The heat wave, which was first designated as a natural disaster in 2018 and causes casualties and property damage due to temperature rises above a certain standard, is becoming more serious at domestic and abroad. In particular, Busan faces the need to establish strategies to cope with the heat wave for its largest number of thermal disease among all metropolitan areas of the Korean Peninsula. This study aims to provide foundations for establishing heat wave strategies by using LSTM techniques, an artificial intelligence (AI) methodology, to reflect the urban-social characteristics of Busan. LSTM optimization analysis results show higher accuracy than conventional regression models and ensemble models, identified by MAE 0,139 and MSE 0.128. In addition, we perform a feature importance analysis to examine the effects of the utilized variables, and the results showed that the temperature-related variables had the highest impact. Significance of this study is found in predicting heat waves by reflecting the urban-social characteristics of Busan beyond simply utilizing climate data through AI methodology. It is expected that heat waves would be more accurately predicted by supplementing future data, adding variables, and improving models.