An Effective Predicting E-Commerce Sales & Management System Based on Machine Learning Methods
Table 1.
Summary of literature review one-commerce-based sales prediction using machine learning
|
| Author |
Methodology |
Data |
Findings |
Limitations |
|
|
| Niu (2020) |
XGBoost with feature engineering for sales prediction |
Walmart sales dataset from Kaggle |
XGBoost outperformed Logistic Regression and Ridge Regression; RMSSE was 0.141 and 0.113 lower than these models, respectively. Feature importance was ranked. |
Limited comparison to other models; only Walmart sales data used. |
|
| Wisesa, Adriansyah & Khalaf (2020) |
Gradient Boosting for B2B Sales Prediction |
B2B sales data |
Gradient Boosting had good accuracy with MSE = 24,743,000,000 and MAPE = 0.18. Generated reliable, accurate sales forecast data. |
Lack of detail on specific data attributes; limited evaluation beyond MSE and MAPE. |
|
| Ding et al. (2020) |
CatBoost with feature engineering for sales prediction |
Walmart sales dataset |
CatBoost outperformed traditional models such as SVM and linear regression with an RMSE of 0.605. Required less fine-tuning and improved generalisation. |
Only Walmart sales data used; limited comparison to other ensemble models. |
|
| Kulshrestha & Saini (2020) |
Sales prediction based on ML, split data by quarters and predicted future income. |
E-commerce sales data |
Predicted future income and analyzed frequently sold commodities. Provided strategic insights for inventory and business planning. |
Did not specify the exact ML algorithm used; limited comparison to other models. |
|
| Kaneko & Yada (2016) |
Deep learning model with L1 regularization for sales prediction |
3 years of POS data from a retail store |
Achieved 86% accuracy for sales prediction. Deep learning model outperformed Logistic Regression, with accuracy falling by 7% versus 13% for Logistic Regression. |
No mention of computational complexity or scalability for larger datasets. Only tested on retail POS data. |
|
|
|
