Claim Missing Document
Check
Articles

Found 4 Documents
Search

KLASIFIKASI FITUR WARNA LEVEL ROASTING BIJI KOPI MENGGUNAKAN ARTIFICIAL NEURAL NETWORK Tri Andre Anu; Rika Rosnelly; Dedi Irawan; Ubaidullah Hasibuan; Progresif Bulolo5
Device Vol 13 No 1 (2023): Mei
Publisher : Fakultas Teknik dan Ilmu Komputer (FASTIKOM) UNSIQ

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32699/device.v13i1.4094

Abstract

Abstract align="justify"Small and Medium Enterprises (SMEs) are using a manual method to notice the roasting level classification of coffee beans. However, the weaknesses in this technique are that the coffee roaster staff consumes time sorting the roasting level of the coffee beans. As a result, the coffee roaster focuses less because they take too long to sort the coffee beans—consequently, the mixed coffee beans in packages that should be elsewhere. Therefore a system is needed to help coffee roaster officers classify coffee beans using an artificial neural network. The data used are 60 coffee beans with three roasting levels: light roasted, medium roasted, and dark roasted. The classification process consists of a training stage and a testing stage. At the testing stage, using a sample of 30 coffee beans and based on the results of this study, the best results were obtained with a training value of 90%. In contrast, the testing accuracy was 66.67%.
Utilization of Digital Image and Convolution Neural Network Algorithm in Customer Satisfaction Survey with Facial Expressions Tri Andre Anu; Rika Rosnelly; Dedi Irawan; Progresif Bulolo; Ubaidullah Hasibuan
Journal of Computer Science, Information Technology and Telecommunication Engineering Vol 4, No 2 (2023)
Publisher : Universitas Muhammadiyah Sumatera Utara, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30596/jcositte.v4i2.15915

Abstract

The human face provides us with a lot of information about a person, and arguably the two most important pieces of information in a face are a person's identity and their emotional state. Judgments of identity and emotion facilitate social interactions. Services are a crucial part of the activities of all organizations, especially those in the service sector. Good services support customer satisfaction and ultimately impact the progress of the organization. The Convolutional Neural Network algorithm has become the most widely used neural architecture in various tasks, including image classification, audio pattern recognition, machine translation of text, and speech recognition. The data groups (angry, fearful, happy, neutral, sad, and surprised) tested with a threshold value of 30 epochs achieved a loss (error) accuracy of 1.5146 on the test data. The accuracy on the test data is 0.61. The proposed Convolutional Neural Network algorithm and digital image utilization achieved high accuracy performance to assist in evaluating a service-related field.
Model Machine Learning untuk Memprediksi Perilaku Konsumen sebagai Dasar Strategi Penargetan Ulang Iklan Antoni Antoni; Mbera Mehuli; Tri Andre Anu
Cosmic Jurnal Teknik Vol 2 No 4 (2025): November
Publisher : Ali Institute or Research and Publication

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

Abstract

Penelitian ini mengkaji bagaimana machine learning dapat digunakan untuk memprediksi perilaku konsumen dan menyediakan dasar berbasis data dalam penyusunan strategi advertising retargeting. Dalam lingkungan periklanan digital yang semakin kompetitif, praktik retargeting generik yang memperlakukan seluruh pengunjung sebagai satu kelompok audiens sering kali menyebabkan pemborosan anggaran, kelelahan iklan (ad fatigue), serta rendahnya relevansi pesan, karena niat konsumen bersifat dinamis dan bervariasi menurut waktu, perangkat, sumber trafik, dan tahapan funnel. Untuk mengatasi permasalahan tersebut, penelitian ini menerapkan desain pemodelan prediktif kuantitatif dengan memanfaatkan data clickstream pengguna dan data peristiwa (event) e-commerce pada tingkat individu. Fitur perilaku direkayasa untuk menangkap indikator seperti recency, frequency, intensitas eksplorasi, durasi sesi, serta sinyal funnel (misalnya add-to-cart), yang kemudian diikuti dengan proses pembersihan data, pengodean, penskalaan, pembagian data latih–uji berbasis waktu guna mengurangi kebocoran informasi, serta penanganan ketidakseimbangan kelas. Algoritma Logistic Regression digunakan sebagai model dasar yang dapat diinterpretasikan untuk mengestimasi probabilitas terjadinya keluaran target (misalnya konversi) dalam rentang waktu tertentu. Kinerja model dievaluasi menggunakan metrik yang sesuai untuk data tidak seimbang, termasuk ROC-AUC dan Precision–Recall (PR-AUC), serta nilai presisi, recall, dan F1-score pada ambang operasional. Hasil penelitian menunjukkan kemampuan diskriminatif yang sangat kuat (ROC-AUC = 0,961) dan efektivitas tinggi pada kelas positif (PR-AUC = 0,913), yang melampaui garis dasar prevalensi sebesar 0,235. Keluaran probabilitas dari model memungkinkan segmentasi audiens yang terukur ke dalam kelompok niat tinggi, sedang, dan rendah, sehingga mendukung penerapan intensitas retargeting dan strategi pesan yang berbeda. Secara keseluruhan, temuan ini menunjukkan bahwa penilaian probabilitas berbasis machine learning dapat meningkatkan presisi operasional retargeting dibandingkan pendekatan yang hanya berbasis intuisi; namun demikian, dampak bisnis terhadap CPA dan ROAS tetap perlu divalidasi melalui eksperimen lapangan seperti pengujian A/B.
Implementasi Algoritma Damerau-Levenshtein Untuk Pemeriksaan Dan Koreksi Kesalahan Ejaan Bahasia Indonesia Okvi Nugroho; Ahmad Rahmatika; Tri Andre Anu; Maulidya Rahmah
JURNAL ILMIAH SAINS TEKNOLOGI DAN INFORMASI Vol. 4 No. 2 (2026): Vol. 4 No. 2 (2026): April : Jurnal Ilmiah Sains Teknologi dan Informasi
Publisher : CV. ALIM'SPUBLISHING

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59024/jiti.v4i2.1943

Abstract

This study implements the Damerau-Levenshtein algorithm for an Indonesian spelling checking and correction system based on the distance editing approach. The main objective of this study is to develop a system capable of automatically detecting and correcting spelling errors at the character level through a matching process against the KBBI dictionary and the Indonesian corpus. The methods used include data collection, text pre-processing, system design, and implementation of the Damerau-Levenshtein algorithm which includes insertion, deletion, substitution, and transposition operations. Testing was conducted using 25 test data consisting of standard words and modified words for typographical errors. The results show that the system is able to measure all test data with an accuracy level of 100% on a limited dataset. In addition, the average Damerau-Levenshtein Distance value of 0.84 indicates that most errors are in the light category. Evaluation using a confusion matrix produces precision, recall, and F1-score values ​​of 100% each. These results indicate that the Damerau-Levenshtein algorithm is effective in handling character-based spelling errors. However, the system still has limitations in handling complex semantic contexts and language variations. Therefore, further research is recommended to integrate language model-based approaches to improve the system's accuracy and generalization on real-world data.