Data background
A telephone company is interested in determining which customer characteristics are useful for predicting churn, customers who will leave their service.
The data set is Churn . The fields are as follows:
State |
discrete. |
account length |
continuous. |
area code |
continuous. |
phone number |
discrete. |
international plan |
discrete. |
voice mail plan |
discrete. |
number vmail messages |
continuous. |
total day minutes |
continuous. |
total day calls |
continuous. |
total day charge |
continuous. |
total eve minutes |
continuous. |
total eve calls |
continuous. |
total eve charge |
continuous. |
total night minutes |
continuous. |
total night calls |
continuous. |
total night charge |
continuous. |
total intl minutes |
continuous. |
total intl calls |
continuous. |
total intl charge |
continuous. |
number customer service calls |
continuous. |
churn |
Discrete |
Data Preparation and Exploration
1. 查看数据概览 2. 3. ## state account.length area.code phone.number 4. ## WV : 158 Min. : 1.0 Min. :408.0 327-1058: 1 5. ## MN : 125 1st Qu.: 73.0 1st Qu.:408.0 327-1319: 1 6. ## AL : 124 Median :100.0 Median :415.0 327-2040: 1 7. ## ID : 119 Mean :100.3 Mean :436.9 327-2475: 1 8. ## VA : 118 3rd Qu.:127.0 3rd Qu.:415.0 327-3053: 1 9. ## OH : 116 Max. :243.0 Max. :510.0 327-3587: 1 10. ## (Other):4240 (Other) :4994 11. ## international.plan voice.mail.plan number.vmail.messages 12. ## no :4527 no :3677 Min. : 0.000 13. ## yes: 473 yes:1323 1st Qu.: 0.000 14. ## Median : 0.000 15. ## Mean : 7.755 16. ## 3rd Qu.:17.000 17. ## Max. :52.000 18. ## 19. ## total.day.minutes total.day.calls total.day.charge total.eve.minutes 20. ## Min. : 0.0 Min. : 0 Min. : 0.00 Min. : 0.0 21. ## 1st Qu.:143.7 1st Qu.: 87 1st Qu.:24.43 1st Qu.:166.4 22. ## Median :180.1 Median :100 Median :30.62 Median :201.0 23. ## Mean :180.3 Mean :100 Mean :30.65 Mean :200.6 24. ## 3rd Qu.:216.2 3rd Qu.:113 3rd Qu.:36.75 3rd Qu.:234.1 25. ## Max. :351.5 Max. :165 Max. :59.76 Max. :363.7 26. ## 27. ## total.eve.calls total.eve.charge total.night.minutes total.night.calls 28. ## Min. : 0.0 Min. : 0.00 Min. : 0.0 Min. : 0.00 29. ## 1st Qu.: 87.0 1st Qu.:14.14 1st Qu.:166.9 1st Qu.: 87.00 30. ## Median :100.0 Median :17.09 Median :200.4 Median :100.00 31. ## Mean :100.2 Mean :17.05 Mean :200.4 Mean : 99.92 32. ## 3rd Qu.:114.0 3rd Qu.:19.90 3rd Qu.:234.7 3rd Qu.:113.00 33. ## Max. :170.0 Max. :30.91 Max. :395.0 Max. :175.00 34. ## 35. ## total.night.charge total.intl.minutes total.intl.calls total.intl.charge 36. ## Min. : 0.000 Min. : 0.00 Min. : 0.000 Min. :0.000 37. ## 1st Qu.: 7.510 1st Qu.: 8.50 1st Qu.: 3.000 1st Qu.:2.300 38. ## Median : 9.020 Median :10.30 Median : 4.000 Median :2.780 39. ## Mean : 9.018 Mean :10.26 Mean : 4.435 Mean :2.771 40. ## 3rd Qu.:10.560 3rd Qu.:12.00 3rd Qu.: 6.000 3rd Qu.:3.240 41. ## Max. :17.770 Max. :20.00 Max. :20.000 Max. :5.400 42. ## 43. ## number.customer.service.calls churn 44. ## Min. :0.00 False.:4293 45. ## 1st Qu.:1.00 True. : 707 46. ## Median :1.00 47. ## Mean :1.57 48. ## 3rd Qu.:2.00 49. ## Max. :9.00 50. ## 51.
