A carregar...
Publicação
Optimization of firefighter response with predictive analytics : practical application to Lisbon, Portugal
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 8.85 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Time is a crucial factor for the outcome of emergencies, especially those that involve human lives.
This paper looks at Lisbon’s firefighter’s occurrences and presents a model, based on city
characteristics and climacteric data, to predict whether there will be an occurrence at a certain
location, according to the weather forecasts. In this study three algorithms were considered, Logistic
Regression, Decision Tree and Random Forest, as well as four techniques to balance the data –
random over-sampling, SMOTE, random under-sampling and Near Miss –, which were compared to
the baseline, the imbalanced data.
Measured by the AUC, the best performant model was a random forest with random under-sampling
at 0.68. This model was well adjusted across the city and showed that precipitation and size of the
subsection are the most relevant features in predicting firefighter’s occurrences.
The work presented here has clear implications on the firefighter’s decision-making regarding vehicle
allocation, as now they can make an informed decision considering the predicted occurrences.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business Intelligence
Palavras-chave
Emergency response Firefighters Predictive modeling Smart city