Classification Using Backpropagation Neural Network on Tweet Emotion Intensity

This Research presents text classification for tweet emotion intensity using Backpropagation Neural Network (BPNN) method. In this study, Twitter data is used for the research, where the data were categorized into four categories (fear, joy, anger and sadness). The dataset is included with emotion intensity with a scale from 0 to 1. The preprocessing steps consist of cleaning the data, normalization, tokenization, Bag Of Word (BoW) method, and steaming process. The intensity features in the dataset are analyzed and filtered with a certain thresholds to ensure only the right emotion is used for training. BPNN process is using 100 neuron for hidden layer, 4 neuron for the output layers with 50 epochs. This research can produce results with a value in accuracy and recall of 93.71% and 94.3%.