Research On IoT Service-oriented Business Process Activity Recommendation Method

The service data information generated in the Internet of Things(IoT)environment will have a significant impact on the execution of enterprise business processes.IoT services connect the Internet of Things with business processes,and the service data generated by IoT services can drive the execution of business processes.In the fields of smart healthcare,smart home,smart elderly care,etc.,predicting subsequent business processes based on real-time scenarios can predict the progress of business processes in advance,avoid risks,and avoid serious losses.Therefore,how to dynamically predict and recommend subsequent processes based on IoT service data generated in the IoT environment combined with process instance information to avoid important risks is a topic worth studying.In response to the above issues,this article has conducted research on business process activity prediction methods for IoT services.The main work is as follows:(1)Based on the large amount of IoT service data generated in the IoT environment,this article explores the impact of additional service data information on process prediction results in the next step of business process activity prediction.At the same time,a method for defining activity classification suitable for different scenarios was proposed.According to the business importance of the activity itself,the additional service data information is classified into case level additional information and event level additional information,and the business process activities in the event log are classified into process activities and environmental activities.The experimental results show that the data partitioning definition method of classifying and predicting based on the business importance of the activity itself can improve the prediction accuracy of the model to a certain extent,and the additional service data information can significantly improve the performance of the business process activity prediction model.This method is suitable for various IoT scenarios.(2)A new expression method for event log data has been designed,which transforms the event log data into a graphical structure with visual features.Based on the graphical structure of the process,various different feature matrices have been constructed,which can clearly display the connections between various activities and their execution status.At the same time,machine learning models such as LSTM,MLP,and RF were also selected for comparison of prediction results.Through experiments,it has been proven that IoT service data can improve the prediction accuracy of models,and traditional machine learning models perform well in prediction performance.At the same time,the results of using different feature matrices to train and predict the model are also different.(3)We have developed a business process activity prediction prototype system for IoT services,which enables the collection,preprocessing,feature extraction,model training,and optimization of business process log data in the IoT environment.The training results can also be manipulated and displayed.