A Dialectical Analysis of Prediction Models for Food Demand Forecasting
Aftab Khan, Department of Industrial Engineering, University of Engineering and Technology, Peshawar, Pakistan.
Corresponding Author:
Aftab Khan (aftabkhan_ind@uetpeshawar.edu.pk)
Abstract:
Accurately gauging the demand volume for the upcoming unit period is pivotal and beneficial for businesses. It can enable the companies to lessen their inventory cost as well as guarantee product delivery. This research investigates the efficacy of multiple time series prediction techniques in the forecasting of food demand. An in-depth study of the current literature led to the selection of the techniques used in this research. These techniques included the Gated Recurrent Unit (GRU) based Recurrent Neural Network (RNN), Long-Short Term Memory (LSTM) based RNN, Auto-Regressive Integrated Moving Average (ARIMA) and the State Space Model (SSM). A food orders dataset was utilized and the weekly usage for various entities was input into the system during the training stage. A testing portion of the weekly consumption was used to estimate the quality of the prediction of these models. Errors measures including Root Mean Square Error (RMSE) and Pearson’s correlation coefficient were used to validate the performance of these models. A better fit of the models was observed for the LSTM-RNN technique in terms of the RMSE. A possible endeavour could be the utilization of heuristic search algorithms in speeding up the computational processes. The work can also be furthered by investigating and possibly estimating the various characteristics of the time series independently before forecasting.
Keywords:
Recurrent Neural Network (RNN); Deep Learning (DL); ARIMA; State Space Model (SSM); Prediction; Pearson’s Correlation Coefficient