Deep Learning Combinatorial Models for Intelligent Supply Chain Demand Forecasting

Ma, Xiaoya; Li, M. Z.; Jin, Tong; Feng, Xinbin

doi:10.3390/biomimetics8030312

Cited by 5 publications

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SARIMA and ANN models have been successfully used to fit and predict time series data in a variety of fields [ 18 – 22 ]. The SARIMA model can fit seasonal fluctuations well, but the fitting accuracy is poor for nonlinear components of TS data [ 23 ], while the LSTM model can compensate for this deficiency well, but another problem is that the mandatory fitting of seasonal fluctuations using a single LSTM model over a longer period increases the risk of overfitting, so a hybrid SARIMA-LSTM model was used to solve the accuracy problem of nonlinear fitting and simulate seasonal fluctuations at the same time [ 24 ]. In the construction of the LSTM model, we initially utilized 10 hidden units and performed iterations from 50 to 500.…”

Section: Discussionmentioning

confidence: 99%

Trend analysis and prediction of gonorrhea in mainland China based on a hybrid time series model

Wang,

Zhang

et al. 2024

BMC Infect Dis

View full text Add to dashboard Cite

Background Gonorrhea has long been a serious public health problem in mainland China that requires attention, modeling to describe and predict its prevalence patterns can help the government to develop more scientific interventions. Methods Time series (TS) data of the gonorrhea incidence in China from January 2004 to August 2022 were collected, with the incidence data from September 2021 to August 2022 as the validation. The seasonal autoregressive integrated moving average (SARIMA) model, long short-term memory network (LSTM) model, and hybrid SARIMA-LSTM model were used to simulate the data respectively, the model performance were evaluated by calculating the mean absolute percentage error (MAPE), root mean square error (RMSE), and mean absolute error (MAE) of the training and validation sets of the models. Results The Seasonal components after data decomposition showed an approximate bimodal distribution with a period of 12 months. The three models identified were SARIMA(1,1,1) (2,1,2)12, LSTM with 150 hidden units, and SARIMA-LSTM with 150 hidden units, the SARIMA-LSTM model fitted best in the training and validation sets, for the smallest MAPE, RMSE, and MPE. Conclusions The overall incidence trend of gonorrhea in mainland China has been on the decline since 2004, with some periods exhibiting an upward trend. The incidence of gonorrhea displays a seasonal distribution, typically peaking in July and December each year. The SARIMA model, LSTM model, and SARIMA-LSTM model can all fit the monthly incidence time series data of gonorrhea in mainland China. However, in terms of predictive performance, the SARIMA-LSTM model outperforms the SARIMA and LSTM models, with the LSTM model surpassing the SARIMA model. This suggests that the SARIMA-LSTM model can serve as a preferred tool for time series analysis, providing evidence for the government to predict trends in gonorrhea incidence. The model's predictions indicate that the incidence of gonorrhea in mainland China will remain at a high level in 2024, necessitating that policymakers implement public health measures in advance to prevent the spread of the disease.

show abstract

Section: Discussionmentioning

confidence: 99%

Trend analysis and prediction of gonorrhea in mainland China based on a hybrid time series model

Wang,

Zhang

et al. 2024

BMC Infect Dis

View full text Add to dashboard Cite

show abstract

“…However, these results must be taken very carefully since many proposed models are compared with models very similar to them, with only slight improvements, while other models are compared with diametrically different models with very different logics and methodologies, which make it more likely large performance differences. Finally, in the documents where the proposed machine learning models have been compared with classical statistical methods, the former have been clearly superior [16], [17], [21], [29], [23], [30], [44], [25], [24], [27], [43], [36], [41], [42]…”

Section: Q4: What Is the Performance Of The New Proposed Models In Re...mentioning

confidence: 96%

“…Complex and nonlinear data Fourteen of the 33 documents analyzed indicate that the main problem that the proposed "machine learning" models are intended to solve is the complexity and non-linearity of the patterns generated by the variables that affect the forecast. [12], [18], [20], [21], [22], [28], [23], [25], [26], [31], [43], [40], [41], [42].…”

