| dc.description.abstract | Demand forecasting has played a crucial role in allowing organizations to
efficiently meet future customer demands. The main focus of current research is to create
models that minimize the margin of error in forecasting. However, it is evident that a
significant proportion of Colombian companies predominantly depend on conventional
approaches to forecasting their requirements, thereby overlooking the potential benefits of
employing machine learning and deep learning models. This study presents a comparative
evaluation of the efficacy of demand projection models for a particular category of fast moving consumer goods (FMCG) provided by Cerescos, Google Colab, and Excel. The
projections for the subsequent months of 2023 and January 2024 were derived from the
demand dataset pertaining to the company's products as well as a collection of
macroeconomic indicators for Vietnam spanning from January to September. The selection
of forecasting models was based on a thorough examination of comparable studies and an
analysis of the demand patterns and time series data of FMCG. This study assesses the
performance of multiple linear regression, SARIMA-MLR, and recurrent neural networks
(RNN) models on the study goods. The evaluation is conducted using real data from the
remaining months of 2023 and January 2024. The measurements employed for forecast
error are MSE, MAD, CFE, TS, and MAPE. The results indicate that the RNN models with
double LSTM layers demonstrate superior forecasting performance for the 250 examined
goods compared to the other models. In addition, it is recommended to utilize a model that
incorporates time-series learning and multivariate learning techniques in order to
effectively manage inventory. The present section concludes by conducting an analysis of
the models' performance and their potential utility in predicting the demand for goods
within Circle-K companies. | en_US |
