Abstract:Abstract: This study discusses the optimization of RAG for a FAQ system in the field of information technology product security certification at BSSN. Although LLM generate reliable responses, they often lack up-to-date…
and domain-specific knowledge, which can be addressed through the RAG approach. This research aims to optimize a domain-specific RAG system by improving embedding performance, enhancing prompt robustness, and increasing retrieval accuracy. The research methods consist of three stages. The first stage involves fine-tuning the bge-m3 embedding model and evaluating its performance using MRR, Recall, and AUC. The second stage applies prompt engineering techniques, namely the SRSM and Autodefense, to mitigate direct-injection and escape-character prompt injection attacks. The third stage evaluates the proposed RAG system using Precision, Recall, and F1-Score metrics against four baseline models. The results of research show that the fine-tuned embedding model achieves higher performance than the original model, with MRR@1 and Recall@1 values of 0.80 and an AUC@100 of 0.7023. In addition, the proposed prompt engineering techniques demonstrate robustness against prompt injection attacks, while the overall RAG system attains a perfect Precision, Recall, and F1-Score of 1.00. In conclusion, the proposed approach effectively enhances retrieval accuracy, embedding quality, and system security, resulting in a more reliable RAG-based FAQ system for information technology product security certification.
Keywords: embedding fine-tuning; large language model; prompt engineering; prompt injection mitigation; retrieval-augmented generation
Abstrak: Studi ini membahas optimasi RAG untuk sistem FAQ di bidang sertifikasi keamanan produk teknologi informasi di BSSN. Meskipun LLM menghasilkan respons yang andal, mereka seringkali kurang memiliki pengetahuan terkini dan spesifik domain, yang dapat diatasi melalui pendekatan RAG. Penelitian ini bertujuan untuk mengoptimalkan sistem RAG spesifik domain dengan meningkatkan kinerja embedding, meningkatkan ketahanan prompt dan meningkatkan akurasi pengambilan. Metode penelitian terdiri dari tiga tahap. Tahap pertama melibatkan fine-tuning model embedding bge-m3 dan mengevaluasi kinerjanya menggunakan Mean Reciprocal Rank (MRR), Recall, dan AUC. Tahap kedua menerapkan teknik rekayasa prompt, yaitu Self- SRSM dan Autodefense, untuk mengurangi serangan direct-injection dan escape-character prompt injection. Tahap ketiga mengevaluasi sistem RAG yang diusulkan menggunakan metrik Presisi, Recall, dan F1-Score terhadap empat model dasar. Hasil penelitian menunjukkan bahwa model embedding yang disempurnakan mencapai kinerja yang lebih tinggi daripada model asli, dengan nilai MRR@1 dan Recall@1 sebesar 0,80 dan AUC@100 sebesar 0,7023. Selain itu, teknik rekayasa prompt yang diusulkan menunjukkan ketahanan terhadap serangan injeksi prompt, sementara sistem RAG secara keseluruhan mencapai Presisi, Recall, dan F1-Score sempurna sebesar 1,00. Kesimpulannya, pendekatan yang diusulkan secara efektif meningkatkan akurasi pengambilan, kualitas embedding dan keamanan sistem, menghasilkan sistem FAQ berbasis RAG yang lebih andal untuk sertifikasi keamanan produk teknologi informasi.
Kata kunci: penyempurnaan embedding; model bahasa besar; rekayasa prompt; mitigasi injeksi prompt; retrieval-augmented generation
Abstract:Abstract: Tourism in Pati Regency currently lacks an integrated digital information system, resulting in suboptimal dissemination of information and trip planning. To address this issue, a tourism website for Pati Regency…
y was developed, equipped with a recommended tourist route feature. This study aims to design and develop a web-based tourism information system that provides destination information based on categories, media galleries, and promotional YouTube videos, as well as a Patiways feature that allows users to select multiple tourist destinations. The system then calculates the most efficient visiting order using a greedy heuristic algorithm, based on the selected starting point. The system was developed using the Waterfall method, consisting of analysis, design, implementation, and testing phases. The system design is illustrated through UML diagrams such as Use Case, Activity, and Class Diagrams. With this system, the distribution of tourism information becomes more effective, and tourists can plan trips with optimized routes. Additionally, the website is expected to serve as a digital promotion medium that contributes to increasing tourist visits to Pati Regency.
