Study On Sentiment Analysis Algorithm Of MOOC Reviews

MOOC(Massive Open Online Course)is a new education mode based on communication and Internet technology.Compared with the traditional offline courses,MOOC breaks the constraints of time and space,reduces the threshold and cost of learning,and is conducive to the sharing and dissemination of educational resources.China’s MOOC started in 2013 and has been developing rapidly since then.By 2019,The number of MOOC learners and courses in China are in the world’s leading level.However,As a new education mode,unlike traditional offline courses,the construction of MOOC is still in the process of exploration,the quality of MOOC is still uneven.Comment is the direct feedback of learners to the teaching effect of MOOC.Therefore,mining the feature of comments and analyzing the sentiment of comments can help MOOC designers grasp the direction of curriculum construction better,which is of great significance to improve the quality of MOOC.Sentiment analysis of MOOC reviews belongs to the problem of text sentiment analysis.There are three methods of text sentiment analysis: dictionary based,machine learning based and deep learning based.The latter two methods are more widely used because they do not need to build artificial dictionaries in advance.As one of the main machine learning classification algorithms,SVM(Support Vector Machine)has good generalization and stability and it’s performance is often better than other algorithms.However,the effect of SVM is greatly affected by the parameters c and γ.In practice,grid search is generally used to determine the optimal parameters by traversing all possible parameters with a certain step size,which is timeconsuming and low accuracy.Therefore,aiming at the problem of sentiment analysis of MOOC reviews,this article construct the GA-SVM,including design of coding method,population initialization method,fitness function and its verification set,selection function,genetic operator,etc.At the same time,fitness calibration,heuristic crossover and other control methods are designed to optimize the convergence of the algorithm.GA-SVM algorithm is designed to improve the classification effect of MOOC reviews.This article crawls MOOC reviews from icourse.163.com,cleans the reviews and segments the words,constructs the word vector model by Word2 Vec,and mines the feature words and topics of the reviews.Then,the cleaned comments were grouped,balanced and screened manually,and the constructed word vector model was used to vectorize the comments combined with TF-IDF weighting,and the experimental sample set was constructed.On the sample set,this article first tests the segmentation ratio of validation set of GA-SVM,and then trains GA-SVM to compare with other traditional classification models such as SVM,NB(Naive Bayesian),LR(Logistic Regression),etc.Finally,the accuracy rate,accuracy rate,recall rate and F1-score were 90.26%,92.59%,94.33% and 93.45% respectively.The experimental results show that GA-SVM is superior to other traditional classification models in general,and it has high accuracy,precision and recall rate in sentiment classification of MOOC reviews,which is suitable for dealing with such problems.