Pemodelan Long Short Term Memory untuk Peramalan Penjualan pada Varian Brand PT. XYZ

Utari, Hanifah Fanidya (2025) Pemodelan Long Short Term Memory untuk Peramalan Penjualan pada Varian Brand PT. XYZ. Masters thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 6032222080-Master_Thesis.pdf] Text
6032222080-Master_Thesis.pdf
Restricted to Repository staff only until 1 April 2027.

Download (2MB) | Request a copy

Abstract

Kemajuan teknologi dan munculnya era Industri 4.0 memberikan dampak signifikan pada berbagai sektor, termasuk industri rokok. PT XYZ, salah satu perusahaan rokok terkemuka di Indonesia, menghadapi tantangan otomatisasi, persaingan teknologi, dan kebijakan cukai rokok yang terus meningkat. Tantangan berupa kenaikan tarif cukai rokok sebesar 15% pada 2022 menyebabkan penurunan pendapatan perusahaan, meskipun data menunjukkan peningkatan jumlah perokok. Berdasarkan data BPS, persentase penduduk Indonesia usia 15 tahun ke atas yang merokok meningkat menjadi 28,62% pada 2023. Untuk mengatasi tantangan ini, PT XYZ mengadopsi sistem berbasis Business Intelligence melalui proyek Marketing Agile Report Simplification System yang mengintegrasikan teknologi OLAP (Online Analytical Processing) dan OLTP (Online Transaction Processing) guna mendukung pengambilan keputusan berbasis data dengan descriptive analytic. Penelitian ini mengembangkan pendekatan tersebut menjadi predictive analytic dengan memanfaatkan metode Long Short-Term Memory (LSTM). Fokus penelitian diarahkan pada empat brand utama yaitu DU, LS, CL, dan CO yang, berdasarkan analisis Pareto, menyumbang 89,9% dari total pendapatan perusahaan. Penelitian menggunakan data historis penjualan harian periode 2022 hingga 2024 sebagai dasar peramalan. Hasil evaluasi menunjukkan MAPE sebesar 33,54% untuk Brand DU, 28,85% untuk Brand CL, 17,77% untuk Brand CO, dan 15,98% untuk Brand LS, dengan nilai R kuadrat lebih besar dari 0,79 pada setiap brand. Selain itu, nilai RMSE yang dihasilkan telah lebih rendah dari setengah standar deviasi data historis, menunjukkan model cukup andal dalam meramalkan penjualan. Hasil peramalan divisualisasikan dalam dashboard Power BI untuk memproyeksikan penjualan selama satu tahun ke depan, yaitu hingga 2025. Visualisasi ini memberikan wawasan bagi PT XYZ dalam memantau tren penjualan, mengidentifikasi risiko, dan menyusun strategi manajerial berbasis data. Pendekatan ini memungkinkan PT XYZ meningkatkan efektivitas pengambilan keputusan dan memperkuat daya saing perusahaan dalam menghadapi tantangan VUCA (Volatility, Uncertainty, Complexity, Ambiguity).
=================================================================================================================================
The rapid advancement of technology and the emergence of the Industry 4.0 era have significantly impacted various sectors, including the tobacco industry. PT XYZ, one of Indonesia's leading tobacco companies, faces challenges from automation, technological competition, and increasingly stringent tobacco excise policies. The 15% increase in tobacco excise taxes in 2022 led to a decline in the company’s revenue, despite data indicating an increase in the number of smokers. According to BPS data, the percentage of Indonesians aged 15 and above who smoke increased to 28.62% in 2023. To address these challenges, PT XYZ adopted a Business Intelligence-based system through the Marketing Agile Report Simplification System project, which integrates OLAP (Online Analytical Processing) technology and OLTP (Online Transaction Processing) to support data-driven decision-making with descriptive analytics. This study extended the approach to predictive analytics by leveraging the Long Short-Term Memory (LSTM) method. The research focused on four main brand DU, LS, CL, and CO which, based on Pareto analysis, contributed 89.9% of the company’s total revenue. The study utilized historical daily sales data from 2022 to 2024 as the basis for forecasting. Evaluation results showed MAPE values of 33.54% for Brand DU, 28.85% for Brand CL, 17.77% for Brand CO, and 15.98% for Brand LS, with R-squared values exceeding 0.79 for all brands. Additionally, the RMSE values were lower than half of the standard deviation of the historical data, demonstrating the model's reliability in sales forecasting. The forecasting results were visualized using Power BI dashboards to project sales for the next year, extending to 2025. This visualization provided PT XYZ with insights into sales trends, risk identification, and the formulation of data-driven managerial strategies. This approach enabled PT XYZ to enhance the effectiveness of decision-making and strengthen its competitive advantage in addressing the challenges of VUCA (Volatility, Uncertainty, Complexity, Ambiguity).

Item Type: Thesis (Masters)
Uncontrolled Keywords: Business Intelligence, Industri Rokok, LSTM, Peramalan Deret Waktu, Power BI
Subjects: T Technology > T Technology (General) > T174 Technological forecasting
T Technology > T Technology (General) > T385 Visualization--Technique
Divisions: Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User: Hanifah Fanidya Utari
Date Deposited: 31 Jan 2025 06:08
Last Modified: 31 Jan 2025 06:08
URI: http://repository.its.ac.id/id/eprint/117069

Actions (login required)

View Item View Item