Keywords: heuristic greedy; recommendation route; tourism; waterfall
Abstrak: Pariwisata di Kabupaten Pati saat ini belum memiliki sistem informasi digital yang terintegrasi, sehingga penyebaran informasi dan perencanaan perjalanan wisata masih belum optimal. Untuk mengatasi permasalahan tersebut, penelitian ini mengembangkan sebuah website pariwisata Kabupaten Pati yang dilengkapi dengan fitur rekomendasi rute wisata terbaik. Penelitian ini bertujuan untuk merancang dan membangun sistem informasi pariwisata berbasis web yang mampu menyajikan informasi destinasi wisata berdasarkan kategori, galeri media, serta video promosi YouTube. Selain itu, sistem ini dilengkapi dengan fitur unggulan bernama Patiways yang memungkinkan pengguna memilih beberapa destinasi wisata dan secara otomatis memperoleh urutan kunjungan paling efisien menggunakan algoritma heuristik greedy berdasarkan titik awal perjalanan. Pengembangan sistem dilakukan menggunakan metode Waterfall yang meliputi tahapan analisis kebutuhan, perancangan sistem, implementasi, dan pengujian. Perancangan sistem direpresentasikan menggunakan diagram UML, meliputi Use Case Diagram, Activity Diagram, dan Class Diagram. Dengan adanya sistem ini, diharapkan penyebaran informasi pariwisata menjadi lebih efektif, wisatawan dapat merencanakan perjalanan dengan rute yang optimal, serta website dapat berfungsi sebagai media promosi digital yang berkontribusi terhadap peningkatan kunjungan wisatawan ke Kabupaten Pati.
Kata kunci: heuristik greedy; pariwisata; rekomendasi rute; waterfall
Abstract:Abstract: One of the main challenges in digital image processing is limited resolution, which makes it difficult to preserve visual details when images are enlarged. Conventional methods such as Bilinear Interpolation are…
e commonly used for image upscaling; however, these approaches often produce blurred images, lose fine textures, and fail to reconstruct complex visual structures. This study aims to enhance digital image resolution by employing a deep learni based approach using a Low-Light Convolutional Neural Network (LLCNN) built upon a Deep Neural Network (DNN) architecture. The dataset used in this study is the DIV2K dataset, which consists of 1,000 high-resolution images. These images were downsampled using scaling factors of ×2, ×3, and ×4 to generate paired Low Resolution–High Resolution (LR–HR) data for training and evaluation. The proposed LLCNN is designed to extract important features such as edges, textures, and local patterns through multiple convolutional layers, followed by non-linear mapping to reconstruct high-resolution images more accurately. Quantitative performance evaluation was conducted using the Peak Signal-to-Noise Ratio (PSNR) and the Structural Similarity Index (SSIM). Model performance was evaluated quantitatively using the Peak Signal-to-Noise Ratio (PSNR) metric. Experimental results showed that the proposed method improved image quality compared to the bilinear method. These results indicate that the deep learning based approach effectively improves image sharpness and structural fidelity, thereby demonstrating its potential for digital image resolution enhancement.
Keywords: deep neural network; image resolution; low-light convolutional neural network; machine learning
Abstrak: Permasalahan utama dalam pengolahan citra digital adalah keterbatasan resolusi yang menyebabkan detail visual sulit dipertahankan ketika citra diperbesar. Metode konvensional seperti Bilinear Interpolation masih banyak digunakan, namun sering menghasilkan citra buram, kehilangan tekstur halus, serta tidak mampu merekonstruksi struktur visual yang kompleks. Penelitian ini bertujuan untuk meningkatkan kualitas resolusi citra digital dengan memanfaatkan pendekatan deep learning berbasis Low-Light Convolutional Neural Network (LLCNN) yang dibangun di atas arsitektur Deep Neural Network (DNN). Data yang digunakan dalam penelitian ini berasal dari dataset DIV2K, yang terdiri dari 1000 citra beresolusi tinggi. Citra tersebut diturunkan menjadi resolusi rendah menggunakan faktor downsampling ×2, ×3, dan ×4 untuk membentuk pasangan data Low Resolution–High Resolution (LR–HR) sebagai data pelatihan dan pengujian. LLCNN dirancang untuk mengekstraksi fitur-fitur penting seperti tepi, tekstur, dan pola lokal melalui beberapa lapisan konvolusi, kemudian melakukan pemetaan non-linear guna merekonstruksi citra resolusi tinggi secara lebih presisi. Evaluasi performa model dilakukan secara kuantitatif menggunakan metrik Peak Signal-to-Noise Ratio (PSNR). Hasil eksperimen menunjukkan bahwa metode yang diusulkan mampu meningkatkan kualitas citra dibandingkan metode bilinear. Hasil ini membuktikan bahwa pendekatan berbasis deep learning efektif dalam meningkatkan ketajaman dan kesesuaian struktur citra digital.
