IMPLEMENTASI METODE XGBOOST DALAM PREDIKSI PENJUALAN PRODUK PAWONKOE DAN ANALISIS ULASAN PELANGGAN MENGGUNAKAN METODE MBERT

afandy, selena nurmanina (2026) IMPLEMENTASI METODE XGBOOST DALAM PREDIKSI PENJUALAN PRODUK PAWONKOE DAN ANALISIS ULASAN PELANGGAN MENGGUNAKAN METODE MBERT. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) cover.pdf Download (780kB)
	Text (Bab 1) Bab 1.pdf Download (217kB)
	Text (Bab 2) bab 2.pdf Restricted to Repository staff only until 9 March 2029. Download (536kB)
	Text (Bab 3) bab 3.pdf Restricted to Repository staff only until 9 March 2029. Download (581kB)
	Text (Bab 4) bab 4.pdf Restricted to Repository staff only until 9 March 2029. Download (5MB)
	Text (Bab 5) bab 5.pdf Download (202kB)
	Text (Daftar pustaka) daftar pustaka.pdf Download (176kB)
	Text (Lampiran) lampiran.pdf Restricted to Repository staff only until 9 March 2029. Download (180kB)

Abstract

Daily sales fluctuations and dynamic customer perceptions became major challenges in managing Pawonkoe MSME in Banyuwangi during the January–December 2024 period. Significant sales variations may lead to inefficiencies in production planning if not supported by an accurate Forecasting system. In addition, 365 customer reviews were available but had not been optimally utilized as a basis for strategic evaluation. This study employed 366 daily sales records and 365 customer reviews using an integrated machine learning approach. The sales Forecasting model was developed using the XGBOOST algorithm with Lag features, a seven-day rolling mean, and day-of-week variables, with an 80% Training and 20% testing data split. The evaluation results showed a Mean Absolute Percentage Error (MAPE) of 20.44% in the 1-step supervised scheme and an average MAPE of 19.66% in Recursive 7-step Forecasting as the best numerical result. However, in the rolling autoregressive scheme, which represents real-world Forecasting conditions, a MAPE of 46.27% was obtained, indicating increased error due to accumulated sequential predictions. Sentiment analysis using Multilingual BERT (MBERT) achieved an accuracy of 0.91, with a distribution of 245 Positive sentiments (67.1%), 117 Neutral (32.1%), and 3 Negative (0.8%). These results indicate that customer perception is predominantly Positive. This study integrates sales Forecasting and sentiment analysis into a web-based system using Streamlit to support data-driven decision-making for Pawonkoe MSME.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Hindrayani, Kartika Maulida	NIDN0009099205	kartika.maulida.ds@upnjatim.ac.id
Thesis advisor	Damaliana, Aviolla Terza	NIDN0002089402	aviolla.terza.sada@upnjatim.ac.id

Subjects:

Q Science > Q Science (General)
Q Science > QA Mathematics
Q Science > QA Mathematics > QA76.6 Computer Programming

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

selena nurmanina afandy

Date Deposited:

10 Mar 2026 01:25

Last Modified:

10 Mar 2026 02:24

URI:

https://repository.upnjatim.ac.id/id/eprint/50296