Here we are implementing our new skill of “Principal Component Analysis” to decide what features to use to engineer our predictions for Churn. Based on the plot, it looks as though the “elbow” occurs around 3. We see the most change around the 3rd component which means the variance of the components following become less significant.
# create some cool features. Make sure you add comments so I know what you are trying to accomplish!
banko <- bank %>%
mutate(Churn = Churn=="yes") %>%
dummy_cols(remove_selected_columns = T)
pr_churn <- prcomp(x = select(banko,-Churn), scale = T, center = T)
screeplot(pr_churn, type = "lines")
Here we are taking a look at the first two components in a biplot to see which variables are most significant. If you zoom in on the plot, it’s clear that gender seems clearly significant. We opted not to use gender in our engineering just to be safe, in case there could be ethical issues of discrimination based on gender. Since we’re analyzing ‘churn’ (whether a customer will leave), perhaps it would be ok? Provide insentives for people identifying as female to stay… But again, to play it safe, we’re leaving gender out of it. Aside from that we can see that on the other ends, variables like total transaction amount, total transaction count and income less thank $40k seem to stand out. We also chose not to use the “blue card” feature because we discussed how humans tend to be easier to predict based on their habits rather than demographics. The type of card they have is given, it’s not an action. However, how much they spend and how often they spend are habits and actions.
biplot(pr_churn)
biplot(pr_churn, choices = c(1,2))
print(ncol(bank)) # Only 5 features allowed for project 2! Not counting the dependent variable.
## [1] 20
pr_churn$rotation
## PC1 PC2 PC3
## Customer_Age -0.010813997 0.0757547936 -0.433019360
## Dependent_count 0.036907819 -0.0377317629 0.112719183
## Months_on_book -0.005699747 0.0713462802 -0.427853027
## Total_Relationship_Count -0.049689093 0.2001415587 -0.102343109
## Months_Inactive_12_mon -0.012256624 0.0265776928 -0.114586464
## Contacts_Count_12_mon 0.016040264 0.0909043204 -0.123300916
## Credit_Limit 0.413014059 -0.1184647817 -0.115635228
## Total_Revolving_Bal -0.034971577 -0.0010124303 0.216509638
## Avg_Open_To_Buy 0.416062573 -0.1183491437 -0.135021215
## Total_Amt_Chng_Q4_Q1 0.007772271 0.0026333373 0.174348201
## Total_Trans_Amt 0.100869534 -0.3699279277 0.289883802
## Total_Trans_Ct 0.044317988 -0.3805716862 0.303259402
## Total_Ct_Chng_Q4_Q1 -0.006905845 -0.0442536169 0.199821499
## Avg_Utilization_Ratio -0.268875657 0.0348746106 0.211170707
## Gender_F -0.355263483 -0.3216570306 -0.190336830
## Gender_M 0.355263483 0.3216570306 0.190336830
## Education_Level_College 0.004373639 0.0003329777 0.009695587
## Education_Level_Doctorate -0.005706353 -0.0036269274 -0.059093066
## Education_Level_Graduate -0.004115878 -0.0106117409 -0.017928153
## Education_Level_High School 0.003513914 0.0246190879 0.029106238
## Education_Level_Post-Graduate 0.005114198 0.0010958051 0.033172039
## Education_Level_Uneducated 0.001962101 -0.0105707065 -0.010319114
## Education_Level_Unknown -0.004074645 -0.0021657338 0.006458350
## Marital_Status_Divorced 0.007488559 -0.0248339495 0.033151739
## Marital_Status_Married -0.030806901 0.1738945600 -0.049881916
## Marital_Status_Single 0.017599981 -0.1478152617 0.008531959
## Marital_Status_Unknown 0.018419018 -0.0310883855 0.046009444
## Income_Category_$40K - $60K -0.049031167 0.0263133355 0.084110862
## Income_Category_$60K - $80K 0.139209300 0.1755923344 0.163215027
## Income_Category_$80K - $120K 0.213389458 0.1521782046 0.054029597
## Income_Category_$120K + 0.172705565 