Application of text mining for classification of community complaints and proposals

Abstract

Enabled by the increased connectivity and ease of access to online services, an increasing number of countries are moving towards participatory decision making. In Jakarta, the Government deploys an e-participation tool to directly involve community associations and citizens in the development planning of the province. The increasing number of participation and the need of immediate response encourage the use of text mining to classify the complaints and proposals automatically. A classification model was built using Support Vector Machine (SVM) algorithm. This study was the first part of research on community complaint and proposals before text clustering and visualization. The result of model development for classifying documents showed that classification model with stemming and synonym recognition was the most accurate among others with 91.37% accuracy rate. Adding the number of training data would improve the accuracy. Based on classification result, the problem of flood and transportation became the most reported problems during January and February 2016. This result indicated that these problems need to be prioritized by the Government of Jakarta.