of this advancement, available information through news have been increased and so program is develop to make summary of news to reduce reading time using the neural network as the basis of this program.The method used for training the model is Recurrent Neural Network. The type of Recurrent Neural Network that being used is Gated Recurrent Unit that is run in 2 level, the word level and then the sentence level. As for making the Recurrent Neural Network model, some experiment can be carried out, like changing initial weight of the word embedding, change the pooling method, removing dropout layer, and some preprocessing for the dataset.The results shows that for the initial model, F1 – Score for ROUGE – 1, ROUGE – 2, and ROUGE – L can reach up to 80% when using extractive summary as the reference and up to 50% when using abstractive summary as the reference. The experiment shows that the best model is using training dataset as the initial word embedding weight, using average pooling and removing the dropout layer. The best experiment result gives F1 – Score 84.10 for ROUGE – 1, 83.10 for ROUGE – 2 and 83.31 for ROUGE – L using the extractive reference and 57.01 for ROUGE – 1, 51.17 for ROUGE – 2 and 55.10 for ROUGE – L using the abstrative reference.
Automatic Playlist Continuation Menggunakan Hybrid Recommender System Martin Andersen Linggajaya; Henry Novianus Palit; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 9, No 2 (2021)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

One of the most popular ways to listen to music is using playlists. The playlist feature can be improved by giving track recommendations to be added to certain playlists. To support the development of this recommendation process, ACM and Spotify held the RecSys Challenge 2018 with the task of automatic playlist continuation. This research is a continuation from [6] that placed 3rd in the RecSys Challenge 2018. The method used consists of 2 phases: candidate selection using a hybrid recommender system called LightFM and ranking using XGBoost. The research gap being developed focuses on one of the calculations for co-occurrence features used in the ranking phase. The result of this research shows that co-occurrence of 3 tracks does not improve the performance of the model used. The model by [6] achieved scores of 0.5251, 0.5582, and 1.295 for R-precision, NDCG, and recommended song clicks respectively. Meanwhile, the model produced in this research achieved an R-precision of 0.5241, an NDCG of 0.5579, and recommend song clicks of 1.312.
Deteksi Rumus Matematika pada Halaman Dokumen Digital dengan Metode Convolutional Neural Network Martina Marcelline Taslim; Kartika Gunadi; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 7, No 2 (2019)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1485.281 KB)

Abstract

Mathematical formulae in academic papers or scientific journals are an important part of said documents. However, mathematical formulae are oftentimes not properly recognized by Optical Character Recognition (OCR) processes. One of the causes of this failure is the difference between mathematical formulae and ordinary text. Therefore, mathematical formula detection in those document pages might help with this problem. The formula detection is done by converting digital document pages into images, then performing text line segmentation and word segmentation and classifying those results with a Convolutional Neural Network. The aim is to help OCR processes by recognizing which parts of the document pages contain formulae and which parts do not. The CNN architectures used to perform classification comes with 64 kernels in each convolutional layer. For displayed formulae (formulae that doesn’t share its space with regular text), the model uses 10 groups of Convolutional-ReLU-Max Pooling layers. For inline formulae (formulae that shares its text line with regular text), 12 groups of Convolutional-ReLU-Max Pooling layers are used. Results of the CNN architectures mentioned above are an F1 score of 0,980 for displayed formulae classification in 1-column documents, 0,940 for 2-column documents, and 0,916 for inline formulae. 
Sistem Rekomendasi Film menggunakan User-based Collaborative Filtering dan K-modes Clustering Ichwanto Hadi; Leo Willyanto Santoso; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 8, No 1 (2020)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Film is one of the popular entertainment media in the community. The number of titles that have been released makes it difficult for people to find which movies they want to watch. To overcome this problem, it is necessary to have information about the film that will make it easier for the public to find films that fit the user's preferences, therefore the user needs a system that can provide movie recommendations.The movie recommendation system using User-based Collaborative Filtering is one method that is able to provide recommendations. K-mode Clustering can also be used as an additional accuracy of recommendations by grouping user preferences history.According to the results of the testing of the k-modes clustering method, the best number of clusters for K-Modes Clustering for film recommendations obtained using the Elbow Method and Silhouette Coefficient is k = 3. From the results of testing the accuracy of the recommendations with Mean Reciprocal Rank (MRR) generated average MRR of 0.17092270381865 for film recommendations with a data train and test ratio of 80%: 20% and an average MRR of 0.15072658511145 for film recommendations with a data train and test ratio of 60%: 40%. From the results of the two tests above, it can be concluded that the level of accuracy of the film recommendations according to the MRR is sufficient because the MRR is close to 0.
Indoor Room Recognition Menggunakan Multiple Instance Learning Convolutional Neural Networks Yehezkiel Wuisang; Djoni Haryadi Setiabudi; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 9, No 2 (2021)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Environment recognition is a modern problem that appears in this modern era. One of them is how a room’s type can be identified. Indoor room is a very challenging environment to identify because the identity of a room is represented by various types of objects in the room which by itself have various sizes and shapes. With the development of technology, especially machine learning, the type of room can be recognized automatically by a system with the help of Image Processing and Artificial Neural Network. This study uses the Mean-Shift algorithm to segment images and the Convolutional Neural Network (CNN) method assisted by the application of Multiple Instance Learning (MIL) so as to form the Multiple Instance Learning Convolutional Neural Network (MILCNN) method to identify room types. During training and testing, adjustments will be made to the method so that it can be applied in recognizing room types only through image labels without looking for individual object labels on images. This study classifies the room that contains an image by recognizing the features of the objects in it. The final result from testing the dataset produces a classification accuracy percentage that reaches 43.05%.
Implementasi Naive Bayes Classifier dalam Analisa Rekrutmen di PT. Mitra Pinasthika Mulia Ardian Teja; Silvia Rostianingsih; Alvin Nathaniel Tjondrowiguno
Jurnal Infra Vol 7, No 2 (2019)
Publisher : Jurnal Infra

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (462.622 KB)

Abstract

PT. Mitra Pinasthika Mulia is a company engaged in the distribution of motorcycles, spare parts and accessories that require applications for applicant analysis or prediction. At present, PT. Mitra Pinasthika Mulia get applicants in several ways, namely from applicant's letter, website, job fair, and job street. There are many applicant data entered and that data is still manually selected. But this is not effective because it will require time and effort, therefore this application was made to help speed up the applicant selection process by providing predictive values to each applicant. The method used is one of the data mining methods, which is Naive Bayes Classifier. The Naive Bayes Classifier method is one method of classification or prediction with a model for calculating probabilities from a category that has existing attribute parameters, and determines which category is the most optimal. The attribute parameters used to measure applicants are age, gender, experience, education, language skills, identity, and applied position. The results of this study showed best f-score of 45,9% and an accuracy of 97,25% for predictions of acceptance, and best accuracy rate of 96,87% for predictions of work position.
Co-Authors Agustinus Noertjahyana Aldy Noah Alles Sandro Oktavio Gandadireja Ananta Kusuma Pangkasidhi Angelika Dibijo Ardian Teja Daniel Adi Daniel Widjojo Djoni Haryadi Setiabudi Fransisco Remon Liemena Henry Novianus Palit Ichwanto Hadi Joshua Maximillian Justinus Andjarwirawan Kartika Gunadi Kristian Halim Leo Willyanto Santoso Lily Puspa Dewi Martin Andersen Linggajaya Martina Marcelline Taslim Maximillian Christianto Nathaniel Edward Nicholas Sutikno Regina Valentina Rolly Intan Silvia Rostianingsih Welly Winata Yehezkiel Wuisang