Focused on the time-consuming and laborious problem of data annotation in review text sentiment analysis, a new automatic data annotation method was proposed. Firstly, the sentiment tendency of the review text was calculated based on the sentiment dictionary. Secondly, the review text was automatically annotated by using the weak tagging information of the user and the sentiment tendency based on the dictionary. Finally, Support Vector Machine (SVM) was used to classify the sentiment of the review text. The proposed method reached 77.2% and 77.8% respectively in the accuracy of sentiment classification on two types of data sets, which were 1.7 percentage points and 2.1 percentage points respectively higher than those of the method only based on user rating. The experimental results show that the proposed method can improve the classification effect in movie reviews sentiment analysis.

Aiming at the shortcomings such as low prediction accuracy and low efficiency of model training in alarm prediction of communication networks, a communication network alarm forecasting scheme based on Canonical-order tree (Can-tree) weighted incremental association rule mining algorithm was proposed. Firstly, the alarm data was preprocessed to determine the alarm data weight and compressed into the Can-tree structure. Secondly, the Can-tree was mined by using the incremental association rule mining algorithm to generate alarm association rules. Finally, a pattern matching method was used to predict real-time alarm information, and the results were optimized. The experimental results show that the proposed method is efficient, and the previously mined results can improve the mining efficiency. The alarm weight assigning scheme can reasonably distinguish the importance of alarm data, help mine the alarm association rules with high importance, speed up the elimination of outdated alarm association rules, and improve the accuracy and precision of the prediction.

To analyze the difficulties of detection under “soft” failures, an improved B-style grey relationship degree based on the moving state propagator for the federated filter (FF) was developed to solve above problems. The mode can obtain the highest fault tolerance quality without feedback, mutual pollution was avoided by using FF. Also, an adaptive balance coefficient and its algorithm was presented to balance the optimal information sharing approach and the fault tolerance one. According to the algorithm in this paper, the information sharing coefficients were adaptively adjusted by the failure grade, and the approximate highest fusion accuracy was ensured, the algorithm has the character of low calculation, simple configuration, high precision, and was suitable to practical application. This simulation results indicate that fusion accuracy under the largest range of failures was increased approximate 28.5%, which shows the fusion accuracy under failures was efficiently improved and the approximate highest fusion accuracy of whole process was realized.

Ultrasound Color Flow Imaging (CFI) is widely used in clinical diagnosis. This paper made two improvements for previous research about Graphics Processing Unit (GPU) framework for CFI: on wall filtering block, the parallel regression filtering was implemented instead of Finite Impulse Response (FIR) filter; on post-processing block, the parallel threshold box filtering was added to improve flow uniformity. The experimental data were acquired from a healthy carotid artery: 88 scan lines, 510 samples along the scan line, and an ensemble size of 16. The experimental results show that, compared with the serial algorithm based on Central Processing Unit (CPU), the computational efficiency of the proposed parallel algorithms based on GPU has been increased by 15.2 times and the frame rate increases up to 70. The color flow image can achieve a high quality by the regression filter instead of FIR filter, and improve the accuracy of tissue/flow detection by threshold box filter.