Klasifikasi Status Stok Barang Distribusi untuk Persediaan Menggunakan Algoritma Naive Bayes Berdasarkan Histori Data Penjualan PT. Berlian Djaya Nusantara
DOI:
https://doi.org/10.70292/pchukumsosial.v4i2.548Keywords:
Data Mining, Histori Penjualan, Klasifikasi, Naive Bayes, Status StokAbstract
Inventory management in distribution companies is exposed to excess-stock and shortage risks when replenishment decisions rely mainly on estimates. The sales history of PT Berlian Djaya Nusantara had not been systematically used to distinguish the stock condition of each item. The analysis classified stock status using the Naive Bayes algorithm and 250 sales records from 2025. The processing stages covered data selection, quality checking, transformation, target-class construction, modeling, and evaluation using a confusion matrix. Stock status was derived from Ending Stock through quartile-based discretization, producing Q1 of 138 and Q3 of 311. These thresholds generated 63 Low, 124 Medium, and 63 High records. Sales Quantity, Sales Frequency, and Initial Stock served as predictors, while Stock Status became the classification target. Five training-to-testing ratios were examined: 90:10, 80:20, 70:30, 60:40, and 50:50. The 90:10 ratio produced the strongest performance, with 70.83% accuracy, 70.27% precision, 72.22% recall, and a 71.23% F1-score. The model distinguished the three stock categories and generated measurable information for inventory control, although classification errors remained because the relationships among several attributes were not fully represented by the model.