Kata kunci: deep neural network; low-light convolutional neural network; machine learning; resolusi citra
Abstract:Abstract: Digital banking has brought various innovations in financial services, one of which is Blu Deposito by BCA Digital. However, the adoption rate of digital deposit services is still relatively low compared to digital…
ital payment services. This study aims to identify and analyze the factors that influence customers' intentions and actual behavior in using Blu Deposito with reference to the Technology Acceptance Model (TAM). This study aims to analyze the factors that influence customers' intentions and actual behavior in adopting Blu Deposito using the Technology Acceptance Model (TAM) framework. Data was collected through a Google Form questionnaire from 54 customers at one BCA branch and analyzed using SPSS through validity and reliability tests, descriptive analysis, and multiple regression. The results show that Behavioral Intention (BI)is significantly influenced by Perceived Ease of Use (PEOU), Perceived Usefulness (PU), and Attitude Toward Using (ATU), with PEOU as the most dominant factor. In addition, BI has a significant effect on Actual System Use (AU), which confirms the relevance of applying the TAM model in the context of digital deposit products. These findings indicate that ease of use plays a greater role than financial benefits in encouraging users to adopt Blu Deposits. This study contributes to the understanding of digital deposit adoption and provides managerial insights to improve the usability and user engagement of digital banking services.
Keywords: actual system use; attitude toward using; behavioral intention; perceived ease of use; perceived usefulness; technology acceptance model
Abstrak: Perbankan digital telah menghadirkan berbagai inovasi dalam layanan keuangan, salah satunya Blu Deposito oleh BCA Digital. Meskipun demikian, tingkat adopsi terhadap layanan deposito digital masih relatif rendah dibandingkan dengan layanan pembayaran digital. Penelitian ini bertujuan untuk mengidentifikasi dan menganalisis faktor-faktor yang memengaruhi niat serta perilaku aktual nasabah dalam menggunakan Blu Deposito dengan mengacu pada kerangka Technology Acceptance Model (TAM). Penelitian ini bertujuan untuk menganalisis faktor-faktor yang memengaruhi niat dan perilaku aktual nasabah dalam mengadopsi Blu Deposito dengan menggunakan kerangka Technology Acceptance Model (TAM). Data dikumpulkan melalui kuesioner Google Form dari 54 nasabah di satu cabang BCA dan dianalisis menggunakan SPSS melalui uji validitas, reliabilitas, analisis deskriptif, dan regresi berganda. Hasil penelitian menunjukkan bahwa Behavioral Intention (BI) dipengaruhi secara signifikan oleh Perceived Ease of Use (PEOU), Perceived Usefulness (PU), dan Attitude Toward Using (ATU), dengan PEOU sebagai faktor paling dominan. Selain itu, BI berpengaruh signifikan terhadap Actual System Use (AU), yang menegaskan relevansi penerapan model TAM pada konteks produk deposito digital. Temuan ini menunjukkan bahwa kemudahan penggunaan memiliki peran lebih besar dibandingkan manfaat finansial dalam mendorong pengguna untuk mengadopsi Blu Deposito. Penelitian ini berkontribusi terhadap pemahaman adopsi deposito digital serta memberikan wawasan manajerial untuk meningkatkan kegunaan dan keterlibatan pengguna pada layanan perbankan digital.
Kata kunci: actual system use; attitude toward using; behavioral intention; perceived ease of use; perceived usefulness; technology acceptance model
Abstract:Abstract: SMP Muhammadiyah 5 Samarinda still relies on manual evaluation with limited data analysis tools in predicting student academic achievement. This study aims develop a system for predicting the learning achievement…
nt of students at SMP Muhammadiyah 5 Samarinda using the Naive Bayes classification method. The dataset used consists of 192 student exam scores covering academic scores, attendance, parents’ education and income, and living conditions as independent variables, while the dependent variable is the achievement label (achieved or not achieved). The preprocessing stage includes label normalization, feature selection, and median imputation to handle missing data. The dataset was divided into 75% training data and 25%. The model was implemented as a pipeline consisting of a median imputer and a Gaussian Naive Bayes classifier. The evaluation results showed that the model achieved an accuracy of 79.2%, with a perfect recall value (1.00) in the high-achieving class and (0.64) in the low-achieving class. This shows that the model is quite effective in identifying high-achieving students. The trained model was then integrated into a Flask-based web application, which enables online predictions through a simple form interface, facilitating contextual interpretation. This system is expected to assist in educational decision-making by helping teachers identify students’ achievement levels early on and design more targeted learning interventions.
