Table 1.
Parameter employed for different models.
Fig 1.
Monthly sales trends for five products over 15 months.
The graph displays actual sales amounts (in Bangladeshi Taka) for five products over 15 months, from June 2022 to August 2023.
Table 2.
Statistical Parameters of all products during 15-month period (In Taka).
Fig 2.
Correlation matrix of total sales and key features.
The heatmap shows correlation coefficients between the sum of sales for five products and features like Discounts, CPI, Inflation Rate, Average Temperature, Rainfall, National Holidays, and Festive Months.
Fig 3.
Predicted sales and uncertainty bands for the XGBOOST model.
Fig 4.
Predicted sales and uncertainty bands for the random forest model.
Fig 5.
Multilayer Perceptron Model Results.
Predicted sales and uncertainty bands for the Multilayer Perceptron (MLP) model.
Fig 6.
Average Mean Absolute Percentage Error (MAPE) of different forecasting methods.
The bar chart compares the MAPE values of forecasting methods. Multilayer Perceptron achieves the lowest MAPE (46.00), followed by Random Forest (52.43), and XGBOOST (56.18).
Table 3.
Root Mean Squared Error of Forecasting Models by Product.
Table 4.
Performance scores of Forecasting Models.
Fig 7.
Actual Sales vs. Predicted Sales using five forecasting models.
The figure presents a comparison of actual sales and forecasted sales for five products over 15 months using XGBOOST, Random Forest, Multilayer Perceptron, Simple Exponential Smoothing, and Croston Method. This figure is a multi-graph comparison, where (a) represents Noodles, followed by (b) Chicken, (c) Hot Coffee, (d) Rice, and (e) Soup.