Professional Documents
Culture Documents
Business Understanding: Introduction ......................................................................................................... 3 Business Objective ........................................................................................................................................ 3 Data Mining Objective .................................................................................................................................. 3 Data Set ......................................................................................................................................................... 3 Data Preparation ........................................................................................................................................... 3 Data Modeling............................................................................................................................................... 4 1. 2. 3. 4. 5. 6. Decision Tree (Binary) ....................................................................................................................... 5 Decision Tree (Three-way tree) ........................................................................................................ 6 Logistic Regression ............................................................................................................................ 6 Logistic Regression with Transform Variables .................................................................................. 7 Neural Networks ............................................................................................................................... 7 Neural Networks after transform variables and variable selection .................................................. 8
Evaluation ..................................................................................................................................................... 9 Profitability of a Proactive Retention Plan.................................................................................................. 10 The key variables predicting churn: ............................................................................................................ 11 Possible Incentives Offered ........................................................................................................................ 11 Test Measures ............................................................................................................................................. 12 Profitability Matrix .................................................................................................................................. 12 Net Additions minus Existing Customers ................................................................................................ 12
1|Page
List of Figures
Figure 1: Process flow Diagram..................................................................................................................... 4 Figure 2: 2-way decision tree that resulted from cell2cell data set ............................................................. 5 Figure 3: Variables in descending order of their importance helping in splits for 2-way Tree .................... 5 Figure 4: Result summary of 2-way decision tree......................................................................................... 5 Figure 5: Variables in descending order of their importance helping in splits for 3-way Tree .................... 6 Figure 6: Result summary of 3-way decision tree......................................................................................... 6 Figure 7: Result summary of Regression without transformation variables ................................................ 7 Figure 8: Result summary of Regression with transformation variables ...................................................... 7 Figure 9: Result summary of Neural Network without Variable transformation and selection ................... 8 Figure 10: Result summary of Network with Variable transformation and selection .................................. 8 Figure 11: Comparison of cumulative lift value for different techniques..................................................... 9
2|Page
Business Objective
Reduce churn for the company Improve profitability Identifying incentives offered to the customers with high risk of churning
Data Set
The given data set consists of 71,047 rows & containing a total of 78 variables (including a variable named CHURN, signifying whether the customer had left the company two months after observation). One of the variables named CALIBRAT was used to differentiate the validation dataset from training dataset. Training dataset contained data of 40,000 customers and validation dataset contained 31,047 customers.
Data Preparation
The dataset was divided in training and validation datasets, using CALIBRAT as the partition variable (value of 1 was used training and value of 0 was used for validation). Variable CHURN was set as target variable and some other variables (those not related to business objective) were rejected. Variable No. 22 26 30 3|Page Original Variable Churn CSA Customer Changed Variable Target Rejected Rejected
Rejected Rejected
Data Modeling
Total of 6 different models were used to predict the churn of customers. These models were:
Decision Tree (binary) Decision Tree (three way tree) Logistic Regression Logistic Regression with Transform Variables Neural Networks Neural Networks after transform variables and variable selection
SAS EM 4.3 was used to run these 6 models. Snapshot of the model is shown below.
4|Page
Figure 2: 2-way decision tree that resulted from cell2cell data set
Figure 3: Variables in descending order of their importance helping in splits for 2-way Tree
As can be seen from above figure, EQPDAYS Number of days of the current equipment (split at <302), MONTHS Months in service (Split < 11 months) are most important variables that resulted in splits.
5|Page
As can be seen from figure above, with number of leaves greater than 34, no significant split happens.
Figure 5: Variables in descending order of their importance helping in splits for 3-way Tree
The important variables are very similar to that used in 2-way decision tree. First four variables are same for 2-way and 3-way tree.
With number of leaves greater than 93, no significant split happens in 3-way tree.
3. Logistic Regression
Here, no transformation of variable was carried out.
6|Page
5. Neural Networks
Here, no transformation and variable selection was done. 7|Page
Figure 9: Result summary of Neural Network without Variable transformation and selection
Figure 10: Result summary of Network with Variable transformation and selection
As can be seen from figure above, the average error has come down a little bit and has less variation as compared to that of neural network model without transformation and variable selection.