Keywords: academic performance; educational data mining; naive bayes; prediction system; student achievement
Abstrak: SMP Muhammadiyah 5 Samarinda masih bergantung pada evaluasi manual dengan alat analisis data terbatas dalam melakukan prediksi prestasi akademik siswa. Penelitian ini bertujuan mengembangkan sistem prediksi prestasi belajar siswa SMP Muhammadiyah 5 Samarinda menggunakan metode klasifikasi Naive Bayes. Dataset yang digunakan terdiri atas 192 data nilai ujian siswa yang mencakup skor akademik, kehadiran, pendidikan dan pendapatan orang tua, serta kondisi tempat tinggal sebagai variabel independen, sedangkan variabel dependen berupa label prestasi (berprestasi atau tidak berprestasi). Tahap preprocessing meliputi normalisasi label, seleksi fitur, serta imputasi median untuk menangani data yang hilang. Dataset dibagi menjadi 75% data latih dan 25%. Model diimplementasikan dalam bentuk pipeline yang terdiri atas median imputer dan Gaussian Naive Bayes classifier. Hasil evaluasi menunjukkan bahwa model mencapai akurasi sebesar 79,2%, dengan nilai recall sempurna (1,00) pada kelas berprestasi dan lebih rendah (0,64) pada kelas tidak berprestasi. Hal ini menunjukkan bahwa model cukup efektif dalam mengidentifikasi siswa berprestasi. Model yang telah dilatih kemudian diintegrasikan ke dalam aplikasi web berbasis Flask, yang memungkinkan prediksi secara daring melalui antarmuka formulir sederhana untuk mendukung interpretasi kontekstual. Sistem ini diharapkan dapat membantu untuk pengambilan keputusan dalam pendidikan dengan membantu guru mengidentifikasi tingkat prestasi siswa sejak dini dan merancang intervensi pembelajaran yang lebih terarah.
Kata kunci: prestasi akademik; penambangan data Pendidikan; naive bayes; sistem prediksi; prestasi siswa
Abstract:This study analyzes the acceptance of teachers and ASN employees of the SINAGA (Sistem Informasi Layanan Kepegawaian) attendance application at SMA Negeri 1 Jatilawang using a modified Technology Acceptance Model (TAM).…
The model was extended by incorporating two external variables: Information Quality and Complexity. This explanatory quantitative research employed the Structural Equation Modeling–Partial Least Square (SEM-PLS) method involving 60 respondents who are civil servants, consisting of teachers and administrative staff. The results reveal that Information Quality has a positive and significant influence on both Perceived Usefulness (PU) and Perceived Ease of Use (PEOU), while Complexity does not show a significant effect on either variable. Furthermore, PEOU and PU have a positive impact on Attitude Toward Use (ATU), which subsequently affects Behavioral Intention to Use (BIU). Behavioral intention, in turn, strongly influences Actual Use (AU). These findings indicate that teachers’ acceptance of the SINAGA digital attendance system in educational settings is primarily driven by information quality and users’ positive attitudes rather than by system complexity. Theoretically, this study contributes to the expansion of TAM application in the educational context. Practically, it provides valuable insights for improving the effectiveness of SINAGA implementation through better information quality and enhanced user experience.
Abstract:Abstract: The policy to increase the Rural and Urban Land and Building Tax (PBB-P2) in Indonesia often elicits mixed reactions from the public. Some support it because they believe it can strengthen regional fiscal capacity,…
ity, while others reject it because they are concerned that it will increase the economic burden on the community. Understanding public sentiment towards this policy is important for evaluating the effectiveness of the policy and formulating appropriate communication strategies. This study aims to analyze public sentiment towards the PBB-P2 increase policy using data uploaded on Platform X (Twitter). The data were collected through crawling with the keyword “building tax,” then processed through several preprocessing stages before classifying tweets into positive and negative sentiments. Two models were used: Support Vector Machine (SVM) and Bidirectional Long Short-Term Memory (BiLSTM). Results show that SVM outperformed BiLSTM, achieving training accuracy of 99.4% and testing accuracy of 85.9%, with accuracy 0.8595, precision 0.8536, recall 0.8595, and F1-score 0.8449. Meanwhile, BiLSTM achieved training accuracy of 86.9% and testing accuracy of 82.9%, with accuracy 0.8294, precision 0.8150, recall 0.8294, and F1-score 0.8080. These findings suggest SVM is more effective in classifying public sentiment and can support better evaluation of regional tax policies.
