The different statistical analyses techniques and machine learning techniques
that I investigated include the predictive models. The main rationale behind
using the techniques was to determine the likelihood or chances that a given
storm can lead to deaths. I investigated five predictive models, which include
neural network, logistic regression, random forest, support vector machine
(SVM), neural network, and a heterogeneous ensemble model (Lind et al.,
2017). The logistic regression model uses logarithmic functions with values
ranging from 0 to 1 for calculating the probability that the target will belong
to a certain class. The neural network uses backpropagation for predicting the
class of the target variable. The support vector machine is a classification
model that separates the classes with the use of a hyperplane. The random
forest is a type of ensemble model that uses decision trees with random data
samples. Finally, the heterogeneous model that I investigated is a
combination of the neural network, the support vector machine, and the
backward linear regression model. The results of my investigation were that
all the statistical analyses techniques and machine learning techniques that I
used attained an accuracy rate of more than 80% and sensitivity rate more
than 70% after changing the cutoff thresholds. It means that all my statistical
techniques have been proved to be effective in classifying the dangerous
storms present in the dataset (Lind et al., 2017). The models mainly classified
the storms as lethal and non-lethal storms. The techniques helped in finding
out the different variables that can have a strong impact on the storm
casualties. In my investigation, I could find out that the most important
variables include the storm location, range, damage to property, occurrence
of injuries, and crop damage. All the statistical analysis and machine learning
techniques that I have investigated prove to be effective in predicting the
potential casualties in the storm in the future and promote safety measures to
save lives.
References
Lind, D. A., Marchal, W. G., & Wathen, S. A. (2017). Statistical techniques
in business & economics. McGraw-Hill Education.
