New machine learning model based on the time factor for e-commerce recommendation systems

Nowadays, thanks to the development of e-commerce websites, businesses can capitalize on many benefits, for example, there are many methods of approaching customers online. Customers can interact with the product on the system, leave comments or reviews about the product, and capitalize on these interactions helps a lot to reach target customers. Many authors have studied machine learning models to build recommendation systems. The common point of the recommendation system is to filter out the products that are most relevant to the customer in order to retain them longer and to improve the customer’s product experience. In this paper, a new recommendation model called ML.Recommend combined with Microsoft’s ML.NET machine learning platform is proposed. This model provides a full cycle of recommendation modeling, including the steps of preprocessing, model training, model evaluation, model saving and usage. ML.Recommend uses the matrix factor and time factor combination for product recommendations based on ratings and logistic regression for customer comments about products. In this model, we provide a set of interactive class libraries, data, and class models that are implemented based on user evaluations of each interactive product over time. The model recommends corresponding products based on the expected score for the customer that a user has configured. The data are experimented on the e-commerce website called UEL Store and the UCI sentiment labeled sentences dataset. Measurement parameters such as mean absolute error, mean square error, root-mean-square error, R -squared and area under the curve are applied to evaluate the model. Finally, the ML.Recommend model was published on Microsoft’s NuGet system, so that other researchers could use and extend this model.