Research On User Recruitment And Individual Behavior Recognition Technology Based On Group Intelligence Perception

With the development of science and technology,the popularity of various wearable intelligent devices and portable mobile devices,crowd sensing technology has been widely used.In crowd sensing,the sensing platform collects a large number of individual behavior data for researchers to study individual behavior recognition technology,which has broad development prospects and strong practical value.The sensing platform needs to recruit users to participate in the data collection task,so how to quickly recruit users to participate in the sensing task,and then how to deal with individual activity data and activity recognition are the main problems to be solved in this paper.The main contents of this paper are as follows:(1)Influence Maximization User Recruitment Program(IMURP)is proposed.In crowd sensing,the sensing platform recruits users to participate in the task of sensing data collection,which requires a large amount of data and high data quality.In social networks,if a user has a high impact,then the user as the information source will spread information more quickly and widely.This paper analyzes the influence of users under the linear threshold propagation model,and the user set with the maximum influence is constructed by combining the greedy idea,so,more participants can be recruited to join the task of perceptual data collection.(2)An individual activity recognition method based on Jaya optimization algorithm is proposed.The k-value of k-NN classifier has a great influence on the final classification results,but there are no pre-defined rules for the selection of k-parameter.Many researchers are committed to find an optimization method that can quickly find k-value.Compared with other swarm intelligence optimization algorithms,Jaya optimization algorithm has no too many parameters and is easy to understand.Therefore,this paper proposes a Jaya-k-NN activity recognition algorithm,which uses Jaya optimization algorithm to optimize the k-value of k-NN classifier.The validity of the recognition algorithm is verified by using the open daily behavior and motion behavior data set.The experimental results show that compared with the traditional classification methods,using Jaya algorithm optimization can quickly find the appropriate k value,which makes the k-NN classifier have a better classification rate.(3)This paper proposes integrated method to identify individual activity.Ada Boost and Random Subspace can change the distribution of sample data,and integrate multiple weak classifiers.The Ada Boost-k-NN and Random Subspace-k-NN activity recognition integrated models are constructed respectively.The experimental results show that,compared with Jaya-k-NN single classifier,the integrated learning model used in this paper can achieve significantly better generalization performance and better classification effect.