I am using KNIME for Sentiment Analysis of Twitter data i.e. the tweets retrieved for a specific keyword using SVM. SVM is trained using labelled data and SVM predictor is provided input of preprocessed tweets. But the SVM Predictor gives error : column "quot" was not found. It exactly wants the same columns (bag of words) in the testing set as given in the training set which is not possible for Twitter data. How can I fix it? Attaching the screenshot of the process.
Hi Jasleen,
the feature space must be equal, for the training and the application of the model. The BoW of the second set of documents has to be filtered according to the BoW use for training.
Attached is an example workflow, that shows the idea.
If you have two separate workflows for model training and application, you need to store the terms of the BoW used for training to a file and read this file in the second workflow and filter the BoW based on that data.
Cheers, Kilian
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.