Garuda - Garba Rujukan Digital

Non-Negative Matrix Factorization Based Recommender System using Female Daily Implicit Feedback Hani Nurrahmi; Agung Toto Wibowo; Selly Meliana
Indonesia Journal on Computing (Indo-JC) Vol. 7 No. 1 (2022): April, 2022
Publisher : School of Computing, Telkom University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.34818/INDOJC.2022.7.1.599

Recommender Systems is widely used by e-commerce to provide recommendations of products that are probably to be the interest to users. One of the recommender system algorithms that can be implemented is Non-negative Matrix Factorization (NMF) which receives explicit feedback in the form of user ratings. Although this method is effective, there are problems faced by explicit feedback as input, e.g. there are users who act as grey-sheep or black-sheep by providing dishonest ratings as explicit feedback. On the opposite, dishonest feedback least frequently occurs in implicit feedback. Therefore, in this study, we used implicit feedback to recommend products by taking the implicit feedback obtained from Female Daily’s mobile application as a case study. There are three types of implicit feedback: View Product Detail, View Review Detail, and Add to Wishlist. We experimented with the NMF algorithm provided by Surprise library using two implicit ratings weighting scenarios: accumulative weighting and maximum weighting. We combined several NMF parameters and run our experiment in 5-fold cross-validation. The best performance result in accumulative weighting is MSE = 1,2969, RMSE = 1,1388, MAE = 0,7909. Meanwhile, the best performance result in maximum weighting is MSE = 0,6742, RMSE = 0,8211, MAE = 0,5924.

Tourism Recommendation System using Weighted Hybrid Method in Bali Island Diffo Elza Pratama; Dade Nurjanah; Hani Nurrahmi
JURNAL MEDIA INFORMATIKA BUDIDARMA Vol 7, No 3 (2023): Juli 2023
Publisher : Universitas Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/mib.v7i3.6409

Tourism is a promising sector for global economic growth, as it has shown resilience during the global crisis. In Bali, tourism is a leading sector alongside agriculture and industry, making a significant contribution to regional and community development. However, Bali's popularity as a sought-after tourist destination also raises the need for an information system that can provide destination recommendations. To overcome the problem of information overload, a recommendation system is needed. This study tested the tourism recommendation system in Bali using the Weighted Hybrid technique which combines two methods, namely Collaborative Filtering and Content-Based using the weighted value technique. Collaborative Filtering, Content-Based, and Weighted Hybrid approaches will be compared in this study to improve the performance and accuracy of current recommendation systems. Utilizing the MAE, MSE, and RMSE values, the evaluation is carried out by comparing the evaluation matrices of the three Collaborative Filtering, Content-Based, and Weighted Hybrid methods. With MAE, MSE, and RMSE values of 0.4854, 0.4034, and 0.6351 respectively, the evaluation findings show that the Weighted Hybrid technique beats Collaborative Filtering and Content-Based with a weight value of 0.4.

Misogyny Text Detection on Tiktok Social Media in Indonesian Using the Pre-trained Language Model IndoBERTweet Perwira Hanif Zakaria; Dade Nurjannah; Hani Nurrahmi
JURNAL MEDIA INFORMATIKA BUDIDARMA Vol 7, No 3 (2023): Juli 2023
Publisher : Universitas Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/mib.v7i3.6438

Social media is a popular communication and information platform due to its ease and speed of access. By using social media, one can express himself freely. This triggers irresponsible individuals to utter hate speech with the aim of bringing down a person or group of people. Misogyny is a form of hate speech directed at women. The problem of misogyny should not be underestimated because misogyny can be one of the main reasons women feel miserable. In this study, a model will be built to detect misogyny text on the Indonesian language TikTok social media using the IndoBERTweet pre-trained model. IndoBERTweet is a pre-trained model based on the BERT model, which has been trained using Indonesian language datasets taken from the previous Twitter social media, resulting in a good performance for detecting misogynous texts on social media by classifying them. The dataset used is in the form of text data taken from misogyny comments by focusing on forms of misogyny in the form of stereotypes, dominance, sexual harassment, and discredit in short video content on women's TikTok social media accounts. The performance of built model performs hyperparameter settings which include batch size 16, epochs 10, and learning rate 7e-5 and is evaluated using a confusion matrix with the best accuracy results of 76.89%.

