Problem statement - How many should be allocated for marketing budget in socmed ads?
A set of data from company X advertisement expenditure in multiple social media platform being used to study and create a machine learning model to predict their unit sales and revenue. Assumption being made that the price for 1 unit is USD5000.
A regression model have been used and the result are as below:
1) Google
Model performance / R square value for google is 0.61
Intercept value for google : 7.03
Coefficients value for google : 0.047
Which means for every USD1000 spent in google , 47 unit being sold.
Expected Revenue: USD 35,167,721.40
2) Facebook
Model performance / R square value for facebook is 0.33
Intercept value for facebook : 9.31
Coefficients value for facebook : 0.202
Which means for every USD1000 spent in facebook , 202 unit being sold.
Expected Revenue: USD 46,578,440.05
3) Instagram
MModel performance / R square value for instagram is 0.052
Intercept value for instagram : 12.35
Coefficients value for instagram : 0.054
Which means for every USD1000 spent in instagram , 54 unit being sold.
Expected Revenue: USD 61,762,504.65
Call for Action!
Based on the obtain model, a multivariable model have been developed and have been used to test for the sales forecast.
Problem statement - Which package that have higher churn rate?
A set of data of telco company Y customers information being used to study and create a machine learning model to predict churn rate and also to advice on package offering for targeted customer classification.
A decision tree model have been developed and the result are as below with accuracy of 80%:
Conclusion
Based on the obtain model, a set of random data have been used to test the churn rate and provide an insight to the company for better package offers.