Section: Keyword Inputmentioning

confidence: 99%

“…Twenty-three of the thirty-three documents analyzed propose some type of deep learning. Three of them propose the deep neural network as part of a base of predictors of different types [19][29] [35]]; fifteen of them propose it in combination with other simpler methods [14], [16], [22], [23], [25], [27], [43], [36], [40], [34] or in combination with other deep neural networks [28], [26], [39], [33] and, finally, five papers propose a single type of neural network as the main predictor [31], [38], [32], [37], [42]. Of note are the studies that propose the use of multiple neural networks of the same type as "stacking" [22], [23], [31], [42].…”

Section: Deep Learningmentioning

confidence: 99%

See 1 more Smart Citation

A Systematic Review of the Literature on the Use of Artificial Intelligence in Forecasting the Demand for Products and Services in Various Sectors

Villar,

Lengua

2024

IJACSA

View full text Add to dashboard Cite

This systematic review, carried out under the PRISMA methodology, aims to identify the recently proposed artificial intelligence models for demand forecasting, distinguishing the problems they try to overcome, recognizing the artificial intelligence methods used, detailing the performance metrics used, recognizing the performance achieved by these models and identifying what is new in them. Studies in the manufacturing, retail trade, tourism and electric energy sectors were considered in order to facilitate the transfer of knowledge from different sectors. 33 articles were analyzed, with the main results being that the proposed models are generally ensembles of various artificial intelligence methods; that the complexity of data and its scarcity are the main problems addressed; that combinations of simple machine learning, "bagging", "boosting" and deep neural networks, are the most used methods; that the performance of the proposed models surpasses the classic statistical methods and other reference models; and that, finally, the proposed novelties cover aspects such as the type of data used, the pattern extraction techniques used, the assembly forms of the applied models and the use of algorithms for automating the adjustment of the models. Finally, a forecast model is proposed that includes the most innovative aspects found in this research.

show abstract

“…The LSTM paradigm not only forecasts the directional momentum of future data points but also demonstrates a reduced goodness-of-fit metric on the test set relative to the ARIMA. Nonetheless, the LSTM's fit on the training dataset is somewhat inferior compared to the ARIMA model, potentially attributable to the LSTM's less robust encapsulation of the seasonal fluctuations inherent in the time series (41).…”

Section: Comparison Of the Simulation And Prediction Effects Of The A...mentioning

confidence: 99%

Comparison of simulation and predictive efficacy for hemorrhagic fever with renal syndrome incidence in mainland China based on five time series models

Wang,

Yang,

et al. 2024

Front. Public Health

View full text Add to dashboard Cite

BackgroundHemorrhagic fever with renal syndrome (HFRS) is a zoonotic infectious disease commonly found in Asia and Europe, characterized by fever, hemorrhage, shock, and renal failure. China is the most severely affected region, necessitating an analysis of the temporal incidence patterns in the country.MethodsWe employed Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Nonlinear AutoRegressive with eXogenous inputs (NARX), and a hybrid CNN-LSTM model to model and forecast time series data spanning from January 2009 to November 2023 in the mainland China. By comparing the simulated performance of these models on training and testing sets, we determined the most suitable model.ResultsOverall, the CNN-LSTM model demonstrated optimal fitting performance (with Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE) of 93.77/270.66, 7.59%/38.96%, and 64.37/189.73 for the training and testing sets, respectively, lower than those of individual CNN or LSTM models).ConclusionThe hybrid CNN-LSTM model seamlessly integrates CNN’s data feature extraction and LSTM’s recurrent prediction capabilities, rendering it theoretically applicable for simulating diverse distributed time series data. We recommend that the CNN-LSTM model be considered as a valuable time series analysis tool for disease prediction by policy-makers.

show abstract

Deep Learning Combinatorial Models for Intelligent Supply Chain Demand Forecasting

Cited by 5 publications

References 21 publications

Trend analysis and prediction of gonorrhea in mainland China based on a hybrid time series model

Trend analysis and prediction of gonorrhea in mainland China based on a hybrid time series model

A Systematic Review of the Literature on the Use of Artificial Intelligence in Forecasting the Demand for Products and Services in Various Sectors

Comparison of simulation and predictive efficacy for hemorrhagic fever with renal syndrome incidence in mainland China based on five time series models

Contact Info

Product

Resources

About