Classification of Community Responses to Service Offices using a Combined CNN-LSTM Algorithm and Random Forest

Abstract

This study analyzes urban issues by examining people's social media responses. The government requires public feedback to assess its performance in addressing these issues. The study employs a combined CNN-LSTM & RF method to classify the responses into 16 agency categories based on their respective responsibilities. We utilize Continuous Bag-of-Words (CBOW) to convert text into vector form, conduct feature extraction using Convolutional Neural Network (CNN) followed by Long Short-Term Memory (LSTM) combination and then perform classification with Random Forest (RF). This research contributes to developing a practical combination algorithm for analyzing and classifying public responses. The combined results of these algorithms provide a multi-class service classification with 83% precision, 80% recall, and 81% F1 score, which enhances accuracy compared to separate models.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint