-1
$\begingroup$

Overview The data set I am working with considers a team that annually plays a 5-game home schedule. My goal is to identify the fans that are most likely to defect for the upcoming season, meaning not renew their season pass. This is my Y variable.

Description of Dataset The dataset contains 2 sheets. The first sheet has 17 x variables that influence whether or not someone will likely defect. (i.e., price of ticket, whether or not they used their ticket for games 1-5, age, income, and so on). This sheet is data taken from the previous year, so I also have the actual Y values (those who did and did not renew their season pass for the previous season)

The second sheet is where I will make my predictions for the upcoming season on whether or not someone will renew their season pass. The X variables are updated and already given for this current year. All I need to do is used the given data and predict the Y variable (whether someone will renew their season pass or default)

Suggestions Wanted I need advice on which classification method would provide me with the highest accuracy. These are the three models I know, so please only suggest one of the three: Decision tree, Random Forests, or Logistic Regression?

Improve this question
$\endgroup$
2
  • 2
    $\begingroup$ Try them all.. And see which performs better but after doing an EDA.... Try RF first and get feature Importances and then do feature engineering and rebuild... Also without seeing the data , it's impossible to say what will and what won't work... Doing some pre data analysis will help.. $\endgroup$
    – Aditya
    Mar 16, 2018 at 16:56
  • $\begingroup$ Here a link which will help you..saedsayad.com/data_mining_map.htm $\endgroup$
    – Aditya
    Mar 17, 2018 at 9:25

1 Answer 1

Reset to default
2
$\begingroup$

In Machine Learning, you can not say that a particular model will perform better than all other in every condition. Some times Logistic Regression may outperform a Random Forest/SVM, sometimes SVM can outperform Logistic Regression and RF. So, use every model you can for your purpose. There is no shortcut here.

Improve this answer
$\endgroup$
1

Not the answer you're looking for? Browse other questions tagged or ask your own question.