Sentiment Analysis Practical Steps
0 likes14 views
The document outlines the steps involved in sentiment analysis, starting from data reading and text corpus building, to data cleaning and sentiment tagging. It discusses the preparation of training and testing datasets, the creation of a term-document matrix, and the training of multiple models using algorithms like SVM and Random Forest. Finally, it covers model evaluation metrics, including confusion matrix and cross-validation.
Sentiment Analysis Practical Steps
- 1. Steps in Sentimental analysis 1) Read the data 2) Build text Corpus 3) Data Transformation / Cleaning 4) Tag the sentiments 5) Checking the overall Sentiment Score >table (analysis$score) 6) Analysis$sentiment<- ifelse(analysis$score>0,”positive”, ifelse (analysis$score <0 , “negative”, “neutral”)) table(analysis$sentiment) 7. Clean the data again 8. Split the data into training & test data sets 9. Tagging the testing & training datasets >train_data$type=“train” >test_data$type=“test” Rupak Roy
- 2. Steps in Sentimental analysis 10. Combine Tweets<-rbind(train_data, test_data) 11. Building TDM matrix=create_matrix(………………..) 12. Convert to matrix data type mat=as.matrix(matrix) 13. Build the data to specify response variable, training set, testing set, container=create_contrainer(……………….) 14. Train the model/create model using algorithm svm,RF,TREE…etc 15. Test the model > results =classify_model(container, models) 16. Model Performance – Confusion Matrix , Recall Accuracy. 17. Model summary- summary(analytics) 18. Ensemble of models – analystics@ensemble_summary 19. Cross Validation Rupak Roy