Predictive analytics for fast moving item using nonlinear regresssion models
A supply chain and logistics company are trying to realign their stock placement in the warehouse based on the type of movement of the stock keeping units (SKU), fast moving or slow moving. This project is executed to construct a nonlinear regression models for order frequency per month of fast movi...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 語言: | English |
| 出版: |
2021
|
| 主題: | |
| 在線閱讀: | http://eprints.utm.my/id/eprint/96404/1/NurArishaMSC2021.pdf.pdf |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | A supply chain and logistics company are trying to realign their stock placement in the warehouse based on the type of movement of the stock keeping units (SKU), fast moving or slow moving. This project is executed to construct a nonlinear regression models for order frequency per month of fast moving items using Python programming language. The variables used for this prediction model is the median order frequency per month for each warehouse, total quantity of item, total volume of item and total value of item. The project framework has been set up with the inclusion of data visualization for the type of movement of each SKU for each warehouse using Tableau software. SKU are segmented by comparing the average frequency of order for each SKU in the span of 33 months with the median frequency of order for each respective warehouse the SKU resides in. Three nonlinear regression based models are used to construct the predictive model which are Decision Tree Regression, Random Forest Regression and Extreme Gradient Boosting Algorithms. Parameters tuning for the model carried out by using RandomizedSearchCV from scikit-learn library. Random forest produce the smallest error rate for prediction by using mean square error with an average value of 1.2608 and mean absolute error with an average value of 0.4496 as model evaluation and holdout method as model validation in this study. |
|---|