Music Recommendation System Using Alternating Least Squares Method Muhammad Rafi Irfansyah; Dade Nurjanah; Hani Nurrahmi
Indonesia Journal on Computing (Indo-JC) Vol. 9 No. 1 (2024): April, 2024
Publisher : School of Computing, Telkom University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.34818/INDOJC.2024.9.1.908

Music is not just entertainment, but it also has a positive impact on psychological well-being. The music landscape is generally dominated by millennials, especially in Indonesia. Music recommendation systems are becoming an important factor in offering songs that match users' preferences. Collaborative Filtering (CF), particularly the Alternating Least Squares (ALS) method, has become a popular solution for data sparsity problems in user-item interactions. Using the Precision@K metric, ALS provides the best results at a 50:50 data split ratio, 0.30225 for the Last FM dataset and 0.19742 for the Taste Profile dataset. Further analysis shows that ALS is more effective on datasets with balanced data distributions, such as Last FM, than on datasets with noisier characteristics, such as Taste Profile. The main conclusion is that ALS is suitable for use on datasets with balanced data distributions and can provide more optimal recommendations. For further development, handling sparsity data on Taste Profile needs to be improved to improve the performance of the recommendation model. This illustrates the importance of adapting the model to the unique characteristics of each dataset to achieve more accurate music recommendations.

Implementation of IndoBERT for Sentiment Analysis of Indonesian Presidential Candidates Primanda Sayarizki; Hasmawati; Hani Nurrahmi
Indonesia Journal on Computing (Indo-JC) Vol. 9 No. 2 (2024): August, 2024
Publisher : School of Computing, Telkom University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.34818/INDOJC.2024.9.2.934

In this modern era, Indonesian society widely utilizes social media, particularly Twitter, as a means to express their opinions. Every day, various opinions of Indonesian citizens are disseminated on this platform, including their views on prospective presidential candidates for the year 2024. Analyzing public opinions regarding prospective presidential candidates in 2024 is crucial to understanding the sentiment of the people toward these candidates. Such sentiment analysis can be conducted using deep learning techniques such as IndoBERT to acquire knowledge regarding the classification of sentiments as positive, neutral, or negative. IndoBERT is employed to generate vector representations that encapsulate the meaning of tokens, words, phrases, or texts. These representation vectors can then be input into a classification model to perform sentiment analysis. The sentiment classification model undergoes testing with a diverse set of tweets in the test dataset, which represent a wide range of public opinions. The evaluation results indicate an overall accuracy rate of 80%, with precision rates of 62% for negative sentiment, 81% for neutral sentiment, and 85% for positive sentiment. Additionally, the recall rates for each sentiment are 64% for negative, 81% for neutral, and 84% for positive, with corresponding F1-scores of 63%, 81%, and 85%, respectively.

Cyberbullying Detection on Twitter using Support Vector Machine Classification Method Putri Waisnawa, Ni Luh Putu Mawar Silveria; Nurjanah, Dade; Nurrahmi, Hani
Building of Informatics, Technology and Science (BITS) Vol 3 No 4 (2022): March 2022
Publisher : Forum Kerjasama Pendidikan Tinggi

Bullying is when someone or a group of individuals is continuously attacked. Because of the advancement of the internet, it has become very easy for society to engage in harmful acts of bullying by attacking a person or group of people who can hurt the victim, this is known as cyberbullying. Twitter is a social media platform that may be used by the society to share information and can also be used to perpetrate cyberbullying actions by sending messages (tweets) that addressed to the victims. This final project was developing a system to detect cyberbullying on Twitter. The system uses the Support Vector Machine method to classify whether the tweets that are shared include cyberbullying or not. In addition, this research also uses Term Frequency-Inverse Document Frequency (TF-IDF) and N-gram feature extraction for data that has gone through the pre-processing stage. In collecting data, the author crawled tweets based on the keywords 'jelek', 'bodoh', 'goblok', 'brengsek', 'bangsat', 'memalukan', 'laknat', 'bacot' and 'pelacur'. The best performance results of the research is 76.2% accuracy, 73.2% precision, 78.2% recall and 75.6% F1-Score generated by the RBF kernel with a total of n=1

Hate Speech Detection on Twitter through Natural Language Processing using LSTM Model Arbaatun, Cepthari Ningtyas; Nurjanah, Dade; Nurrahmi, Hani
Building of Informatics, Technology and Science (BITS) Vol 4 No 3 (2022): December 2022
Publisher : Forum Kerjasama Pendidikan Tinggi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bits.v4i3.2718

Currently, social media is a place to express opinions. This opinion can be positive or negative. However, lately, the opinion that often appears is a negative opinion, such as hate speech. Hate speech is often found on social media, such as malicious comments intended to insult individuals or groups. Based on WeAreSocial data in 2021, one of the most used social media platforms in Indonesia is Twitter, with 63.6% of users. According to the Indonesia National Police, hate speech cases were more dominant during the period from April 2020 to July 2021. Therefore, efforts are needed to identify hate speech on the Twitter platform. One way to detect hate speech is by using deep learning. In this research, we use a deep learning model of Long Short-Term Memory (LSTM) with word embedding. FastText and Global Vector (GloVe) is the word embeddings that we use as input for word representation and classification. FastText embeddings make use of subword information to create word embeddings and GloVe embeddings using an unsupervised learning method trained on a corpus to generate distributional feature vectors. From the evaluation results on the experimental model, LSTM-FastText using random oversampling has an advantage with an F1-score of 89.91% compared to LSTM-GloVe to obtain an F1-score of 82.14%.

