19828121_BUS5CA_Assignment 3

ASSIGNMENT-03 Churn Analysis SUBMITTED BY: RUHI MIGLANI 19828121

TABLE OF CONTENTS  Task 1  Task-2A  Task2B  Task 2.B1  Task 2 B2  Task2 B3  Task3

100 296 988 0 250 500 750 1000 Two year One year Month-to-month Contract T o t a l N u m b e r Contract 365 287 732 0 200 400 600 Yes No No internet service TechSupport T o t a l N u m b e r TechSupport  First if we look onto variable Partner , we have two categories Yes & No. Here, Yes means the count of people who are loyal to the company have partners and count value is 1053 whereas, no means these loyal customers don’t have partners and their count is 331. It can be easily analysed that in loyalty segment for variable partner those who have partners are more loyal in compare to those who don’t have partners. It is necessary for the company to target and focus those who have partners as they are giving more profit to the company.  Then, considering another variable named Gender, here also we have two categories MALE is 711and FEMALES is 673. With the plot it can easily observe that the Males customers are more loyal in compare to the females, so it is essential to focus on females as they are less loyal and giving less profit to the company whereas, it is essential to maintain the loyalty of Males customers as well’  Next variable is Payment Method with the graph we can observe that here major preference of the loyal customers is the Bank Transfer i.e. 525 following with Credit card and they are least preferring Electronic check and mailed check. But it is important for the bank to promote all services so for this an action needs o be taken.  Next is Internet services offered by the company to their customers. Here company is offering two types of services DSL and Fiber optic, if we see the graph of Internet services it can be easily observed that in the loyal customer segment the count of opting Fiber optic is high its count is 582 whereas, the count of opting DSL is less and here few loyal customers are not opting any of the services so it is essential to do something so that people will opt more services and for those who are not opting anything some initiatives needs to be taken so that they will encourage to opt any of the services offered.  Then next variable is Contract here the customers are more loyal towards the long term plans this states that people are happy with the long -term services so here we can do something for the contracts of short- term nature.  Another is Tech -Support those who opt tech support are more loyal and their count is 732 those who not -opt their count is less than it is essential to promote these services to those who didn’t opt that. All-in-all, to maintain the customer base it is essential to have a good relation with the customers who are generating profit for the company along with that it is also essential to focus on those who are not generating much profit if both will taken into account the profit and the customer base automatically improve. Churned (Segment-02)- These are those customers who are no more- happy with the services provided by the company and they don’t want to use more services of the company as they are using earlier. This data set is obtained after filtering the data and following results are obtained:

1200 669 0 250 500 750 1000 1250 No Yes Partner T o ta l N u m be r Partner 939 930 0 250 500 750 Female Male Gender T o ta l N u m b e r Gender 459 1297 113 0 500 1000 Fiber optic DSL No InternetService T o ta l N u m b e r InternetService 1655 166 48 0 500 1000 1500 Month-to-month One year Two year Contract T o ta l N u m b e r Contract 1446 113 310 0 500 1000 1500 No Yes No internet service TechSupport T ota l N um be r TechSupport  The firs variable taken here is partner , there are two categories in it YES and NO here also we can see that the count of NO is more in compare to YES means those who don’t have partners are less happy with the services provided by the company and they are less or not using the services or they are not going to use their services for longer period of time. So, it is essential for the company to focus on those who are not happy and comes under the category of churn so that some initiative can be taken to retain these customers.  Next, is Gender , in this the number of Females are more in compare to Males the difference is not big but still females churning count is more in respect to males so Females needs to be targeted in order to check why their churning count is high and then according to all such answers the action needs to be taken.  Then Payment Method , in this the count of Electronic transfer is more in respect to other mode of payments as its count is 1071, it is essential to know the reason behind this and also essential to focus on the all modes and to check the reasons why the churning ratio is more and why these services are least preferred.  Next is Internet Services if we see the graph of it we can easily observe that here trend is same the count of using Fiber optic is more and DSL is less here also we have few peoples who are not using anything means they are already churned .So it is essential to know the

