I have tested out a multiclass text classification using multinomialNB() and LogisticRegression() , both have 10-fold cross validation accuracy of around 98% where multinonialNB is a little higher. I have used TFIDF for word embeddings. Should I go with multinomialNB as my final choice since it has 10 fold accuracy of 98% or should I try algorithms like LSTM and CNN with word embeddings like word2vec? (Total data size is around 1000 records only)
Since MultinomialNB and Logistic Regression require less computational power I started with these and thought if its less accurate to try LSTM or CNN deep learning , is this is the right way to approach?
Thank You in Advance.