Diversity Balancing in Two-Stage Collaborative Filtering for Book Recommendation Systems Rifqi Fauzia Muttaqien; Dade Nurjanah; Hani Nurrahmi
JURNAL TEKNIK INFORMATIKA Vol 16, No 2 (2023): JURNAL TEKNIK INFORMATIKA
Publisher : Department of Informatics, Universitas Islam Negeri Syarif Hidayatullah

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.15408/jti.v16i2.36580

A book recommender system is a system used to provide relevant book recommendations for readers. One approach that is often used in recommender systems is Collaborative Filtering (CF). CF provides book recommendations based on books liked by other similar users. However, CF only provides recommendations for items that are popular, so items that are less popular will be difficult to recommend. Therefore, we propose a book recommendation system based on Two-stages CF using the Diversity Balancing method. Diversity Balancing method in CF is used to balance diversity in the recommendation results by replacing popular items with less popular relevant items. System accuracy is measured using precision and recall, while diversity is measured using personal diversity and aggregate diversity. The test results show that the accuracy of the proposed system increases with the increasing number of recommended items. meanwhile, the diversity of recommended items continues to decrease as more items are included in the recommendation list. In consideration of the trade-off between accuracy and diversity, our system achieves a recall score of 0.301, a precision score of 0.282, a PD score of 0.048, and an AD score of 0.095 with a recommendation list size of 8 items.

Design and Implementation of a Performance Dashboard for Public Relations at Telkom University: Perancangan dan Implementasi Dashboard Kinerja Bidang Humas Universitas Telkom Dea Reskyadita, Feddy; Nurrahmi, Hani; Maulana , Daris
NUANSA INFORMATIKA Vol. 19 No. 2 (2025): Nuansa Informatika 19.2 Juli 2025
Publisher : FKOM UNIKU

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.25134/ilkom.v19i2.377

The Public Relations and Analytics (PRA) division at Telkom University faces challenges in managing manual processes and integrating data from diverse sources, hindering efficient reputation management. This study developed PANDA (Public Relations and Analytics Dashboard Application), a web-based system with an interactive performance dashboard to optimize PRA operations. Using the Waterfall methodology, the research encompassed requirements analysis, system design, implementation, and initial evaluation. PANDA integrates real-time key performance indicators (KPIs) such as media coverage, social media engagement, website analytics, and enabling data-driven decision-making. The System Usability Scale (SUS) evaluation resulted in a high score of 82 (categorized as 'Good to Excellent'), demonstrating PANDA's effectiveness in enhancing usability and streamlining operational workflows. The system enhances PR performance monitoring and offers a scalable model for educational institutions.

Music Recommendation System Using Alternating Least Squares Method Irfansyah, Muhammad Rafi; Dade Nurjanah; Hani Nurrahmi
Indonesian Journal on Computing (Indo-JC) Vol. 9 No. 1 (2024): April, 2024
Publisher : School of Computing, Telkom University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.34818/INDOJC.2024.9.1.908

Music is not just entertainment, but it also has a positive impact on psychological well-being. The music landscape is generally dominated by millennials, especially in Indonesia. Music recommendation systems are becoming an important factor in offering songs that match users' preferences. Collaborative Filtering (CF), particularly the Alternating Least Squares (ALS) method, has become a popular solution for data sparsity problems in user-item interactions. Using the Precision@K metric, ALS provides the best results at a 50:50 data split ratio, 0.30225 for the Last FM dataset and 0.19742 for the Taste Profile dataset. Further analysis shows that ALS is more effective on datasets with balanced data distributions, such as Last FM, than on datasets with noisier characteristics, such as Taste Profile. The main conclusion is that ALS is suitable for use on datasets with balanced data distributions and can provide more optimal recommendations. For further development, handling sparsity data on Taste Profile needs to be improved to improve the performance of the recommendation model. This illustrates the importance of adapting the model to the unique characteristics of each dataset to achieve more accurate music recommendations.

Title

Found 13 Documents
Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title Search

Found 13 Documents Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title

Found 13 Documents
Search