Garuda - Garba Rujukan Digital

Application of YOLOv8 Model for Early Detection of Diseases in Bean Leaves Indra Yustiana; Alun Sujjada; Tirawati
bit-Tech Vol. 8 No. 1 (2025): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i1.2514

Bean plant is one of the high economic value horticultural commodities widely cultivated in Indonesia. However, its productivity declines due to pest attacks and leaf diseases. Farmers' limitations in accurately identifying disease types also pose obstacles in early mitigation efforts. Therefore, technology-based solutions capable of quickly and accurately detecting plant diseases are needed. This research aims to develop and evaluate the performance of a leaf disease detection model for bean plants using the You Only Look Once version 8 (YOLOv8) algorithm with a transfer learning approach. The dataset used consists of 1,037 images of bean leaves, classified into three categories: angular leaf spots, leaf rust, and healthy leaves. Data were obtained from two sources, namely field documentation in Sindang Village, Sukabumi Regency, and an open repository on GitHub. The dataset was divided into training data (70%), validation (20%), and testing (10%). The model was trained using the YOLOv8s architecture for 30 epochs and achieved a detection accuracy of 85%. Performance evaluation was conducted using precision, recall, and mean average precision (mAP) metrics. The results of this study are expected to be an initial contribution to the application of artificial intelligence in agriculture, particularly in helping farmers efficiently detect leaf diseases in beans to improve productivity and quality of harvest.

Implementation of Machine Learning Using Decision Tree Method for Social Assistance Recipient Classification Akbar Ilham Perhan; Indra Yustiana; Imam Sanjaya
bit-Tech Vol. 8 No. 1 (2025): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i1.2755

The distribution of social assistance in Indonesia often faces challenges in accuracy, where individuals who are financially capable still receive aid, while those truly in need are excluded. To address this issue, this study applies a Machine Learning approach using the C4.5 Decision Tree algorithm to classify the eligibility of recipients in Bojonggenteng Village. This algorithm was chosen because it is easy to interpret, performs well, and is suitable for categorical data. The main objective of the study is to develop a classification model that enhances the objectivity and accuracy in determining aid recipients, ensuring that assistance is directed to those who truly need it. The research process involves several stages, including problem identification, literature review, data collection, preprocessing, classification, and model evaluation. A total of 904 records from the 2023 BPNT and PBI-JK programs were obtained in collaboration with the local village authorities. The classification process was conducted using RapidMiner, which allows for visual data processing and model building without requiring programming. The model evaluation was carried out using a confusion matrix, yielding an accuracy of 98.90%, precision of 100%, recall of 97.60%, and an AUC score of 0.988. These results indicate that the C4.5 algorithm is effective for prediction tasks and can be a valuable tool in supporting fair and data-driven decision-making in social assistance programs. This study concludes that the application of Machine Learning in this context improves the fairness and transparency of aid distribution and recommends future research to involve larger datasets for broader implementation.

Implementation of Technique for Order Preference by Similarity to Ideal Solution for Selecting Content Ivana Lucia Kharisma; Indra Yustiana; Falya Amrina Zahra
bit-Tech Vol. 8 No. 1 (2025): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i1.2768

This study addresses the challenge faced by the Sukabumi Creative Hub Instagram team in identifying the most engaging content by proposing a web-based Decision Support System (DSS) utilizing the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method. Instagram, as a dominant social media platform in Indonesia, serves as a vital tool for promoting local creative industries, yet current content evaluation lacks systematic analysis. The system developed ranks 62 content items based on three engagement metrics—likes, views, and shares—weighted at 5, 3, and 1 respectively. Data were processed using Microsoft Excel and visualized through an Input-Process-Output (IPO) model. The results show that “Rekap Merangkum Sukabumi” achieved the highest relative closeness (RC = 0.8793), demonstrating TOPSIS’s effectiveness in ranking content based on proximity to ideal engagement levels. Compared to previous studies that applied TOPSIS in different contexts, this research offers a novel contribution by applying it to localized social media content, filling a gap in digital content analytics literature. Despite limitations such as subjective weighting, platform specificity, and manual calculations, the system offers a replicable, structured approach to content evaluation, with implications for improved social media strategy and future research in automated, cross-platform DSS applications. Ultimately, this study bridges practical needs in creative content management with theoretical development in decision support systems for digital engagement analysis.

