Abstract In recent years, with the completion of the new library of Qinghai University, the collection of books in the library has greatly increased. The library has a total collection of 880,000 volumes, covering a dozen disciplines such as science, engineering, agriculture, literature, history, economics, philosophy, law, education, management and medicine. It is difficult for users to find the books they are interested in among the numerous materials. Based on the actual situation of the library of Qinghai University, the differences of different professional users and their personal interests, this paper chooses the item-based collaborative filtering algorithm to realize personalized recommendation. First of all, in the calculation of book similarity, the traditional user score data is not chosen to calculate the similarity, but to calculate the similarity between books and books according to the feature vector of book name. Secondly, in order to avoid the problem of cold start, the system recommends the users who have no borrowing record, but the most borrowed books in their department. The combination of the two realized the personalized recommendation of books. By comparing with other traditional recommendation algorithms, it is found that the algorithm adopted in this paper has better recommendation effect.
The quality of the data determines the quality of the model. In this paper, the grassland degradation data in the Headwaters of the Three Rivers were processed in the early stage and labeled with multiple classification. Guided clustering and semi-supervised clustering were used for comparison. The two methods were combined to classify and label the data, so as to improve the accuracy and completeness of the classification data.
The Three Rivers Source Region, situated on the Qinghai-Tibet Plateau, is an important natural reserve in China, with its grassland ecosystem playing a crucial role in maintaining regional ecological balance and water conservation. However, in recent years, the area has experienced severe grassland degradation due to factors such as overgrazing, grassland fires, climate change, and rodent infestations. Therefore, the urgent need for grassland restoration and management has emerged. To effectively address grassland degradation, accurate identification of its severity levels (e.g., first-degree degradation, second-degree degradation, etc.) is essential, followed by the implementation of appropriate restoration measures corresponding to each level. This study focuses on rodent infestations as a perspective to investigate grassland degradation. Through surveys, it has been found that the coverage ratio of rodent burrows, specifically those made by plateau pikas and field mice, can be used to determine the degree of grassland degradation. Accordingly, this research attempts to utilize semantic segmentation techniques to identify rodent burrows, particularly those of plateau pikas, and subsequently calculate the coverage ratio, enabling the determination of grassland degradation levels. Currently, due to the absence of a comprehensive publicly available rodent burrow dataset, we have conducted on-site data collection in the Three Rivers Source Region to establish a dataset based on rodent burrows. Additionally, we have improved the semantic segmentation model, SeMask-FPN. Addressing the insufficiency of global interactions in the SeMask-FPN model, we propose two different global fusion models: the first incorporates burrows convolution and self-attention mechanisms, introducing the Atrous Self-Attention (ASA) module to create the ASA-based SeMask-FPN model. The second model combines pooling and self-attention mechanisms, introducing the Pooling Self-Attention (PSA) module to establish the PSA-based SeMask-FPN model. Experimental comparisons reveal that the ASA-based SeMask-FPN model and PSA-based SeMask-FPN model achieved mIou values of 86.43% and 81.9%, respectively, representing improvements of 11.4% and 6.87% over the original SeMask-FPN model's mIou value.
[Purpose] Personalized recommendation is one of the hottest research areas in recent years Recommendation systems developed by Google, Amazon, Alibaba and other companies have brought them huge benefits, which are recommendations based on big data analysis. However,For data sets with large data sparseness, traditional recommendation algorithms for historical records cannot obtain satisfactory recommendation results, and traditional recommendation algorithms often cannot discover the potential interests of users. In this paper, we managed to extend the personalized recommendation system to the University Library Lending system [Methodology]Firstly, to meet the challenge of data sparsity, we collected the information of readers and books in the borrowing records of Qinghai University Library in recent 20 years. Secondly, through the analysis and research of the Wide and Deep model, the recommendation model is obtained by joint training of LR (Logistic Regression) and DNN (Deep Neural Network) networks. Moreover, we improved the double-label of the Wide and Deep model into multiple labels and got the final model after extensive training. [Findings]The experimental results show that the accuracy of our book recommendation model is significantly better than traditional recommendation algorithms and hybrid recommendation algorithms. [Originality]Firstly we set up a large Qinghai University book data set for training and testing and verification.Secondly, completed the model migration and improvement of W & D. Through a large number of experiments for comparative research, it is concluded that the improved W & D model is suitable for book recommendation systems.The value of the AUC index of the traditional collaborative filtering model is the lowest. The AUC value of the weighted bipartite graph model is greater than the collaborative filtering model. The AUC value of the hybrid model is basically the same as that of the weighted bipartite graph model. The Wide & Deep model has the highest AUC value. Reached 0.75. Therefore, the Wide & Deep model is suitable for a book personalized recommendation system with sparse characteristics of big data.
