Financial fraud detection is critical for market transparency and regulatory compliance. Existing methods often ignore the temporal patterns in financial data, which are essential for understanding dynamic financial behaviors and detecting fraud. Moreover, they also treat companies as independent entities, overlooking the valuable interrelationships. To address these issues, we propose ACRF-RNN, a Recurrent Neural Network (RNN) with Attention-based Conditional Random Field (CRF) for fraud detection. Specifically, we use an RNN with a sliding window to capture temporal dependencies from historical data, and an attention-based CRF feature transformer to model inter-company relationships. This transforms raw financial data into optimized features, fed into a multi-layer perceptron for classification. Besides, we also use the focal loss to alleviate the class imbalance problem caused by rare fraudulent cases. This work presents a novel real-world dataset to evaluate the performance of ACRF-RNN. Extensive experiments show that ACRF-RNN outperforms the state-of-the-art methods by 15.28% in KS and 4.04% in Recall. Data and code are available at: https://github.com/XNetLab/ACRF-RNN.git.

The detection of oblique wedge is a crucial task in the detection and repair of railway freight car bogies, which involves multiple inspection items. Among them, the wear degree of rubbing surface directly affects the damping effect and stability of the train, which is a very important inspection item. the exiting method is the manual measurement that uses the non-standard measuring tools and visual inspection. which cannot achieve the ever-increasing measurement efficientcy and accuracy of automate maintenance. In order to solve this problem, the measurement method and algorithm based on point cloud are researched to measure the wear of the rubbing surface. The test results shows that the method is superior to the existing manual measurement method in terms of accuracy and efficiency, and provides important technical support for the automatic maintenance and upgrading of the oblique wedge.

Based on that the normal lines of points on the same latitude of a surface of revolution point to the same point on the axis, a strategy is proposed to find the axis of the surface of revolution by fitting the intersections of normal lines. The k-d tree is used in this paper to address the computational challenges. The shortest distance between the normal lines of two points in the neighborhood of a certain point is calculated to determine whether the intersection of normal lines needs to be calculated. The RANSAC algorithm is used to fit the line to the normal lines intersections to obtain the axis. The axis extraction method is tested on cone, cylinder, spindle, and incomplete shapes of them. the error of extracted axis is small, indicating the feasibility of this method.

Based on the state-space representation of inertia wheel pendulum, Euler method is used to update the state of inertia wheel pendulum system. The effect of actor-critic algorithm on inertia wheel pendulum system control is studied with different sample time and different proportion of angle in reward. The results show that the training effect is better with smaller sample time. Because the main goal of inertia wheel pendulum control is to control the angle of the pendulum, more weight should be given to the angle in reward to make the training to be stable more easily.