从数据概览中我们可以发现没有缺失数据，同时可以发现电话号 地区代码是没有价值的变量，可以删去
Examine the variables graphically
从上面的结果中，我们可以看到churn为no的样本数目要远远大于churn为yes的样本，因此所有样本中churn占多数。
从上面的结果中，我们可以看到除了emailcode和areacode之外，其他数值变量近似符合正态分布。
1. ## account.length area.code number.vmail.messages total.day.minutes 2. ## Min. : 1.0 Min. :408.0 Min. : 0.000 Min. : 0.0 3. ## 1st Qu.: 73.0 1st Qu.:408.0 1st Qu.: 0.000 1st Qu.:143.7 4. ## Median :100.0 Median :415.0 Median : 0.000 Median :180.1 5. ## Mean :100.3 Mean :436.9 Mean : 7.755 Mean :180.3 6. ## 3rd Qu.:127.0 3rd Qu.:415.0 3rd Qu.:17.000 3rd Qu.:216.2 7. ## Max. :243.0 Max. :510.0 Max. :52.000 Max. :351.5 8. ## total.day.calls total.day.charge total.eve.minutes total.eve.calls 9. ## Min. : 0 Min. : 0.00 Min. : 0.0 Min. : 0.0 10. ## 1st Qu.: 87 1st Qu.:24.43 1st Qu.:166.4 1st Qu.: 87.0 11. ## Median :100 Median :30.62 Median :201.0 Median :100.0 12. ## Mean :100 Mean :30.65 Mean :200.6 Mean :100.2 13. ## 3rd Qu.:113 3rd Qu.:36.75 3rd Qu.:234.1 3rd Qu.:114.0 14. ## Max. :165 Max. :59.76 Max. :363.7 Max. :170.0 15. ## total.eve.charge total.night.minutes total.night.calls total.night.charge 16. ## Min. : 0.00 Min. : 0.0 Min. : 0.00 Min. : 0.000 17. ## 1st Qu.:14.14 1st Qu.:166.9 1st Qu.: 87.00 1st Qu.: 7.510 18. ## Median :17.09 Median :200.4 Median :100.00 Median : 9.020 19. ## Mean :17.05 Mean :200.4 Mean : 99.92 Mean : 9.018 20. ## 3rd Qu.:19.90 3rd Qu.:234.7 3rd Qu.:113.00 3rd Qu.:10.560 21. ## Max. :30.91 Max. :395.0 Max. :175.00 Max. :17.770 22. ## total.intl.minutes total.intl.calls total.intl.charge 23. ## Min. : 0.00 Min. : 0.000 Min. :0.000 24. ## 1st Qu.: 8.50 1st Qu.: 3.000 1st Qu.:2.300 25. ## Median :10.30 Median : 4.000 Median :2.780 26. ## Mean :10.26 Mean : 4.435 Mean :2.771 27. ## 3rd Qu.:12.00 3rd Qu.: 6.000 3rd Qu.:3.240 28. ## Max. :20.00 Max. :20.000 Max. :5.400 29. ## number.customer.service.calls 30. ## Min. :0.00 31. ## 1st Qu.:1.00 32. ## Median :1.00 33. ## Mean :1.57 34. ## 3rd Qu.:2.00 35. ## Max. :9.00
Relationships between variables
从结果中我们可以看到两者之间存在显著的正相关线性关系。
Using the statistics node, report
1. ## account.length area.code 2. ## account.length 1.0000000000 -0.018054187 3. ## area.code -0.0180541874 1.000000000 4. ## number.vmail.messages -0.0145746663 -0.003398983 5. ## total.day.minutes -0.0010174908 -0.019118245 6. ## total.day.calls 0.0282402279 -0.019313854 7. ## total.day.charge -0.0010191980 -0.019119256 8. ## total.eve.minutes -0.0095913331 0.007097877 9. ## total.eve.calls 0.0091425790 -0.012299947 10. ## total.eve.charge -0.0095873958 0.007114130 11. ## total.night.minutes 0.0006679112 0.002083626 12. ## total.night.calls -0.0078254785 0.014656846 13. ## total.night.charge 0.0006558937 0.002070264 14. ## total.intl.minutes 0.0012908394 -0.004153729 15. ## total.intl.calls 0.0142772733 -0.013623309 16. ## total.intl.charge 0.0012918112 -0.004219099 17. ## number.customer.service.calls -0.0014447918 0.020920513 18. ## number.vmail.messages total.day.minutes 19. ## account.length -0.0145746663 -0.001017491 20. ## area.code -0.0033989831 -0.019118245 21. ## number.vmail.messages 1.0000000000 0.005381376 22. ## total.day.minutes 0.0053813760 1.000000000 23. ## total.day.calls 0.0008831280 0.001935149 24. ## total.day.charge 0.0053767959 0.999999951 25. ## total.eve.minutes 0.0194901208 -0.010750427 26. ## total.eve.calls -0.0039543728 0.008128130 27. ## total.eve.charge 0.0194959757 -0.010760022 28. ## total.night.minutes 0.0055413838 0.011798660 29. ## total.night.calls 0.0026762202 0.004236100 30. ## total.night.charge 0.0055349281 0.011782533 31. ## total.intl.minutes 0.0024627018 -0.019485746 32. ## total.intl.calls 0.0001243302 -0.001303123 33. ## total.intl.charge 0.0025051773 -0.019414797 34. ## number.customer.service.calls -0.0070856427 0.002732576 35. ## total.day.calls total.day.charge 36. ## account.length 0.0282402279 -0.001019198 37. ## area.code -0.0193138545 -0.019119256 38. ## number.vmail.messages 0.0008831280 0.005376796 39. ## total.day.minutes 0.0019351487 0.999999951 40. ## total.day.calls 1.0000000000 0.001935884 41. ## total.day.charge 0.0019358844 1.000000000 42. ## total.eve.minutes -0.0006994115 -0.010747297 43. ## total.eve.calls 0.0037541787 0.008129319 44. ## total.eve.charge -0.0006952217 -0.010756893 45. ## total.night.minutes 0.0028044650 0.011801434 46. ## total.night.calls -0.0083083467 0.004234934 47. ## total.night.charge 0.0028018169 0.011785301 48. ## total.intl.minutes 0.0130972198 -0.019489700 49. ## total.intl.calls 0.0108928533 -0.001306635 50. ## total.intl.charge 0.0131613976 -0.019418755 51. ## number.customer.service.calls -0.0107394951 0.002726370 52. ## total.eve.minutes total.eve.calls 53. ## account.length -0.0095913331 0.009142579 54. ## area.code 0.0070978766 -0.012299947 55. ## number.vmail.messages 0.0194901208 -0.003954373 56. ## total.day.minutes -0.0107504274 0.008128130 57. ## total.day.calls -0.0006994115 0.003754179 58. ## total.day.charge -0.0107472968 0.008129319 59. ## total.eve.minutes 1.0000000000 0.002763019 60. ## total.eve.calls 0.0027630194 1.000000000 61. ## total.eve.charge 0.9999997749 0.002778097 62. ## total.night.minutes -0.0166391160 0.001781411 63. ## total.night.calls 0.0134202163 -0.013682341 64. ## total.night.charge -0.0166420421 0.001799380 65. ## total.intl.minutes 0.0001365487 -0.007458458 66. ## total.intl.calls 0.0083881559 0.005574500 67. ## total.intl.charge 0.0001593155 -0.007507151 68. ## number.customer.service.calls -0.0138234228 0.006234831 69. ## total.eve.charge total.night.minutes 70. ## account.length -0.0095873958 0.0006679112 71. ## area.code 0.0071141298 0.0020836263 72. ## number.vmail.messages 0.0194959757 0.0055413838 73. ## total.day.minutes -0.0107600217 0.0117986600 74. ## total.day.calls -0.0006952217 0.0028044650 75. ## total.day.charge -0.0107568931 0.0118014339 76. ## total.eve.minutes 0.9999997749 -0.0166391160 77. ## total.eve.calls 0.0027780971 0.0017814106 78. ## total.eve.charge 1.0000000000 -0.0166489191 79. ## total.night.minutes -0.0166489191 1.0000000000 80. ## total.night.calls 0.0134220174 0.0269718182 81. ## total.night.charge -0.0166518367 0.9999992072 82. ## total.intl.minutes 0.0001320238 -0.0067209669 83. ## total.intl.calls 0.0083930603 -0.0172140162 84. ## total.intl.charge 0.0001547783 -0.0066545873 85. ## number.customer.service.calls -0.0138363623 -0.0085325365
如果把高相关性的变量保存下来，可能会造成多重共线性问题，因此需要把高相关关系的变量删去。
Data Manipulation
从结果中可以看到，total.day.calls和total.day.charge之间存在一定的相关关系。
特别是voicemial为no的变量之间存在负相关关系。
Discretize (make categorical) a relevant numeric variable
对变量进行离散化
construct a distribution of the variable with a churn overlay
construct a histogram of the variable with a churn overlay
Find a pair of numeric variables which are interesting with respect to churn.
