A Novel Approach for E-Commerce System for Sale Prediction with De noised Auto Encoder and SVM Based Approach

Abstract

E-commerce,oronlineshopping,hasgrownin prominenceoverthepastfewyearsthankstotheproliferation oftheInternet.Yet,therearealotofthingsthatcanaffectan online store's success, and if the operators don't correctly assess their supply and marketing partnerships, they could lose a lot of money. Thus, it is crucial to create a model that can reliably produce high precision sales prediction in order toguaranteethelong-termsuccessofe-commercebusinesses. Thesuggestedmethodcomprisesthreestages:preprocessing, Feature selection, and model training. This work uses zero- phase component analysis and normalization in the preprocessing phase to get rid of noise and inconsistent data. Finally, the model is trained with DAE-SVM after information gain is employed for feature selection. When compared to convolutional neural network and support vector machine models, the proposed model excels.