ANALISIS PREDIKSI PENJUALAN TOKO FURNITUR DENGAN METODE LONG SHORT-TERM MEMORY (LSTM)

Abstract

This study aims to analyze and predict furniture store sales using the Long Short-Term Memory (LSTM) method, focusing on time series datasets from 2014 to 2017. The LSTM method was chosen because of its ability to handle remote dependencies in time series data, which is relevant in understanding furniture sales patterns and trends for strategic planning. The research stages include literature study, library research, field research, data acquisition, and data preprocessing using Python and Google Colab. Exploratory analysis of the data was carried out to understand the characteristics of the dataset, followed by the development of the LSTM model, data normalization, and model evaluation with RMSE, MAE, and MAPE metrics. The evaluation results show that the LSTM model produces RMSE of 39.27%, MAE of 32.74%, and MAPE of 42.28%. Nonetheless, there is potential to improve accuracy by integrating more variables, exploring different LSTM architectures, and utilizing regularization techniques. This research is expected to contribute to improving the effectiveness of furniture sales management strategies and inspire further development in the application of ESG in business prediction.