Using predictive analytics to solve a newsvendor problem / S. Sarifah Radiah Shariff and Hady Hud
Purpose Forecasting accuracy is an ongoing challenge for many companies. The pandemic has resulted in unpredictable customers’ purchasing patterns post-pandemic has rendered heuristic-based forecasting large forecast errors, which leads to poor decision-making. It is debatable whether such behavio...
保存先:
| 主要な著者: | , |
|---|---|
| フォーマット: | Book Section |
| 言語: | English |
| 出版事項: |
Malaysia Institute of Transport (MITRANS), UiTM
2023
|
| 主題: | |
| オンライン･アクセス: | https://ir.uitm.edu.my/id/eprint/101938/1/101938.pdf https://ir.uitm.edu.my/id/eprint/101938/ |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
| 要約: | Purpose
Forecasting accuracy is an ongoing challenge for many companies. The pandemic has resulted in unpredictable customers’ purchasing patterns post-pandemic has rendered heuristic-based forecasting large forecast errors, which leads to poor decision-making. It is debatable whether such behaviour may persist long after the pandemic ends or settle down at a new normal level.
The objective of this research is to forecast the daily demand of a perishable product during the new norm post-pandemic, ensuring minimal unsold items are discarded as waste.
Findings
This research makes two key observations. Firstly, an algorithm with good performance metrics over a small data set collection may obtain worse results when the data set collection is extended. The best algorithm will not be the same for all the data sets. Secondly, in solving every Machine Learning problem, there is no one algorithm superior to other algorithms. Every algorithm makes its own respective prior assumptions about the relationships between the features and target variables, which create different types and levels of bias. The assumptions adopted in the Decision Tree Model and K-Nearest Neighbour are derived from symbolic artificial intelligence and data mining, whilst the assumptions in the Artificial Neural Network are derived from the connectionist approach.
Practical implications
In terms of managerial implications, the findings in this research help to frame the adoption of a more advanced analytical approach to forecasting, using a Machine Learning algorithm, in solving a newsvendor problem. The unpredictability of customers’ purchasing patterns postpandemic has rendered heuristic-based forecasting large forecast errors. This research attempts to solve the problem statement on how to forecast the daily optimal quantity of a perishable product during the new norm post-pandemic, ensuring minimal unsold items are discarded as waste.
Originality/value
Overall, this research provides initial insights into adopting Machine Learning algorithms in making better-informed managerial decisions among SMEs in Malaysia |
|---|