Documente Academic
Documente Profesional
Documente Cultură
Problem statement:
To identify the churn rate of the company and identify measures to reduce the churn rate.
2. Multi-collinearity:
Collinearity Diagnosticsa
MDimension
Eigenvalu
Condition
Index
V
(Constant)
Account
VMail Message
Day Mins
Length
e
l
1
8.819
1.000
.00
.00
.00
.00
.718
3.504
.00
.00
.99
.00
.126
8.381
.00
.94
.00
.02
.080
10.500
.00
.01
.00
.83
5
1
6
.067
11.486
.00
.00
.00
.05
.061
12.065
.00
.00
.00
.00
.048
13.563
.00
.01
.00
.04
.038
15.160
.00
.00
.00
.00
.037
15.436
.00
.00
.00
.00
10
.006
37.460
1.00
.03
.00
.06
Descriptive Statistics
N
Range
Minimum
Maximum
Mean
Std. Deviation
Account Length
3333
242
243
101.06
39.822
VMail Message
3333
51
51
8.10
13.688
Day Mins
3333
350.8
.0
350.8
179.775
54.4674
Eve Mins
3333
363.7
.0
363.7
200.980
50.7138
Night Mins
3333
371.8
23.2
395.0
200.872
50.5738
Intl Mins
3333
20
20
10.24
2.792
Day Calls
3333
165
165
100.44
20.069
Day Charge
3333
59.64
.00
59.64
30.5623
9.25943
Eve Calls
3333
170
170
100.11
19.923
Eve Charge
3333
30.91
.00
30.91
17.0835
4.31067
Night Calls
3333
142
33
175
100.11
19.569
Night Charge
3333
16.73
1.04
17.77
9.0393
2.27587
Intl Charge
3333
5.4
.0
5.4
2.765
.7538
Area Code
3333
102
408
510
437.18
42.371
Valid N (listwise)
3333
Decision tree:
1)CHAID- Cross validation
Classification
Observed
Predicted
0
Percent Correct
2786
64
97.8%
252
231
47.8%
91.1%
8.9%
90.5%
Overall Percentage
Growing Method: CHAID
Dependent Variable: Churn
Risk
Method
Estimate
Std. Error
Resubstitution
.095
.005
Cross-Validation
.101
.005
2) CRT-Cross validation
Risk
Method
Estimate
Std. Error
Resubstitution
.092
.005
Cross-Validation
.112
.005
Classification
Observed
Predicted
0
Percent Correct
2785
65
97.7%
240
243
50.3%
90.8%
9.2%
90.8%
Overall Percentage
Growing Method: CRT
Dependent Variable: Churn
Test sample
Risk
Sample
Estimate
Std. Error
Training
.092
.006
Test
.101
.008
Classification
Sample
Observed
Predicted
0
Training
Percent Correct
1703
29
98.3%
157
131
45.5%
92.1%
7.9%
90.8%
1089
29
97.4%
104
91
46.7%
90.9%
9.1%
89.9%
Overall Percentage
Test
Overall Percentage
Risk
Sample
Estimate
Std. Error
Training
.104
.007
Test
.111
.009
Classification
Sample
Observed
Predicted
0
Training
1649
84
95.2%
129
179
58.1%
87.1%
12.9%
89.6%
1057
60
94.6%
83
92
52.6%
88.2%
11.8%
88.9%
Overall Percentage
Growing Method: CRT
Dependent Variable: Churn
5. Split validation(75.25)
CHAID
Percent Correct
Overall Percentage
Test
Risk
Sample
Estimate
Std. Error
Training
.111
.006
Test
.139
.012
Classification
Sample
Observed
Predicted
0
Training
2109
62
97.1%
220
139
38.7%
92.1%
7.9%
88.9%
654
25
96.3%
87
37
29.8%
92.3%
7.7%
86.1%
Overall Percentage
Growing Method: CHAID
Dependent Variable: Churn
6CRT
Percent Correct
Overall Percentage
Test
Risk
Sample
Estimate
Std. Error
Training
.105
.006
Test
.112
.011
Classification
Sample
Observed
Predicted
0
Training
Percent Correct
2009
102
95.2%
158
198
55.6%
87.8%
12.2%
89.5%
699
40
94.6%
57
70
55.1%
87.3%
12.7%
88.8%
Overall Percentage
Test
Overall Percentage
Growing Method: CRT
Dependent Variable: Churn
Regression:
Model Summary
Model
.257
R Square
.066
Adjusted R
Square
Estimate
.063
.341