|
| Method | Paper | Domain | Findings |
|
| Fuzzy | Holt Winter | [33] | New items | The proposed fuzzy-adaptive model controls the weight factors of an exponential-smoothing forecasting method, and it can be applied to new item sales forecasting. |
|
CCX | [34] | Mean term | It uses fuzzy logic abilities to map the nonlinear influences of explanatory variables to conduct forecasting, but expert judgment is required for the learning process. |
| [35] | Mean term | It allows a textile company to obtain mean term forecasting to pass commands to providers. |
| NN | [36] | Short term | The method performs short-term forecasting by readjusting mean-term model forecasts from load real sales. |
| Distribution of aggregated forecast and classification | [37] | New items
Insufficient data | The method (items forecasting model based on distribution of aggregated forecast and classification) estimates the item sales of the same family without requiring historical data. |
| NN | [38] | Short term | The model promotes greatly the accuracy of forecasting results for the horizon of one month. |
| NN | [39] | Fast forecasting | Using fuzzy logic, the combiner calculates final forecast for each week’s demand as a weighted average of forecasts that are generated by different methods. The combined forecast achieves better accuracy than any of the individual forecasts. |
|
| ANN | CCX | [40] | Mean term | Considering noisy data and multiple explanatory variables (controlled, available or not) related to the sales behavior, the proposed model performs well. |
| Classification | [41] | New items | Neural clustering and classification model globally increases the accuracy of midterm forecasting in comparison with the mean sales profile predictor. |
ELM +
Harmony search | [42] | Mean term | The learning algorithm integrates an improved harmony search algorithm and an extreme learning machine to improve the network generalization performance and is better than traditional ARIMA models and two recently developed neural network models in fashion forecasting. |
| ART | [43] | Two stages:
long term and short term | Combining the ART model the and error forecasting model based on neural network, an adjustment improving model which can be applied to the fashion retail forecasting is developed. |
| GM | [44] | Color trend
Insufficient data | GM+ANN hybrid models are examined in the domain of color trend forecasting with a limited amount of historical data. The GM+ANN hybrid model is the best one for forecasting fashion sales by colors where only very few historical data is available. |
|
| ELM | Statistic | [45] | Fast forecasting | A comparison with other traditional methods has shown that the ELM fast forecasting model is quick and effective. |
| Metrics | [46] | Sufficient data | The adaptive metrics of inputs can solve the problems of amplitude changing and trend determination and reduce the effect of the overfitting of the neural networks. The model outperforms autoregression (AR), ANN, and ELM models. |
| GRA | [47] | Color trend | With real data analysis, the results show that the ANN family models, especially for ELM with GRA, outperform the other models for forecasting fashion color trend. |
|
| SARIMA | Wavelet | [48] | Highly
volatile sales | For real data with relatively weak seasonality and highly variable seasonality factor, the SW hybrid model performs well. |
|
| Decision tree | Clustering | [49] | Mean term | The proposed model based on existing clustering technique and decision tree classifier is useful to estimate sales profiles of new items in the absence of historical sales data. |
| Auto-regressive technique | [43] | Short term | A two-stage dynamic short-term forecasting model is proposed |
|
