In traditional financial analysis discriminant analyses with statistic methods are considered as primary and widely applied corporate credit rating bases in which human impact plays an essential role. Machine learning approaches enormously assist credit rating in reducing human interference and improving rating efficiency. However, existing methods did not efficiently leverage the time series features of financial data that is beneficial for the classification tasks. We propose a novel end-to-end architecture, the SMAGRU, based on self multi-head attention with gated recurrent neural networks, which is capable of evaluating a corporate credit rating by capturing the time series features assigned the weights of market benchmarks. Experiments compared with statistic and machine learning baseline models on different datasets show that our model contains superiority in multi-class credit rating classification. The result of ablation experiments suggests that time series features are crucial in classification performance, and self multi-head attention can significantly enhance these features and thereby improve the accuracy and convergence speed. We show that the SMAGRU generalizes well, even on sparse data during K-fold cross validation experiments.

In light of the escalating global concerns surrounding climate change, the significance of sustainable development in the realm of logistics cannot be overstated. This study undertakes the imperative task of devising strategies aimed at mitigating carbon emissions, reducing logistics costs, minimizing transportation time, and enhancing customer satisfaction. The research delves into the intricacies of an optimization model tailored for a specific iteration of the Location-Routing Problem (LRP), namely the Multi-Objective Multi-Period Low-Carbon Location-Routing Problem (MMLCLRP). This variant of the LRP takes into meticulous consideration several crucial parameters, such as the overall logistics cost, the arrival times of demand points, and carbon emissions. These factors are pivotal in determining both the optimal location for depots and the programming of routes within a multi-period planning horizon. The proposed model guarantees the long-term sustainability of logistics operations while flexibly adapting location routing decisions for each period in response to evolving market demands. To tackle the inherent complexity of this problem, an improved version of the Non-dominated Sorting Genetic Algorithm (NSGA-II) was employed. This approach integrates a pioneering similarity distance metric to quantify the resemblance between potential solutions. Additionally, a crowding clustering strategy was implemented to enhance the diversity within the NSGA-II. Empirical results illustrate the capability of the proposed optimization model in effectively harmonizing various objectives, encompassing economic, efficiency, and environmental aspects within the logistics domain. Additionally, the enhanced algorithm exhibits notable advantages in addressing the complexities inherent in the optimization model.

Grammatical error correction has been considered as an application closely related to daily life and an important shared task in many prestigious competitions and workshops. The neural machine translation with an encoder-decoder architecture containing language models has been the fundamental solution for the grammatical error correction. Whereas Grammatical error correction task on texts of hearing impaired people or its solution has not been seen yet, and common Grammatical error correction tasks are suffering several challenges, such as insufficient training data, insufficient accuracy due to the unsatisfactory capacity of extracting semantic and grammatical patterns. Under these circumstances, we proposed a novel encoder-decoder architecture based on multi-head self-attention along with multiple strategies, which excels at extracting deep representations from the corrupted sentences of hearing impaired students and further reconstructing the sentences into grammatical ones. Via the re-ranking strategy, our model can correct various kinds of errors including spelling and complex syntax errors. The ablation experiments prove that the semantic extracting of self-attention mechanism excluding the position encoding with the word order shuffle operation can significantly learn the hearing impaired students’ sentence patterns whose word order is quite different from the ones of hearing people and improve the correction scores. The pre-training can enhance the restoring efficiency of sentence structure in the decoding process. The comparison experiments with baseline models show that our model obtains superior performance either in the hearing impaired students’ grammatical error correction or in a common grammatical error correction shared task.

The particle swarm optimization (PSO) is a wide used optimization algorithm, which yet suffers from trapping in local optimum and the premature convergence. Many studies have proposed the improvements to address the drawbacks above. Most of them have implemented a single strategy for one problem or a fixed neighborhood structure during the whole search process. To further improve the PSO performance, we introduced a simple but effective method, named adaptive particle swarm optimization with Gaussian perturbation and mutation (AGMPSO), consisting of three strategies. Gaussian perturbation and mutation are incorporated to promote the exploration and exploitation capability, while the adaptive strategy is introduced to ensure dynamic implement of the former two strategies, which guarantee the balance of the searching ability and accuracy. Comparison experiments of proposed AGMPSO and existing PSO variants in solving 29 benchmark functions of CEC 2017 test suites suggest that, despite the simplicity in architecture, the proposed AGMPSO obtains a high convergence accuracy and significant robustness which are proven by conducted Wilcoxon’s rank sum test.

To construct a neural network model (ATBP) for predicting susceptibility to Post-inflammatory hyperpigmentation (PIH), which is a rapid, objective, and reliable decision-support method before physical and chemical interventions in dermatology clinics for pigment disorders. A dataset was established based on the VISIA Skin Analysis System detection results of 1953 patients with pigment disorders including 93,477 labeled data under 8 indicators. A novel Post-inflammatory hyperpigmentation susceptibility prediction model incorporating Multi-head self-attention mechanism and Back-propagation neural network is proposed to capture the patterns of skin detection data to predict PIH susceptibility. The results of comparison experiments indicate that Attentive BP (Back Propagation Neural Network) has a significant superiority in prediction accuracy (0.8604) compared with other machine learning models. The ablation experiments prove that the Multi-head self-attention mechanism substantially improves the accuracy and the stability of prediction. The results of the 10-fold cross-validation experiment prove that ATBP is robust and avoids turbulence in predicting. Leveraging Multi-head self-attention mechanism and the architecture advantage of BPNN, the proposed model ATBP obtains the robust and efficient prediction performance in predicting PIH susceptibility via processing large-scale and hi-dimension data, i.e., considering comprehensive skin conditions of individual patient. It can be proved from the experimental results that the proposed model is reliable for decision-support work of PIH susceptibility.