A Novel Machine Learning Approach to Predict Sales of an Item in E-commerce

Predicting the sales of the item through online comments given by the user is challenging research today. Consumer comments, user transactions, and item ratings are taken into account by the majority of existing prediction systems in order to determine which item is most likely to be predicted. But these comments may be changed over time since the interest to the item for the user may vary. Due to the fact that users' interests change over time, present sale prediction are ineffective in identifying items that are currently relevant to the consumer interest. This article presents a novel sales prediction technique that merges an incorporated selection approach for features and convolutional neural network classifier along with fuzzy logic. The empirical evaluation of the sales prediction system demonstrate that it outperforms the existing sales prediction system in the context of accuracy in predicting appropriate items for target consumers and in the time required to offer such predictions.