Description
Acme Telephonica (AT) is a mobile phone operator that has customers across every state of the U.S.A.
AT struggles with customer churn prediction—customers leaving AT for other mobile phone operators.
AT hired us to take a new approach to reducing customer churn.
This case study is to develop a machine learning solution to this business problem.
AT did not approach us with a well-specified machine learning problem. Instead, the company approached us with a business problem—reducing customer churn.Our first goal is to convert this business problem into a machine learning problem and develop a concrete solution.
To evaluate the available data, we have the data definitions.
| Feature | Description |
| BILLAMOUNTCHANGEPCT | The percent by which the customer’s bill has changed from last |
| CALLMINUTESCHANGEPCT
AVGBILL AVGRECURRINGCHARGE AVGDROPPEDCALLS PEAKRATIOCHANGEPCT AVGRECEIVEDMINS AVGMINS AVGOVERBUNDLEMINS AVGROAMCALLS PEAKOFFPEAKRATIO |
month to thismonth
The percent by which the call minutes used by the customer has changed from last month to this month The average monthly bill amount The average monthly recurring charge paid by the customer The average number of customer calls dropped each month The percent by which the customer’s peak calls to off-peak calls ratio has changed from last month to this month The average number of calls received each month by the customer The average number of call minutes used by the customer each month The average number of out-of-bundle minutes used by the customer eachmonth The average number of roamingcallsmade by the customer each month The ratio between peak and off peak calls made by the customer thismonth |
| NEWFREQUENTNUMBERS | How many new numbers the customer is frequently calling  this month? |
| Feature | Description |
| CUSTOMERCARECALLS | The number of customer care calls made by the customer last |
| NUMRETENTIONCALLS
NUMRETENTIONOFFERS AGE CREDITRATING INCOME LIFETIME OCCUPATION REGIONTYPE HANDSETPRICE HANDSETAGE NUMHANDSETS SMARTPHONE CHURN |
month
The number of times the customer has been called by the retentionteam The number of retention offers the customer has accepted The customer’sage The customer’s credit rating The customer’s incomelevel The number of months the customer has been with AT The customer’soccupation The type of region the customer lives in The price of the customer’s current handset The age of the customer’s current handset The number of handsets the customer has had in the past 3 years Is the customer’s current handset a smart phone? The targetfeature |
Midterm analysis
- Understand the data. Note any issues with missing or damaged data and how you handle it. Comment on whether normalization of features would be helpful. One paragraph.
- Develop an analysis of the churn data using two different classifiers. First, justify why you chose each classifier. Provide a description of your classifiers. More than one paragraph.
- Compare the performance of the classifiers. Determine which is preferable and justify your decision. More than one paragraph.
- Extra credit. The company now tells you new information. Adjust your analysis for the case where giving service to a churner costs the company $700 and excluding a customer who is not a churner costs $100, while other cases are 0. Now optimize for the least cost. Describe how this changed your classifier performance.