company. Last is, Internet services here, DSL is most preferred its count is high in compare to Fiber optic but here also those who are with company are not using any of the services among two so we can do something for that so that they will also opt one of the service among those two. Conclusion By analysing all three different customers we can easily conclude that in each type variables are playing different roles but after analysing all the segments or profiles we can say that we need to look out those who don’t have partners, in gender we need to focus on females , in mode of payment all services needs to be taken into consideration whereas, in internet services we need to check why DSL is not famous in loyal customers and why there are few customers who are not using any of the services offered by the company. Task-02 A. Over- all churn rate: In this part it is asked to calculate over all churn rate and the churn rate of the categorical variable. The results are given below: No; 73.4 2% Yes; 26.5 8% Total Churn Rate(%) No Yes Here with the table and graph we can easily conclude that overall rate of churn is 26.57% means the ratio of the customers who are churned is 26.57% and rest 73.42% are the loyal ones. Group churn Rate Gender  Males= ( 930/3549) *100 = 26.20%  Females = (939)/3483) *100=26.95% Senior citizens  Yes= (476/1142) *100= 41.68%  No = (1394/5891} *100= 23.66% Dependent  Yes = (669/3393) *100 = 19.71%  No = (1200/3639) * 100 = 32.97% Phone Services  Yes = (1687/6340) *100 = 26.60%  No = (170/680) *100 = 25% Paperless Billing Yes = (1400/4168) *100 = 33.58% Row Labels Count of Churn Churn Rate (%) No 5163 73.42150171 Yes 1869 26.57849829 Grand Total 7032

No = (469/2864) * 100 = 16.37% Above mentioned are the calculation of Group churn. Task-02 B.1 why we apply standardization (z-score normalization) on the continuous variables? When all the Continuous variable are measured at different scales, we need to apply standardization (z-score normalization) this states that in the analysis there is no equal contribution of all the continuous variables. for instance, if we conduct the customer segmentation, in which we seek to group customers based on their diverse qualities. A' transaction amount ' variable, which is in b/w $10 and $10,000, has a greater weighting than the number of transactions between 0 and $20. The variable is called a transaction amount. It is therefore necessary to transform the data into weights which can be compared. It is a matter of rescaling an original parameter to the same scope and/or variability . B.2 What are the selected variables used for building the prediction models? With the above- mentioned diagram, we can easily identify the variables in order to build prediction models and this can be get after making changes in running the transformation node by making changes in the properties of Transformation node where we did Interval Inputs and in sample properties taken method as Random and Size as max. Selected variables are given below: B.3 What are the predictive performance of various models and how they rank against one another? Model 1 Neural network Model The neural network model helps in understanding the complex and non-linear relations in between the different variables.The first graph is the graph of Score Ranking Matrix with this we can observe that 3.7 is the best score whereas our score is 2.9and it is closer to the Best Commulative Scorewhich is a good indication. Then second graph is of Misclassification Rate if we see the value of Misclassification rate is 0.208 means there are less errors in the model.

With the above-mentioned fit statistics diagram, it is easy to check the performance of the model as it depicts all the values which is essential to check the overall performance of the model. Regression Model The Regression is used in order to get the existing relationship between the different variables,to get the clear picture here, also few properties are changed like in Regression Type we take Logistic Regression then in Link Function, Logit is there, Selection Model is Backwars and the criteria of selection is taken as Misclassification Validation. However, the results after running the Regression Node are given below: Here Blue Bar represents the Negative Relation and Red bars represents the Positive Relation, so here few variables are representing positive and few are representing negative relations.Then next is the graph of Score Ranking Matric with this we will able to get the commulative score the value for bes Commulative Score is approximately 3.7 whereas, our commulative score is approximately 2.9 this means the our commulative score is very close to the best commulative score this is a positive sign means model is good.

