6th North American International Conference on Industrial Engineering and Operations Management

A Hybrid Artificial Neural Network and Logistics Regression Model for Fashion Sales Strategies Prediction

Rifdah Zahabiyah & Andi Cakravastia
Publisher: IEOM Society International
0 Paper Citations
Track: Graduate Student Paper Competition

Sales prediction for fashion product is known to be very difficult, as we know the fashion demand is highly volatile with ever-changing customer's tastes and the product life cycle tends to be short. The purpose of this study is to predict the right strategy, between pre-order or vice versa using the deep learning function of Artificial Intelligence (AI). This research was conducted by combining AI using Artificial Neural Networks (ANN) with statistical methods using Logistics Regression (LR). The LR model was adopted as a pre-selection of input factors for the ANN prediction model (ANN-Logit). We also calculated the probability of each case using the logistic function then added as a new input variable for the ANN model (ANN-Plogit). The comparison was made to compare each model’s performance. There is no significant difference between the outputs of each model. All models are considered good in predictions with fitness values above 0.85 and MSE tends to be small, but the hybrid ANN-Plogit provides slightly better than others with the highest fitness value of 0.995 and the lowest error (MSE) for 0.0012 with 98.6% prediction accuracy. This shows that the more input variables used as predictors, the better the fitness model. Managerial insight as the results of prediction for upcoming product is presented in this study.

Published in: 6th North American International Conference on Industrial Engineering and Operations Management, Monterrey, Mexico

Publisher: IEOM Society International
Date of Conference: November 3-5, 2021

ISBN: 978-1-7923-6130-2
ISSN/E-ISSN: 2169-8767