AbstractEdible oil adulteration is a mostly practiced phenomenon. However, the traditional discriminant methods fail to detect oil adulteration involving more than one adulterant. Recently, one‐class classifiers were built for food or oil authentication. Unfortunately, as it is hard to determine the application domain of the one‐class classifier, high prediction error was obtained for real samples in market surveillance. In this study, a new method was developed based on one‐class classification and outlier detection for edible oil adulteration detection in market surveillance. The model population was constructed using Monte Carlo sampling of unidentified inspected samples to select the plateau region exhibiting the highest accumulated absolute centered residual (ACR) values. Subsequently, the number of models in the plateau region was validated by the theoretical ones calculated by the classical probability model. The models in the plateau region with the highest cumulative accumulated ACR values were used to identify adulterated oils. Furthermore, the cross‐validation was conducted by comparing identification results from two different Monte Carlo sampling ratios to ensure the accuracy of our method. Both single adulteration and multiple adulteration of peanut oils were prepared to validate our method. Moreover, this method was used to detect adulteration of sesame oils, which have already been identified by the markers in our previous study. The validation results of three datasets indicated that this method could effectively identify adulterated samples and therefore provide a novel solution for inspecting potential adulteration in practice.

Endotoxins, also known as lipopolysaccharides (LPS), are present within the cell walls of Gram-negative bacteria and are released upon cellular death, which can pose a significant risk to human and animal health. Due to the minimal amount of endotoxin required to trigger an inflammatory response in human body, the demand for sensitive methods with low endotoxin detection limits is essential necessary. This paper presents a straightforward aptamer sensor which can enhance the conductivity and specific surface area of molybdenum disulfide (MoS2) by incorporating carboxylated multi-walled carbon nanotubes (MWCNTs-COOH) and polyaniline (PANI). Doping with gold nanoparticles (AuNPs) improves biocompatibility and sensitivity while providing binding sites for thiolated endotoxin-binding aptamers (LBA). This biosensor achieved a remarkable detection limit as low as 0.5 fg mL−1, enabling trace-level identification of LPS. It also exhibits excellent repeatability, selectivity, and stability, facilitating rapid and accurate LPS detection. Moreover, this method demonstrates high recovery rates and specificity for LPS analysis in food samples, showcasing its promising application prospects in trace-level LPS detection within the food industry.

In text classifier models, the complexity of recurrent neural networks (RNNs) is very high because of the vast state space and uncertainty of transitions, which makes the RNN classifier’s explainability insufficient. It is almost impossible to explain the large-scale RNN directly. A feasible method is to generalize the rules undermining it, that is, model abstraction. To deal with the low efficiency and excessive information loss in existing model abstraction for RNNs, this work proposes a PSO (Particle Swarm Optimization)-based model abstraction and explanation generation method for RNNs. Firstly, the k-means clustering is applied to preliminarily partition the RNN decision process state. Secondly, a frequency prefix tree is constructed based on the traces, and a PSO algorithm is designed to implement state merging to address the problem of vast state space. Then, a PFA (probabilistic finite automata) is constructed to explain the RNN structure with preserving the origin RNN information as much as possible. Finally, the quantitative keywords are labeled as an explanation for classification results, which are automatically generated with the abstract model PFA. We demonstrate the feasibility and effectiveness of the proposed method in some cases.

In complex security markets, uncertainty and randomness may coexist. The investors hold different risk attitudes toward different investment objectives in reality. In order to reflect this phenomenon, this paper applies mental accounts to portfolio optimization in an uncertain random environment. Firstly, considering the influence of transaction costs and diversification degree, this paper establishes an uncertain random mean-absolute semi-deviation-entropy bi-objective optimization model. Then the bi-objective model is converted to two single-objective models through three steps. Secondly, the equivalent forms of the models are deduced when the return rates of random risky securities are assumed to be normal random variables and the return rates of uncertain risky securities are assumed to be linear and zigzag uncertain variables. Furthermore, we propose an improved butterfly optimization algorithm (IBOA) to solve the two single-objective models. Finally, numerical simulations are presented to analyze the practicability and effectiveness of the models with different mental accounts and the IBOA algorithm. The results indicate that the IBOA algorithm is effective and putting money into more mental accounts may gain higher returns.