从结果中可以看到，total.day.calls和total.day.charge之间存在一定的相关关系。
Model Building
特别是churn为no的变量之间存在相关关系。
1. ## Estimate Std. Error t value Pr(>|t|) 2. ## (Intercept) 0.3082150 0.0735760 4.189 2.85e-05 *** 3. ## stateAL 0.0151188 0.0462343 0.327 0.743680 4. ## stateAR 0.0894792 0.0490897 1.823 0.068399 . 5. ## stateAZ 0.0329566 0.0494195 0.667 0.504883 6. ## stateCA 0.1951511 0.0567439 3.439 0.000588 *** 7. ## international.plan yes 0.3059341 0.0151677 20.170 < 2e-16 *** 8. ## voice.mail.plan yes -0.1375056 0.0337533 -4.074 4.70e-05 *** 9. ## number.vmail.messages 0.0017068 0.0010988 1.553 0.120402 10. ## total.day.minutes 0.3796323 0.2629027 1.444 0.148802 11. ## total.day.calls 0.0002191 0.0002235 0.981 0.326781 12. ## total.day.charge -2.2207671 1.5464583 -1.436 0.151056 13. ## total.eve.minutes 0.0288233 0.1307496 0.220 0.825533 14. ## total.eve.calls -0.0001585 0.0002238 -0.708 0.478915 15. ## total.eve.charge -0.3316041 1.5382391 -0.216 0.829329 16. ## total.night.minutes 0.0083224 0.0695916 0.120 0.904814 17. ## total.night.calls -0.0001824 0.0002225 -0.820 0.412290 18. ## total.night.charge -0.1760782 1.5464674 -0.114 0.909355 19. ## total.intl.minutes -0.0104679 0.4192270 -0.025 0.980080 20. ## total.intl.calls -0.0063448 0.0018062 -3.513 0.000447 *** 21. ## total.intl.charge 0.0676460 1.5528267 0.044 0.965254 22. ## number.customer.service.calls 0.0566474 0.0033945 16.688 < 2e-16 *** 23. ## total.day.minutes1medium 0.0502681 0.0160228 3.137 0.001715 ** 24. ## total.day.minutes1short 0.2404020 0.0322293 7.459 1.02e-13 ***
从结果中看，我们可以发现 state total.intl.calls 、number.customer.service.calls 、 total.day.minutes1medium 、 total.day.minutes1short 的变量有重要的影响。
Use K-Nearest-Neighbors (K-NN) algorithm to develop a model for predicting Churn
1. ## Direction.2005 2. ## knn.pred 1 2 3. ## 1 760 97 4. ## 2 100 43 5. 6. 7. [1] 0.803
混淆矩阵（英语：confusion matrix）是可视化工具，特别用于监督学习，在无监督学习一般叫做匹配矩阵。矩阵的每一列代表一个类的实例预测，而每一行表示一个实际的类的实例。
1. ## Direction.2005 2. ## knn.pred 1 2 3. ## 1 827 104 4. ## 2 33 36 5. 6. 7. 8. [1] 0.863
从测试集的结果，我们可以看到准确度达到86%。
Findings
我们可以发现 ，total.day.calls和total.day.charge之间存在一定的相关关系。特别是churn为no的变量之间存在相关关系。同时我们可以发现 state total.intl.calls 、number.customer.service.calls 、 total.day.minutes1medium、 total.day.minutes1short 的变量有重要的影响。同时我们可以发现，total.day.calls和total.day.charge之间存在一定的相关关系。最后从knn模型结果中，我们可以发现从训练集的结果中，我们可以看到准确度有80%，从测试集的结果，我们可以看到准确度达到86%。说明模型有很好的预测效果。
