Article Per Year (5 Year)
p-Index From 2021 - 2026
0.444
P-Index
This Author published in this journals
All Journal Journal of Educational Science and E-Learning JITCoS : Journal of Information Technology and Computer System
Hafiz Aryanda
Universitas Islam Negeri Sumatera Utara
Computer Science & IT Control & Systems Engineering Education Languange, Linguistic, Communication & Media Other

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Employee Performance Classification Using K-Nearest Neighbor and Random Forest with Work Behavior Scenario Simulation Hafiz Aryanda; Alwi Syahputra; Muhammad Farhan; Ade Aulia Dharma
Journal of Information Technology and Computer System Vol. 2 No. 1 (2026): June
Publisher : CV. Multimedia Teknologi Kreatif

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.65230/jitcos.v2i1.83

Abstract

Employee performance evaluation is a critical aspect of human resource management, directly influencing productivity, promotion decisions, and career development planning. Traditional approaches often fail to capture hidden patterns in complex employee data, making machine learning (ML) a promising solution for more accurate and objective classification. This study aims to model and compare two ML algorithms, K-Nearest Neighbor (KNN) and Random Forest, for employee performance classification, followed by scenario-based simulation using the best-performing model. The research employed a quantitative computational approach with a dataset of 100,000 employee records and 20 features. Preprocessing steps included feature selection, binarization of performance scores, handling class imbalance using Synthetic Minority Over-sampling Technique (SMOTE), and feature engineering to enrich data representation. The dataset was split into 80% training and 20% testing. KNN was first built as a baseline model, then optimized through hyperparameter tuning using GridSearchCV with cross-validation. Random Forest was implemented with 100 decision trees and bootstrap sampling to enhance accuracy and stability. Results show that the tuned KNN model achieved an accuracy of 70.88%, improving from 63.85% baseline. However, Random Forest outperformed KNN significantly, reaching 97.17% accuracy with lower error rates. Scenario simulations using Random Forest demonstrated practical applicability in predicting employee performance based on work behavior profiles. In conclusion, Random Forest provides a more robust and reliable model for employee performance classification compared to KNN. The study contributes by integrating algorithm comparison with simulation design, offering actionable insights for human resource decision-making. Future research may explore additional ensemble methods and real-time evaluation frameworks.
Analisis Performa CNN Berbasis MobileNetV2 pada Dataset Citra Dokumen Laporan Praktikum Hafiz Aryanda; Lailan Sofinah Harahap; Dimas Aqila Aptanta
Journal of Educational Science and E-Learning Vol. 2 No. 2 (2025): Desember
Publisher : CV Rena Cipta Mandiri

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.62354/jese.v2i2.44

Abstract

Penelitian ini bertujuan untuk menganalisis performa CNN berbasis MobileNetV2 dalam klasifikasi citra dokumen laporan praktikum. Pendekatan yang digunakan adalah transfer learning dengan bobot pra-latih dari ImageNet, serta fine-tuning pada lapisan atas untuk menyesuaikan model terhadap karakteristik dokumen akademik. Dataset terdiri dari ribuan citra dokumen yang telah melalui tahap pra-pemrosesan berupa resizing, normalisasi, dan augmentasi. Hasil eksperimen menunjukkan bahwa pada baseline, akurasi validasi mencapai 64,88% namun akurasi evaluasi manual hanya 10,54%, menandakan adanya masalah domain gap dan kesulitan separasi antar kelas. Fine-tuning meningkatkan akurasi validasi training menjadi 70,43%, tetapi akurasi evaluasi biner justru menurun menjadi 47,50%, sehingga performa model tetap rendah. Temuan ini menegaskan bahwa CNN kurang tepat untuk tugas validasi kepatuhan dokumen yang membutuhkan analisis semantik dan aturan tata letak yang kompleks serta dinamis. Sebagai solusi, integrasi OCR dengan rule-based validation direkomendasikan karena mampu memberikan akurasi lebih tinggi, interpretasi jelas, serta umpan balik detail sesuai pedoman penulisan akademik modern.
Co-Authors Ade Aulia Dharma Alwi Syahputra Dimas Aqila Aptanta Lailan Sofinah Harahap Muhammad Farhan