Kraken will use the Target you select to determine what kind of algorithms to use in the analysis process.
- If you select a Target that has only two unique values, Kraken will use algorithms that work best with binary classification problems. For example: customer retention (will my customer leave, yes/no), employee retention (will my employee leave, yes/no), etc.
- If you select a Target that is a string value and has more than two unique values, Kraken will use algorithms that work best with multi class classification problems. For example: campaign mix, product recommendation, up-sell opportunity, etc.
- If you select a Target that is a number and there are no dates in your dataset, Kraken will use algorithms that work best with regression problems. For example: how much will this customer purchase, what will be the value of this sale, etc.
- If you select a Target that is a number and there are dates in your dataset, Kraken will give you the option to create a time series model or a regression model. Time series work best for problems that forecast values over time. For example: sales forecast, inventory demand, etc.