Abstract When entering the heliosphere, galactic cosmic rays (GCRs) will encounter the solar wind plasma, reducing their intensity. This solar modulation effect is strongly affected by the structure of the solar wind and the heliospheric magnetic field (HMF). To address the effect during the solar maximum of cycle 24, we study the solar modulation under a scenario in which the weights for A = ±1 are determined by the structure of HMF, and the traveling time of GCRs simulated by SOLARPROP is taken into account. We then fit the cosmic-ray proton data provided by AMS-02 and Voyager in the energy range 4 MeV–30 GeV, and confirm that the modulation time lag in this model is about 9 months, which is consistent with the previous studies. This model incorporates a more realistic description of the polarity reversing and provides a more reliable estimation of the solar modulation effect during the maximum activity period.
When the Galactic Cosmic Rays (GCRs) entering the heliosphere, they encounter the solar wind plasma, and their intensity is reduced, so-called solar modulation. The modulation is caused by the combination of a few factors, such as particle energies, solar activity and solar disturbance. In this work, a 2D numerical method is adopted to simulate the propagation of GCRs in the heliosphere with SOLARPROP, and to overcome the time-consuming issue, the machine learning technique is also applied. With the obtained proton local interstellar spectra (LIS) based on the observation from Voyager 1 and AMS-02, the solar modulation parameters during the solar maximum activity of cycle 24 have been found. It shows the normalization and index of the diffusion coefficient indeed reach a maximal value in February 2014. However, after taking into account the travel time of particles with different energies, the peak time was found postponed to November 2014 as expected. The nine-month late is so-called time lag.
The evaluation of grassland degradation is an important part of ecological conservation research, and rodent infestation is a significant factor in grassland degradation. The presence of a large number of mouseholes means that the environmental balance of grassland has been destroyed, so the coverage of mouseholes can be used as an evaluation method for grassland degradation levels. In this paper, the image segmentation method is used to segment the mousehole images, Upernet is used as the segmentation network, and Swin Transformer as the Backbone. FAM and FSM modules are added to the Upernet network to solve the target misalignment problem when upsampling the network. The mIoU is improved by 5.3% according to the experimental results.
Studies by grassland workers have shown that the occurrence of degradation indicator plants in grassland is an important sign of grassland degradation. The detection of degradation indicator plants can provide a certain data basis for the study of grassland degradation. In this paper, a target detection algorithm for improving YOLOv5 model is proposed to detect the degradation indicator plants (wolfsbane) in grassland. Firstly, the target detection dataset of the grassland degradation indicator plant(wolfsbane)is constructed, and then the backbone network is optimized by adding a coordinate attention mechanism on the basis of the original YOLOv5 model; The original feature pyramid module in the feature fusion module is replaced by a weighted bidirectional feature pyramid (BiFPN) network structure, which realizes effective weighted feature fusion and bi-directional cross-scale connection; A small target detection layer is also added to further improve the detection accuracy of small targets. Experimental results show that the proposed improved algorithm achieves an average precision (AP) of 80.4%, which is 3.4% better than the original YOLOv5 model, and verifies the effectiveness of the improved model for the detection of degraded indicator plants (wolfsbane).
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
