|
Abstract
|
Background/Objectives: Suicide is one of the great social problems of modern society. We conducted data mining to predict the actual suicide rate in Republic of Korea. The purpose of this study was to find practical ways to prevent suicide.
METHODS/Statistical analysis: We collected the actual suicide rate data of The Statistics Korea. Then, after extracting words related to suicide through the Social Matrix, we looked at the trends in word-by-word searches from 2004 to 2017 in Google Trends. This study the difference between the actual suicide rate and the suicide-related words at 1, 2, and 3 months and how to affect the suicide words to actual suicide.
FINDINGS: When the two-month time difference was considered, the suicide-related keywords were the best predictors of the actual suicide rate. The relevant data was analyzed by using Cart, Neural Network, and Linear Model. Hybrid model predictive method, which combined the above-mentioned 3 analytics models, was used in order to gain more improved functions. Hybrid model predictive method among various data mining techniques was the best prediction method. The weight of 4 analytics models was determined by using MAE(Mean Absolute Error) value. MAE shows average prediction errors. Suicide, society, poor, and depression were the most frequently associated words in suicide rate. Improvements/Applications: We identifies a correlation between actual suicide rate and suicide related keywords found on the Internet, and further suggests the practical implication of the use of social data for prevention. © 2018, Institute of Advanced Scientific Research, Inc.. All rights reserved.
Language: en |