Indonesian Finance News Sentiment from Hybrid Deep Learning and Support Vector Machine

Abstract

One common action to keep aware of the current investment progress is by updating finance news continuously. Indeed, we can read a bunch of news relating to finance from social media, which is often difficult to figure out at a glance. Hence, this work aims to propose hybrid models that can help us to classify whether the finance news is positive to follow. Also, we may sort a few articles containing neutral ones. More specifically, we incorporate deep neural networks: deep convolutional neural networks and long short term memory, to draw diverse word representations, and support vector machines to categorize them as a multi-class classification case. In this work, we evaluated the proposed models on Indonesian finance news that was officially reported from the Bank of Indonesia around 2019 before the pandemic started. In the evaluation results, we showed the DCNN-SVM better accuracy compared to others.