Problem Statement - Build a model to predict the conversion rate and provide recommendations to better the conversion rate. Data utilized - Conversion Rate table gives information about the users and whether they converted. Technology Utilised - Python Models Utilised- DecisionTree, LogisticRegression, RandomForest, AdaBoost, Bagging Evaluation Metric - F-Score [Focused on], Accuracy[Just for contrast] Best Performing Model - AdaBoost with 98.6% accuracy and 0.76 F-Score. Recommendations -
- The most crucial aspect in determining whether the user will convert is whether they are repeat users. Repeat users are likely not to convert and new users are more likely to convert. It would be essential to attract more new users to the website in order to increase the conversion rate. This should be supplemented by coming up with ideas to get return users to convert.
- The US provides the maximum number of converted users. Targeting the users based in the US would be fruitful in efforts to increase conversion rate.
- SEO is the most popular source for users that convert. It would be sensible to invest in potentially scaling it even further.
- Users in China are least likely to convert even though it has the second-largest user base. This seems to be a hindrance when trying to increase conversion rate. Users in China that are likely to convert are routed to the website via SEO. Therefore, focusing on optimizing this source especially in China would help increase the conversion rate.
- The younger user base is more likely to convert when compared to the older user base. Marketing should target the older user base to increase the popularity of the website in that age group and also increase their engagement. This would help increase the likelihood of them converting as increased engagement positively affects the chances of conversion.