| In recent years,with the intensification of competition in modern society,an increasing number of groups have been troubled by psychological health issues that severely affect their normal learning,work,and life.However,the traditional method of identifying psychological health through filling out psychological questionnaires has subjective and costly drawbacks and is not suitable for large-scale use.With the widespread application of terminal devices such as smartphones and wearable devices,research on psychological health identification systems based on perceptual data is becoming increasingly popular in both academia and industry.However,before applying it to practical scenarios,there are still several issues that need to be addressed.Firstly,perceptual data is highly heterogeneous,and psychological health identification involves multiple complex related tasks,requiring the design and construction of high-performance psychological health intelligent identification algorithms to solve the above issues.Secondly,the storage and computation of massive perceptual data demand higher requirements for system performance and stability.Finally,for real-time psychological health identification tasks,the system needs to complete real-time data processing and respond promptly under low latency conditions.The highly efficient and real-time psychological health intelligent identification system proposed in this paper successfully addresses the above issues.Firstly,the system designs a psychological health intelligent identification algorithm based on multi-task and multi-view learning.On the one hand,to address the high heterogeneity of perceptual data,the algorithm model employs a fully connected neural network for unified representation at the multi-view layer.Then,based on attention mechanism,the algorithm model performs featureweighted fusion of the feature vectors from different views to obtain important and comprehensive information from multiple views.On the other hand,to address the multi-task identification problem,the algorithm model proposes a multi-layer multi-task learning framework,which reduces the time and computation cost of the model compared to the singletask model.Additionally,by fully utilizing the complex relationships between tasks,the multitask learning framework can significantly improve the algorithm’s performance and generalization ability.Moreover,extensive experiments were conducted on a real dataset,and the results proved that the intelligent identification algorithm proposed in this paper had the best performance in all indicators compared to the baseline.Furthermore,this paper uses a finergrained psychological health dual-factor model based on the definition of psychological health.In addition to psychological disorders,it also considers individuals’ positive experiences,which can more comprehensively and accurately reflect users’ psychological health status.Secondly,in the face of massive perceptual data,the system uses an efficient and reliable storage and computation scheme.Unlike traditional single-machine or simple distributed architectures,the system stores data in memory for processing,avoiding frequent disk read/write operations,thereby accelerating data processing speed.Meanwhile,the data is divided into multiple blocks and replicated across multiple nodes in the cluster to prevent a single-point failure.The system can flexibly expand according to data size and computational demand,providing high-performance data processing capabilities.Finally,the system completes real-time computing tasks under low latency conditions through a native stream computing framework.By processing data in the form of streams instead of waiting for data to accumulate to a certain extent before batch processing,the system can respond to real-time data more quickly.Meanwhile,to maximize data processing efficiency,the system employs a lighter data consistency scheme,avoiding data duplication or loss at a lower cost. |
PDF Full Text Request