Keywords: sentiment analysis; PBB-P2; BiLSTM; SVM; X platform
Abstrak: Kebijakan kenaikan tarif Pajak Bumi dan Bangunan Perdesaan dan Perkotaan (PBB-P2) di In-donesia sering memunculkan beragam reaksi dari masyarakat. Sebagian mendukung karena dianggap dapat memperkuat kapasitas fiskal daerah, sementara lainnya menolak karena kha-watir menambah beban ekonomi masyarakat. Pemahaman terhadap sentimen publik atas ke-bijakan tersebut penting untuk mengevaluasi efektivitas kebijakan dan merumuskan strategi komunikasi yang tepat. Penelitian ini bertujuan menganalisis sentimen masyarakat terhadap kebijakan kenaikan PBB-P2 menggunakan data unggahan di Platform X (Twitter). Data dik-umpulkan melalui proses crawling dengan kata kunci “pajak bangunan” kemudian diproses melalui beberapa tahap preprocessing sebelum diklasifikasikan menjadi sentimen positif dan negatif. Dua model digunakan dalam penelitian ini, yaitu Support Vector Machine (SVM) dan Bidirectional Long Short-Term Memory (BiLSTM). Hasil penelitian menunjukkan bahwa SVM memiliki kinerja lebih baik dibandingkan BiLSTM, dengan akurasi pelatihan 99,4% dan akurasi pengujian 85,9%. Nilai akurasi 0,8595, precision 0,8536, recall 0,8595, dan F1-score 0,8449. Sementara itu, BiLSTM memperoleh akurasi pelatihan 86,9% dan akurasi pengujian 82,9%, dengan akurasi 0,8294, precision 0,8150; recall 0,8294; dan F1-score 0,8080. Temuan ini menunjukkan bahwa SVM lebih efektif dalam mengklasifikasikan sentimen publik serta dapat mendukung evaluasi kebijakan pajak daerah dengan lebih baik.
Kata kunci: analisis sentimen; PBB-P2; BiLSTM; SVM; platform X
Abstract:Abstract: The rapid growth of the digital music industry requires accurate music genre classification systems to enhance user experience in streaming services. This study compares a domain-specific Long Short-Term Memory…
(LSTM) network with three Large Language Models (LLMs)—HuBERT, WavLM, and WAV2Vec 2.0—for Music Genre Classification (MGC). The LSTM model was trained using Mel-spectrograms transformed from the GTZAN dataset, while the LLMs were fine-tuned using a smaller set of raw audio samples due to computational constraints. All models were tested on datasets with identical genre labels to ensure a fair evaluation. Results show that the LSTM model achieved the highest accuracy of 97.10%, outperforming HuBERT (86.00%), WavLM (83.00%), and WAV2Vec 2.0 (80.00%). The LSTM demonstrated superior generalization and stability without overfitting, while the LLMs struggled to differentiate between genres with similar acoustic characteristics. These findings indicate that general-purpose pre-trained models, although powerful, are less effective in music-specific tasks due to domain mismatch. Therefore, incorporating music-specific features and architectures remains essential for achieving higher accuracy and reliability in automatic genre classification systems.
Keywords: audio large language models; comparative deep learning; music genre classification.
