A series of machine learning methods for user rating prediction
Jingbo Wang · Applied and Computational Engineering · 2023
Users’ feedback online for services or products reflects their satisfaction. For example, higher rating implies better user satisfaction and experience. It is crucial to provide users satisfying services or products to ensure user experience and thus improve the long-term value of the platform. Manually reviewing all candidate services and products to select potentially user-preferred ones it time-costuming due to large amount of candidates. Therefore, this paper studies a series of machine learning based algorithms to automatically predict the user feedback for an online shopping website, i.e., user ratings in this scenario. To be specific, three machine learning models are investigated, including logistic regression, multi-layer perceptron, and convolution neural network. Starting with the pre-processing of raw data crawled from Amazon, an international online shopping website such three models are first trained over the training set and then evaluated on the testing set. The experiments demonstrate the convolutional neural networks give the highest accuracy for the unseen data.