In the current era of rapid Internet and artificial intelligence development, the explosion of information requires effective filtering to match user interests. Accordingly, this paper proposes a personalized recommendation algorithm based on sentiment analysis and matrix factorization (PRSAMF). A sentiment analysis model is constructed utilizing a Long Short-Term Memory (LSTM) network for deep learning, with ongoing parameter adjustments for training and validation. Through this approach, the LSTM network effectively captures the emotional polarity in user reviews. This emotional polarity, combined with the user rating matrix, enhances the accuracy of representing user reviews. Subsequently, the user sentiment score matrix and the high-frequency matrix of user search items undergo factorization to uncover potential user preferences and item attributes. Recommendations are then made based on the users sentiment towards these attributes. Experimental results on the dataset demonstrate the models effectiveness, showing optimal accuracy and low loss rates in sentiment analysis. Additionally, the error rate remains within acceptable limits, indicating the feasibility and robustness of the proposed recommendation algorithm.

Genetic information becomes more and more important in the process of biological research. Gene analysis is an effective mean in biological research, especially the analysis of differentially expressed genes. Robust principal component analysis (RPCA) is an effective method to identify differentially expressed genes. But tensor robust principal component analysis (TRPCA) performs better than RPCA when processing multi-dimensional data. The traditional TRPCA method also has limitations in restoring low-rank sparse components. To further improve the accuracy of the TRPCA method in restoring low-rank components and sparse components, we propose a novel TRPCA method to obtain high-order correlations information of multi-dimensional data. It uses a new nuclear norm based on t-product operator to approximate the rank function. The L2,1-norm is used to improve the sparsity of tensors and reduce the negative effects caused by noises and outliers. At the same time, the introduction of L2,1-norm enhances the sparsity of error components, and improves the accuracy of low-rank component recovery. The low-rank sparse components are obtained by solving the convex problem of the new tensor nuclear norm. It can well preserve the spatial structure and make full use of complementary information to improve the clustering effect. Alternating direction method of multiplier (ADMM) is used to solve the optimization problem of this method. Experimental results on different cancer genomic datasets indicate that our method is superior to other methods.

In mobile computing, machine learning models for natural language processing (NLP) have become one of the most attractive focus areas in research. Association rules among attributes are common knowledge patterns, which can often provide potential and useful information such as mobile users′ interests. Actually, almost each attribute is associated with a hierarchy of the domain. Given an relation R = (U, A) and any cut αa on the hierarchy for every attribute a, there is another rough relation RΦ, where Φ = (αa : a ∈ A). This paper will establish the connection between the functional dependencies in R and RΦ, propose the method for extracting reducts in RΦ, and demonstrate the implementation of proposed method on an application in data mining of association rules. The method for acquiring association rules consists of the following three steps: (1) translating natural texts into relations, by NLP; (2) translating relations into rough ones, by attributes analysis or fuzzy k‐means (FKM) clustering; and (3) extracting association rules from concept lattices, by formal concept analysis (FCA). Our experimental results show that the proposed methods, which can be applied directly to regular mobile data such as healthcare data, improved quality, and relevance of rules.

Formal Concept Analysis (FCA) is a valid tool for data mining and knowledge discovery, which identifies conceptual structures from (formal) contexts. As many practical applications involve non-binary data, non-binary attributes are introduced via a many-valued context in FCA. In FCA, conceptual scaling provides a complete framework for transforming any many-valued context into a context, in which each non-binary attribute is given a scale, and the scale is a context. Each relation in relational databases is a many-valued context of FCA. In this paper, we provide an approach toward normalizing scales, i.e., each scale can be represented by a nominal scale and/or a set of statements. One advantage of normalizing scales is to avoid generating huge (binary) derived relations. By the normalization, the concept lattice of a derived relation is reduced to a combination of the concept lattice of a derived nominal relation and a set of statements. Hence, without transforming a relation into a derived relation, one can not only determine concepts of the derived relation from concepts of given scales, but also determine concepts of the derived relation from concepts of a derived nominal relation and a set of statements. The connection between the concept lattice of a derived nominal relation and the concept lattice of a derived relation is also considered.