OPTIMASI PREDIKSI PENJUALAN PRODUK FASHION DI TIKTOKSHOP MENGGUNAKAN XGBOOST DAN BAYESIAN OPTIMIZATION

Ali Sofi, Firhan (2025) OPTIMASI PREDIKSI PENJUALAN PRODUK FASHION DI TIKTOKSHOP MENGGUNAKAN XGBOOST DAN BAYESIAN OPTIMIZATION. Undergraduate thesis, Universitas Pembangunan Nasional "Veteran" Jawa Timur.

	Text (Cover) Cover.pdf Download (1MB)
	Text (BAB I) BAB 1.pdf Download (338kB)
	Text (BAB II) BAB 2.pdf Restricted to Repository staff only until 27 May 2028. Download (447kB)
	Text (BAB III) BAB 3.pdf Restricted to Repository staff only until 27 May 2028. Download (1MB)
	Text (BAB IV) BAB 4.pdf Restricted to Repository staff only until 27 May 2028. Download (1MB)
	Text (BAB V) BAB 5.pdf Download (192kB)
	Text (Daftar Pustaka) Daftar Pustaka.pdf Download (185kB)

Abstract

Sales prediction is a crucial aspect of inventory management and business strategy planning. This study aims to develop a sales prediction model for fashion products using the XGBoost algorithm optimized with Bayesian Optimization. The data used in this study is sales transaction data from the fashion business D'Busana for the period of 2022 to 2025, with a focus on the e-commerce platform TikTok Shop. The model is developed to predict sales quantity based on factors such as product lifecycle, seasonal patterns, and stock availability. The data processing phase includes preprocessing, time-series feature creation, as well as experimentation with feature engineering techniques to improve model performance. The data is split with an 80:10:10 ratio for training, validation, and testing. The best configuration is obtained by using Bayesian Optimization with a range of n_estimators from 50 to 500 and 5-fold cross-validation. The hyperparameters optimized in this study include max_depth, learning_rate, n_estimators, colsample_bytree, and subsample, all of which were explored within identical value ranges. The experimental results show that the XGBoost model optimized using Bayesian Optimization delivers the best performance compared to the baseline model, achieving an MSE of 0.0001, RMSE of 0.0105, MAE of 0.0013, and R² of 0.9591. In this study, R² is used as the primary metric to assess whether the model's performance has improved. The model evaluation indicates that it can predict sales with high accuracy, although challenges remain in accounting for seasonal fluctuations and stock availability. The results of this study indicate that optimization using Bayesian Optimization is effective in improving the sales prediction model's performance and can be implemented to support more efficient stock planning and sales strategies on e-commerce platforms.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Diyasa, I Gede Susrama Mas	NIDN0710117803	igsusrama.if@upnjatim.ac.id
Thesis advisor	Junaidi, Achmad	NIDN0019067008	achmadjunaidi.if@upnjatim.ac.id

Subjects:

Z Bibliography. Library Science. Information Resources > Z665 Library Science. Information Science

Depositing User:

Firhan Firhan Ali Sofi

Date Deposited:

27 May 2025 08:45

Last Modified:

27 May 2025 08:45

URI:

https://repository.upnjatim.ac.id/id/eprint/36792