Application Of Random Forest Algorithm in Music Recommendation System Using Content-Based Filtering Rubby Malik Fajar; Indra Yustiana; Alun Sujjada
bit-Tech Vol. 8 No. 2 (2025): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i2.2803

The rapid growth of digital technology has revolutionized how people access and listen to music, especially through online streaming platforms. However, the overwhelming number of available songs often confuses users, particularly new users who have no listening history. To address this, the study proposes a music recommendation system using a content-based filtering approach that recommends songs based on similarities in both textual and numerical features, such as genre, artist, lyrics, tempo, energy, and danceability. The system operates in two main stages. First, it classifies the popularity of songs into two categories, “High” and “Low,” using three classification algorithms: Random Forest, K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). Second, it generates music recommendations based on content similarity using TF-IDF and cosine similarity. Random Forest is chosen as the main algorithm due to its superior performance in high-dimensional data and its ensemble learning mechanism. The evaluation uses confusion matrix metrics including accuracy, precision, recall, and F1 score, tested across multiple data split ratios (90:10, 80:20, 70:30, 60:40). The results show that Random Forest consistently delivers better classification and recommendation performance compared to KNN and SVM. It demonstrates higher accuracy and F1 score, making it suitable for real-world applications. The system is developed using Streamlit, allowing users to interactively receive music recommendations through a user-friendly web interface. The findings support the integration of Random Forest in content-based recommendation systems to improve accuracy and solve cold-start problems effectively in digital music platforms.

Web-Based Drug Inventory System with FIFO Method and SCM Approach at Risa Farma Pharmacy Mayang Selpiyana; Anggun Fergina; Indra Yustiana
bit-Tech Vol. 8 No. 2 (2025): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i2.2951

Drug inventory management plays a crucial role in pharmacy operations as it directly relates to drug availability, service accuracy, and the prevention of losses caused by expired medicines. However, many small-scale pharmacies still rely on manual recording, which is prone to stock discrepancies, delayed detection of near-expired drugs, and low administrative efficiency. To address these issues, this study developed a web-based drug inventory system implementing the First In First Out (FIFO) method combined with a Supply Chain Management (SCM) approach. The system was developed using the System Development Life Cycle (SDLC), which includes planning, requirements analysis, system design, implementation, and functional as well as user testing. The results demonstrate that the system successfully reduces potential stock recording errors by up to 75% compared to manual methods and improves transaction recording efficiency by an average of 40%. The SCM approach enables the system to automatically issue alerts when stock reaches a minimum threshold and to provide restock recommendations based on demand data. A User Acceptance Test (UAT) involving 10 respondents produced a satisfaction score of 86%, indicating that the system is effective, user-friendly, and beneficial for pharmacy operations. This study concludes that the integration of FIFO and SCM in a web-based drug inventory system improves data accuracy, enhances operational efficiency, and minimizes the risk of expired medicines. Nevertheless, the system still faces limitations related to data security, internet dependency, and its suitability primarily for small-scale pharmacies.

Utilization Of Content-Based Filtering Method In Game Recommendation With Support Vector Machine Algorithm Indra Yustiana; Ivana Lucia Kharisma; Ade Arian
bit-Tech Vol. 8 No. 2 (2025): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i2.2966

The explosive growth of the gaming industry has led to a "paradox of choice," where an overwhelming number of titles on platforms like Steam makes it difficult for players to discover games aligned with their personal preferences. This study addresses the critical challenge of game discovery by developing a novel recommendation system that integrates a Content-Based Filtering (CBF) approach with the Support Vector Machine (SVM) algorithm, a combination not extensively explored in the gaming domain. The system provides accurate, attribute-driven recommendations to enhance user experience. Utilizing a dataset of over 55.691 Steam games, we processed textual data such as genre, tags, and categories using TF-IDF before applying the SVM classifier. To validate its effectiveness, the model was benchmarked against K-Nearest Neighbor (KNN) across various training-to-testing ratios. The results demonstrate SVM's consistent superiority, achieving up to 98% accuracy. Notably, the high F1-score of 97.94% in genre-based recommendations signifies a well-balanced model that excels at both minimizing irrelevant suggestions and identifying relevant titles, directly translating to higher user satisfaction. The successfully deployed system, built on the Streamlit framework, was validated through black-box testing, confirming its functionality. This research confirms that the CBF-SVM model offers a highly effective solution to the game discovery problem, with future potential to incorporate hybrid filtering techniques for even greater personalization.

Title

Found 6 Documents
Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title Search

Found 6 Documents Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title

Found 6 Documents
Search