Abstrak: Pertumbuhan industri musik digital yang pesat menuntut sistem klasifikasi genre musik yang akurat untuk meningkatkan pengalaman pengguna dalam layanan streaming. Penelitian ini dilatarbelakangi oleh perkembangan pesat model pembelajaran mendalam, khususnya jaringan LSTM dan model bahasa berskala besar LLM seperti HuBERT, WavLM, dan WAV2Vec 2.0, yang telah menunjukkan kemampuan representasi audio yang kuat. Tujuan penelitian ini ini membandingkan jaringan Long Short-Term Memory (LSTM) khusus domain dengan tiga model Large Language Models (LLM)—HuBERT, WavLM, dan WAV2Vec 2.0—untuk tugas Klasifikasi Genre Musik (MGC). Metode penelitian melibatkan pelatihan LSTM menggunakan data Mel-spectrogram hasil transformasi dari dataset GTZAN, sementara LLM disesuaikan (fine-tuning) menggunakan data audio mentah dalam jumlah lebih kecil karena keterbatasan komputasi. Seluruh model diuji pada dataset dengan label genre yang sama untuk memastikan evaluasi yang adil. Hasil penelitian menunjukkan bahwa model LSTM mencapai akurasi tertinggi sebesar 97,10%, sedangkan model HuBERT, WavLM, dan WAV2Vec 2.0 masing-masing memperoleh 86,00%, 83,00%, dan 80,00%. Model LSTM menunjukkan kemampuan generalisasi yang lebih baik tanpa overfitting, sedangkan model LLM cenderung kesulitan membedakan genre dengan karakteristik akustik yang mirip. Kesimpulan penelitian ini adalah ketidaksesuaian domain secara signifikan membatasi performa model umum saat diterapkan pada tugas berbasis musik. Oleh karena itu, penggunaan fitur dan arsitektur khusus musik sangat penting dalam membangun sistem klasifikasi genre yang lebih akurat.
Kata kunci: klasifikasi genre musik; model bahasa besar; perbandingan pembelajaran mendalam.
Abstract:Abstract: Soil moisture is an important factor in determining the watering needs of plants for optimal growth. Therefore, accurate monitoring of soil moisture is necessary. This research aims to design and build a soil moisture…
oisture detection tool based on the Decision Tree algorithm with the support of the YL-69 sensor for humidity measurement and the DHT11 sensor for temperature measurement to increase data accuracy. This system uses NodeMCU ESP8266 as a microcontroller and is integrated with an Android application as a user interface. Sensor interpretation data is analyzed using the Decision Tree algorithm to determine soil conditions (dry, damp or wet). The test results show an accuracy level of 95% from 300 data samples. Thus, this system is able to detect soil moisture effectively and can help increase the efficiency of crop management on a household and commercial agricultural scale.
Keywords: agriculture, android, decision tree algorithm, sensors, soil moisture detection
Abstract:Abstract: The rapid growth of the cosmetics industry on e-commerce platforms has intensified competition, creating a critical need for effective, data-driven marketing strategies. This study aims to conduct a comparative…
analysis of machine learning algorithms to predict the sales categories (High, Medium, Low) of cosmetic products on the Tokopedia marketplace. Four classification models; Random Forest, XGBoost, Logistic Regression, and Naive Bayes were trained and evaluated on data collected via web scraping. The methodology incorporates the Synthetic Minority Over-sampling Technique (SMOTE) to address significant class imbalance and GridSearchCV for hyperparameter optimization to ensure a fair and robust comparison. The experimental results conclusively show that the Random Forest model achieved the best performance, yielding the highest F1-Score Macro Average of 0.75 and an accuracy of 85.3%. The superior model was subsequently implemented in a simple recommendation system to simulate optimal discount strategies, demonstrating its practical utility in providing actionable insights for business decisions.
Keywords: classification; comparative analysis; machine learning; sales prediction; SMOTE
Abstrak: Pertumbuhan pesat industri kosmetik pada platform e-commerce telah membuat persaingan ketat, sehingga menciptakan kebutuhan krusial akan strategi pemasaran yang efektif dan berbasis data. Penelitian ini bertujuan untuk melakukan analisis komparatif terhadap algoritma machine learning untuk memprediksi kategori penjualan (Tinggi, Sedang, Rendah) produk kosmetik di marketplace Tokopedia. Empat model klasifikasi, yaitu Random Forest, XGBoost, Regresi Logistik, dan Naive Bayes, dilatih dan dievaluasi menggunakan data yang dikumpulkan melalui web scraping. Metodologi penelitian ini menerapkan Synthetic Minority Over-sampling Technique (SMOTE) untuk mengatasi ketidakseimbangan kelas yang signifikan dan GridSearchCV untuk optimisasi hyperparameter guna memastikan perbandingan yang adil. Hasil eksperimen menunjukkan bahwa model Random Forest mencapai performa terbaik, dengan menghasilkan F1-Score Macro Average tertinggi sebesar 0,75 dan akurasi 85,3%. Model unggul ini kemudian diimplementasikan dalam sebuah sistem rekomendasi sederhana untuk menyimulasikan strategi diskon yang optimal, yang menunjukkan kegunaan praktisnya dalam memberikan wawasan yang dapat ditindaklanjuti untuk pengambilan keputusan bisnis.
Kata kunci: analisis komparatif; klasifikasi; machine learning; prediksi penjualan; SMOTE