I am a machine learning beginner and decided to do sentiment analysis for tweets. I don't wanna use libraries like textblob etc but rather build my own model. I am using this dataset from Kaggle link
to train my model. So my doubt is that i wanna build a model that can give number from -1 to 1 where -1 being negative and 1 being positive. Since sentiment analyzer are mostly classifiers , so they just can predict between 0 to 1. Can MSE or RMSE be used since i want to label the values between -0.5 to 0.5 as Neutral. So is it possible? or can I use tanh activation as my last layer and MSE as loss function