On April 15th in 1912, one of the most tragic shipwreck happened. The Titanic ship sunk under the Atlantic Ocean after colliding with an iceberg. A lot of people died during the accident. Small number of available lifeboats was one of the main reasons that led to such a high number of deaths. Such factors as sex, class, age, and others also significantly contributed to the likelihood of survival. Introducing different statistical methods, I will classify what sorts of people had a better chance of survival the shipwreck. For the data analysis, I have used the dataset “Titanic_Passanger_List” that consists of 1309 observations of passenger on Titanic. “Survived” is a predictor variable that describes whether or not a person survived the accident (1 – survived, 0 – died). In this sample,
…show more content…
The area under the curve is also increasing on the validation set and indicates that survived passengers 84.4671% more often are classified as belonging to class of survival than those passenger who have died. Finally, I will estimate such classification method as the Naïve Bayes Classifier that is based on determining probabilities for the outcomes. Implementing the full model,
I receive some evidence of overfitting for the full model due to the high dimensions of the model. As implemented in a previous method, I will create a reduced model with the variables that were identified as the most significant for the reduced logistic regression model:
Model Error Rate (training) Error Rate (validation)
Full model 25.5102
The area under the curve is also increasing on the validation set and indicates that survived passengers 84.4671% more often are classified as belonging to class of survival than those passenger who have died. Finally, I will estimate such classification method as the Naïve Bayes Classifier that is based on determining probabilities for the outcomes. Implementing the full model,
I receive some evidence of overfitting for the full model due to the high dimensions of the model. As implemented in a previous method, I will create a reduced model with the variables that were identified as the most significant for the reduced logistic regression model:
Model Error Rate (training) Error Rate (validation)
Full model 25.5102