Authors :
Swarna Chaithanya Kollipara; Satya Krishna M B; Sai Vishal Golem; Purvaja Fursule; Bharani Kumar Depuru
Volume/Issue :
Volume 9 - 2024, Issue 1 - January
Google Scholar :
http://tinyurl.com/mrxxj4sf
Scribd :
http://tinyurl.com/3rpdr6kx
DOI :
https://doi.org/10.5281/zenodo.10635066
Abstract :
Wooden pallet manufacturers contend with
erratic demand patterns, impeding optimal resource
allocation and operational performance. In the dynamic
industry of wooden pallet manufacturing, the imperative
for precise demand forecasting arises from this inherent
variability in customer demand, demanding accuracy in
inventory management, warehouse capacity utilization,
and production planning. This study harnesses deep
learning models for enhancing demand forecasting in the
wooden pallet manufacturing industry because
conventional forecasting methodologies encounter
difficulties adapting to these dynamic conditions,
resulting in inaccuracies and consequential inventory
mismanagement, which incur substantial costs.
A comprehensive evaluation of 14 deep learning
models, including Autoformer, Informer, Neural Basis
Expansion Analysis for Interpretable Time Series
Forecasting (N-BEATS), Neural Basis Expansion
Analysis for Interpretable Time Series Forecasting with
Exogenous Variables (N-BEATSx), Neural Hierarchical
Interpolation for Time Series Forecasting (N-HiTS),
PatchTST, Prophet, Temporal Convolutional Network
(TCN), Temporal Fusion Transformer (TFT), TimeGPT,
TimesNet, TSmixer, and the AutoTS library, culminated
in the identification of AutoTS as the most effective for
consistently accurate predictions. AutoTS library
autonomously analyzes customer data, tests
approximately 800 models for each customer, and
adeptly selects the most suitable model from its
expansive library, ensuring optimal forecasting accuracy
tailored to each unique customer. The amalgamation of
multiple models through AutoTS mitigates risks
associated with reliance on a singular algorithm,
contributing to producing more robust and reliable
forecasts. Rigorous testing on historical data from 3,710
unique customers across India revealed AutoTS's
capability to generate precise weekly, bi-weekly, and
monthly forecasts, surpassing an accuracy benchmark of
80%.
Integrating an interactive dashboard in the study
facilitates real-time data analysis and visualization,
fostering informed decision-making in critical
operational domains of our client. By delivering highly
accurate demand forecasts, this approach empowers
wooden pallet manufacturers to efficiently manage
inventory, optimize production schedules, and ultimately
enhance operational efficiency and profitability.
Keywords :
Deep Learning Models, AutoTS Library, Machine Learning, Predictive Modelling, Demand Forecasting, Supply Chain Optimization, Inventory Management.
Wooden pallet manufacturers contend with
erratic demand patterns, impeding optimal resource
allocation and operational performance. In the dynamic
industry of wooden pallet manufacturing, the imperative
for precise demand forecasting arises from this inherent
variability in customer demand, demanding accuracy in
inventory management, warehouse capacity utilization,
and production planning. This study harnesses deep
learning models for enhancing demand forecasting in the
wooden pallet manufacturing industry because
conventional forecasting methodologies encounter
difficulties adapting to these dynamic conditions,
resulting in inaccuracies and consequential inventory
mismanagement, which incur substantial costs.
A comprehensive evaluation of 14 deep learning
models, including Autoformer, Informer, Neural Basis
Expansion Analysis for Interpretable Time Series
Forecasting (N-BEATS), Neural Basis Expansion
Analysis for Interpretable Time Series Forecasting with
Exogenous Variables (N-BEATSx), Neural Hierarchical
Interpolation for Time Series Forecasting (N-HiTS),
PatchTST, Prophet, Temporal Convolutional Network
(TCN), Temporal Fusion Transformer (TFT), TimeGPT,
TimesNet, TSmixer, and the AutoTS library, culminated
in the identification of AutoTS as the most effective for
consistently accurate predictions. AutoTS library
autonomously analyzes customer data, tests
approximately 800 models for each customer, and
adeptly selects the most suitable model from its
expansive library, ensuring optimal forecasting accuracy
tailored to each unique customer. The amalgamation of
multiple models through AutoTS mitigates risks
associated with reliance on a singular algorithm,
contributing to producing more robust and reliable
forecasts. Rigorous testing on historical data from 3,710
unique customers across India revealed AutoTS's
capability to generate precise weekly, bi-weekly, and
monthly forecasts, surpassing an accuracy benchmark of
80%.
Integrating an interactive dashboard in the study
facilitates real-time data analysis and visualization,
fostering informed decision-making in critical
operational domains of our client. By delivering highly
accurate demand forecasts, this approach empowers
wooden pallet manufacturers to efficiently manage
inventory, optimize production schedules, and ultimately
enhance operational efficiency and profitability.
Keywords :
Deep Learning Models, AutoTS Library, Machine Learning, Predictive Modelling, Demand Forecasting, Supply Chain Optimization, Inventory Management.