Then first graph is of Score Ranking Matric with this we will able to get the commulative score the value for best Commulative Score is approximately 3.7 whereas, our commulative score is approximately 2.8 this means the our commulative score is very close to the best commulative score this is a positive sign means model is good. Then second graph is of Iteration plot which gives out the value of Misclassification Rate its 0.211 Model Comparison & selection All three models are compared where the Selection statistics is taken as Misclassification Rate and below are the results:Here, Neural Network is selected as best model because it has less value of Misclassification in comparision with the other two models and less would be the misclassification rate more would be the model is accurate or good.Then to clear this thing another chart i.e ROC chart is taken into consideration:

If we see this here the misclassification rate for Neural Network is less in compare to all other models weather in case of training or validation. Confusion Matrix: Model Partition of data True + True - False+ False- Total Accuracy Decision Tree Training 650 3251 362 657 4920 0.79 Validation 258 1410 140 304 2112 0.78 Neural Network Training 690 3264 349 617 4920 0.80 Validation 274 1397 153 288 2112 0.79 Regression Training 668 3272 341 639 4920 0.80 Validation 264 1402 148 298 2112 0.78 TP + TN / TP + TN + FP + FN. With the help of confusion matrix, the accuracy of the model is calculated and we can clearly observe that The validation accuracy of Neural Network is high i.e. 0.79 in compare to other two models this means this is the best model among all. Conclusion After analysing all three graph the below mention conclusions are drawn:

 The value of Average square error is less of Neural Network model in compare to all other models.  The ROC curve of Neural network is closer to the top left corner menas the performance of this model is good in compare to other models.  If we focus on the lift score the lift score of Neural Network model is high in compare to other models.  Value of accuracy of Neural network is more with value 0.79, means in compare to other this model is more accurate and reliable. All-in-all the Neural Network is best in compare to Decision tree model and Regression model Task-03 In this part it is asked to give recommendations based on the analysis done in previous task(1,2).so, recommendations are mainly based on the variables, means which is the important variable for the Alpha Telecom so that they will focus on that first in order to retained their customers for longer period of time.After deep analysis the major insights drawn are, In case of loyal customers they are providing the profit to the company means they are happy with the services offered and they are happily using the services offered by the company.Based on the insights drawn and our understanding below mentioned are the few recommendations which Alpha Telco needs to take into consideration in order reduce the churn rate of their customers:  The churn out ratio of the people who don’t have partners are high, means those who are single are churrning out more, so it is essential for the company to introduce some extra benefits and discounts to those who are single, like 10% off on fees of the service ,as retaining them is essential for the company in order to reduce the churn out ratio. And to retained those who have partners and loyal family plans are the best tactics but for singles who are churrning at a good pace some initial discounts and schemes or some lomg term benefits and dioscounts in the price of service if they use that service for longer period this will attract them and helped to stay with the company for ;longer period of time.  Then the churn out ratio is high for the contact of 2 year, so for this some attractive long term benefits needs to offer to the customer so they finf fruitful to invest for long term contract and this will reduce the churn ratio for long term contracts. Here company can offere few monetary discounts for those who adopt 2 year contract or some rewards points to use later this will attract people to invest more in longer period of contracts.  The churn ratio is high for the customers in case of Electronic check whereas, there is less count of churn of all other modes of payment so it essential for the company to provide some discount vouchers for opting Electronic checks this will attracts new customers also this will help to retain the existing customers as every one wants to opt those services which give monetary benefit in return. Also it is good to opt because this will reduce the handling of

opting mofre number of services and accordingly some iniotiatives needs to be taken so that they will use the maximum services of the company and will remain loyal for the longer period of time.  Integrated data management is essential to get rid from the churn out ratio as this will help to understand the activities of the customer and accordingly company can provide special discounts and offers to the interface of the customer.  Real time communiaction or personal contact with the customer is essential and accordingly we can offer value services like extra minutes talk time just like vodaphone is doing in their pack they are providing 1000 free international minutes.These type of offeres attract more customers and also help in retaining the existing base of customer.  It is essential to track the the performance of the services offered this will help to know the lacking areas and help to know what initiative is required to tackle the problem. All-in -all the churning ratio of the customers is 27%, so it is essential for the company to take all above mentioned initiatives in order to reduce this percentage and to convert them as loyal or potential customers, because for the success of the firm it is essential to satisfy the customers through their services better would be the service more would be the potential customer along with that customer frelationship management is important now-a-days only with this company can gety to know the demand and preference of the customer and accordingly they can deliever the services to them this is good for both company and the customers.