Commodity Width Measurement Based on Big Data

Abstract

Evidences about the impact of commodity width on brand performance remain fragmented. The traditional measurement of commodity category width is characterized by a quantity of manual work and repetition. A fusion model of convolutional neural network (CNN) and long short-term memory (LSTM) is proposed to solve the issue. In order to assure the universality and applicability of the findings, a vast consumer data set covering two retailers and two good categories is used for the measurement. The calculation results show that a composite model has the higher extraction accuracy than any single model on the average, and CNN or LSTM model alone will lead to the lower accuracy and higher error value. Convolutional neural network model possesses of powerful feature extraction, and the accuracy capacity which can be improved by CNN-LSTM fusion model. The mentioned fusion opens a new way for the measurement of commodity width.