Research On Recommendation Method Of Safety Examination System Based On Collaborative Filtering Algorithm

Safety production of power plants has always been the most important issue in power production.Power plants pay more and more attention to their own safety standards,and pay more and more attention to safety training for employees.The safety training management system based on B/S structure adopts the server-browser mode to realize the safety assessment and learning tasks of the power plant online,which effectively improves the learning efficiency of the power plant personnel.However,as the number of questions in the question bank continues to expand,it will take a lot of time to manually retrieve the questions that are suitable for the user's practice.The recommendation system proposed in this paper is to add the test recommendation function to the safety assessment system.The system can target each user.The mastery of the situation is recommended to be suitable for the subject of the exercise,to improve the efficiency of learning.This article has mainly completed the following:1.Learn user-based collaborative filtering algorithm and item-based collaborative filtering algorithm,and compare their applications in different recommendation systems.This research is implemented by the collaborative filtering algorithm based on items.According to the requirements,a data table is designed to save the user-item score matrix and the item-item similarity matrix in the database.We analyze the event report contained in the system,summarize the content information in the event report,analyze and select the useful extraction scope to decompose it by the IK Analyzer segmentation method,and Set up the sequence of key words.2.Learn Chinese word segmentation technology,collect questions in question bank,remove punctuation marks from text for data preprocessing.Several commonly used Chinese word segmentation tools,such as Stanford,SnownLp and stuttering,are tested and compared.Finally,a stuttering participle is selected for word segmentation.3.Statistical part of the user's mistake information,generate the user-item score matrix and save it to the database;Chinese word segmentation is carried out for each topic,word vector is constructed based on the 2-generated thesaurus,distance between the two topics is calculated,and similarity matrix of the item-item is constructed and saved to the database.4.Recommendation is completed according to user-item score matrix and item-item similarity matrix.When the user uses this function,the system can recommend similar topics to users according to the situation of the user's previous wrong questions.