With the rapid development of e-commerce, merchants put more categories on the shelves. Understanding consumer preferences and prompting consumers to quickly find their favorite products in many categories requires recommendation algorithms. Wu (2019) analyzed the characteristics of various data in different information sources, proposed a novel recommendation model, which can alleviate the sparsity problem by seamlessly integrating multi-relational data and visual content, and designed a computationally efficient learning algorithm MSRA to optimize the proposed model. The prediction method was introduced into a new vector to represent disease and applied the new vectorized data to a positive unlabeled learning algorithm to predict and rank long non-coding RNA (lncRNA) genes associated with disease (Peng et al., 2017). Bai et al. (2017) studied two main text representations for predicting cross-site purchase preferences, including shallow and deep text features learned by deep neural network models. Using extensive experiments on a large, linked dataset, they provide experimental results showing that leveraging social text to predict purchase preferences is promising. In listening experiments, fan noise signals were adjusted to the same loudness and the same preference compared to the normal reference sound by varying their levels in an adaptive program to quantify how changes in these two indices affect subjects’ preference and loudness judgments (Töpken & van de Par, 2019).