Year: 2024
Type: article
Abstract: In this study, we predict inventory for an IoT-enabled vending machine warehouse servicing approximately 1,500 vending machines with the goal of timely replenishing, achieving cost effectiveness, reducing stock waste, optimising the available resources and ensuring fulfilment of consumer demand. The study deploys four different ML algorithms, namely, Extreme gradient boosting, Autoregressive integrated moving average with/without exogenous variables (ARIMA/ARIMAX), Facebook Prophet (Fb Prophet), and Support Vector Regression (SVR). The study unfolds in two phases. First, we utilise conventional historical sales data variables to make the prediction whereas in the second phase, we systematically introduced external variables including weekday, sales deviation flag, and holiday flags into our ML algorithms. The results indicate a significant performance boost using external variables with extreme gradient boosting achieving the lowest (Mean Absolute Error) MAE of 22, followed by ARIMAX, FB Prophet, and SVR with MAE values of 27, 37, and 38, respectively.
Source: IEEE Internet of Things Magazine
Authors: Umair Mehmood, John Broderick, Simon Davies, Ali Kashif Bashir, Khaled Rabie
Institution: Manchester Metropolitan University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $2,995
Funder: Innovate UK
Award: G5232767972
OpenAlex ID: W4402769072