Data Analytics Implementation for Surabaya City Emergency Center

Abstract

Quick response service and emergency reports handling is one of the main aspects in the data-driven government system, oriented to people service in the city of Surabaya through an emergency center called as Command Center 112. Our idea is to implement descriptive and predictive analytics to be able to provide a detailed picture of the intensity of the number of reports of each category and sub-district in the city of Surabaya as well as make predictions to find out future public report projections by analyzing spatial and temporal data. For descriptive analysis, we apply the unsupervised learning method with agglomerative hierarchical clustering combined with K-Means clustering for centroid initialization. After the data is preprocessed, such as imputation and data structure improvement, the data is then transformed into a report number format for each month and category, then segmented with the K-Means clustering hierarchical model, this model will get 3 final labels. These labels will be projected (grounding) to the level of intensity of community reports in the month and category, ranging from the low, medium and high categories. As for the prediction model, in this study we use combination of timeseries prediction methods, such as Exponential Smoothing, Moving Average and Auto Regressive Integrated Moving Average (ARIMA) by modifying the parameters according to the characteristics of movement, trends and seasonal data. We applied the model that we proposed for research purposes with a dataset of reports from the people of Surabaya to the Command Center 112 in 2019 with a total of 169,937 data.