8|Page
Evaluation
The lift chart of all the techniques is shown below:
Legend: Tree : 2-way decision tree Tree-2 : 3-way decision tree Neural : Neural Network without transformations Neural-2 : Neural Network with transformed variables and variable selection Reg : Regression without variable transformation Reg-2: Regression with variable transformations Table below shows cumulative lift values for different techniques at different percentiles (deciles).
Neural Networks Regression with Neural with Transformation Networks Transformation 1.511 2.102 1.691 1.42 1.79 1.593 1.336 1.604 1.489 1.318 1.457 1.416 1.297 1.411 1.346 1.21 1.314 1.284 1.159 1.217 1.194 1.127 1.133 1.131 1.076 1.071 1.062 1 1 1
Percentile 2-way Tree 3-way Tree Regression 10 1.931 1.904 1.658 20 1.706 1.699 1.642 30 1.451 1.573 1.5 40 1.353 1.444 1.4 50 1.294 1.325 1.343 60 1.254 1.263 1.253 70 1.236 1.192 1.203 80 1.127 1.131 1.131 90 1.056 1.07 1.067 100 1 1 1
Table 2. Showing cumulative lift values at different deciles for different techniques As can be seen from table, the performance of Regression with transformed variables is best among the different techniques used. The lift value at first deciles with Regression with transformed variables is
9|Page
2.102 which is highest. It is also evident from table that Regression with transformed variables performs best at most deciles and significantly better till top 3 deciles. The performance of Neural Network technique is worse among the 3 techniques. Transformation and variable selection does help in improving the performance of Neural Network technique. 3-way decision tree, though underperforms when compared to 2-way decision tree at first deciles, performs better than 2-way decision tree at other deciles. So, based on the table above, the best technique is Regression with transformed variables which gives a lift value of 2.102 which is well above the target of 1.75.
Using other monthly revenues we have: Monthly revenues 30 50 70 90 110 130 150 Lifetime Value of Customer 97.05876356 161.7646059 226.4704483 291.1762907 355.8821331 420.5879754 485.2938178 Maximum Incentive Cost 440.6614469 734.4357448 1028.210043 1321.984341 1615.758639 1909.532936 2203.307234
Variables like months, MOU, RETCALL, and Change in Revenues are Key performance indicator of telecom businesses. It indicates longer the time a customer remains with a Cell2Cell lesser will be its churn. Also number of calls made to the retention team has direct effect on the churn rate. Variable like MOU, RETCALL, RECCHARGE are actionable. Cell2Cell can offer incentive plans to improve minutes of usage as this will result in lower churn rate. Also Retention team can play an active role in improving EQPDAYS, MONTHS and RECCHARGE through loyalty and service programs.
From the model we got EQPDAYS as one of the primary factors for churn prediction. It makes business sense, as a customer who changes his old cell phone is likely to churn, because many mobile service providers gives new cell connection bundled with cell phone. From decision tree we can see that after 302 days customers change their handset. o So the company can come up with a plan of offering customers new cell phone (at a price slightly higher than the cost price) without changing their connection. This offered at 280 days mark. This offer will allow the company to keep hold of their customers (ensuring future revenue) without any incremental cost. From the decision tree it is visible that the probability of churn increases if EQPDAYS is more than 301 and CHANGEM value is less than -131% (i.e. decrease in minutes by 131%). Hence the company should try to increase the usage minutes by the customers having old cell phones. o This can be done by providing discount plan to these customers. Customers with CHANGEM value less than -131% along with MOU value less than 416mins and the customers with CHANGEM value more than -131% along with MOU value less than 1875mins have a very high tendency of churn. o Such customers can also be offered with discounted calling plans to increase their minutes of usage. In case of LCV customers, the company should offer little more incentives to such customers. This is because even if in the short term the company incurs more cost, but retaining such customers will increase cash flow to the company in the long run.
Test Measures
Profitability Matrix
Profitability in this case can be represented as shown below: Decision (from Predictive Modeling) Churn Not Churn Churn Actual Not Churn False Positive (Loss = Incentive Cost) True Negative True Positive (Profit = LCV - Incentive Cost) False Negative
12 | P a g e