Financial Risk Assessment Based On Disaster Induce

Abstract

The security, authenticity and protection of electric power framework is a major dilemma now-a days because it is not a single system many other systems are dependent on it. Severe weather conditions are badly effecting our power infrastructures causing billions of dollar of financial losses. Apart from the economic losses, power outages disrupt the lives of millions of people including industrial and commercial customers. United States has a huge power infrastructure and face many problems due to these disasters. The aim of this thesis is to assess the financial risk associated with disaster-induced power outages. These disasters include mostly severe weather events including hurricanes, thunder storm, wildfire, snowfall, heavy wind and winter storms. It is necessary to take precautionary measures to avoid huge losses and also save the humanity from disasters. The prediction of financial losses and the factors being involved for power outage events is important. To accomplish this task a detailed exploratory and statistical analysis is required to see the correlation between different key parameters. For this purpose a machine learning algorithm Random Forest is used on publicly available data set of the United States from 2000 2016, containing data of about 49 States and comprising of 51 different variables. Exploratory analysis is carried out to form a base for this research work. Random Forest is a useful classifier used both for classification and regression. In this algorithm predicted results are compared with actual output in order to find the error and accuracy. In this research random forest is used for the purpose of predicting the financial losses against each disaster category for the top 8 financially affected states of US. It has one more advantage of predicting the importance of the features being involved for the losses. For the evaluation of results three types of error are calculated including MAE, MAPE and RMSE. It was found that outage duration and customers affected have the highest importance for each disaster in parameters ranking, and the major losses are due to Hurricanes, thunders storm and winter storms.