0.0621837568 -0.062625192
## Income_Category_Less than $40K -0.284878123 -0.2090096715 -0.089170360
## Income_Category_Unknown -0.046235765 -0.1327596236 -0.157001203
## Card_Category_Blue -0.257857151 0.3513883431 0.159134939
## Card_Category_Gold 0.113877056 -0.1500641047 -0.035769610
## Card_Category_Platinum 0.047305369 -0.0797260734 -0.029431473
## Card_Category_Silver 0.223199987 -0.3035764523 -0.153834541
## PC4 PC5 PC6
## Customer_Age 0.443499234 -0.266842219 0.003345576
## Dependent_count -0.081193438 0.109100516 0.097514325
## Months_on_book 0.439072869 -0.274734380 0.006509941
## Total_Relationship_Count -0.147836648 0.115075283 -0.293862484
## Months_Inactive_12_mon -0.001921710 -0.059288100 0.039294366
## Contacts_Count_12_mon -0.163137334 0.035553306 -0.047800261
## Credit_Limit -0.003755843 0.080990599 -0.052492125
## Total_Revolving_Bal 0.334555958 0.062542679 -0.459716873
## Avg_Open_To_Buy -0.033748268 0.075366592 -0.011267112
## Total_Amt_Chng_Q4_Q1 0.117915538 0.139278369 -0.163881656
## Total_Trans_Amt 0.262444253 -0.149981471 0.275895881
## Total_Trans_Ct 0.225273171 -0.183203010 0.245451826
## Total_Ct_Chng_Q4_Q1 0.182273387 0.067487809 -0.156661702
## Avg_Utilization_Ratio 0.257176086 0.025783374 -0.392825422
## Gender_F -0.067697739 0.113851814 0.023559659
## Gender_M 0.067697739 -0.113851814 -0.023559659
## Education_Level_College -0.024100754 0.014402905 0.022332369
## Education_Level_Doctorate -0.008318849 -0.025293962 0.046743197
## Education_Level_Graduate 0.016036359 0.047124906 -0.020742580
## Education_Level_High School 0.031582327 -0.018024493 -0.012086005
## Education_Level_Post-Graduate -0.010007543 0.026901088 -0.014524292
## Education_Level_Uneducated -0.018053749 -0.023400278 0.019641146
## Education_Level_Unknown -0.006932294 -0.031694456 -0.015945692
## Marital_Status_Divorced -0.072002664 0.002107928 0.030005373
## Marital_Status_Married 0.320106969 0.536301770 0.257409204
## Marital_Status_Single -0.266356421 -0.548948741 -0.292834274
## Marital_Status_Unknown -0.041657024 -0.001109133 0.025184945
## Income_Category_$40K - $60K 0.019636144 -0.044926389 -0.032068162
## Income_Category_$60K - $80K 0.025995880 -0.075116106 -0.014035928
## Income_Category_$80K - $120K 0.028692459 -0.018811267 0.010896864
## Income_Category_$120K + 0.038971138 -0.017014774 0.018740391
## Income_Category_Less than $40K -0.005177428 0.061605305 -0.075558313
## Income_Category_Unknown -0.109854134 0.079331308 0.142053374
## Card_Category_Blue -0.022247279 -0.209020498 0.278774578
## Card_Category_Gold 0.032067496 0.060138308 -0.086741539
## Card_Category_Platinum 0.023654477 0.017904628 0.003559924
## Card_Category_Silver 0.005039861 0.199955899 -0.268982568
## PC7 PC8 PC9
## Customer_Age -0.03336354 0.036362109 0.034871555
## Dependent_count 0.04081514 -0.044681645 -0.076070481
## Months_on_book -0.02940410 0.041067875 0.044670014
## Total_Relationship_Count -0.10860198 0.119361902 -0.002946349
## Months_Inactive_12_mon 0.04268363 -0.010710413 -0.042706255
## Contacts_Count_12_mon 0.05699464 -0.062727730 0.018184995
## Credit_Limit -0.08112809 0.054445251 -0.094530295
## Total_Revolving_Bal 0.02699446 -0.098915085 -0.078540266
## Avg_Open_To_Buy -0.08353113 0.063301638 -0.087469248
## Total_Amt_Chng_Q4_Q1 -0.35981356 0.384838953 0.056653514
## Total_Trans_Amt 0.03854855 -0.011702758 -0.010484331
## Total_Trans_Ct 0.01691837 0.009851743 -0.014565211
## Total_Ct_Chng_Q4_Q1 -0.35909704 0.395611732 0.033996975
## Avg_Utilization_Ratio 0.08621415 -0.121938678 -0.054130302
## Gender_F -0.07901224 0.042123098 0.019427834
## Gender_M 0.07901224 -0.042123098 -0.019427834
## Education_Level_College -0.09527680 -0.077086143 0.001828881
## Education_Level_Doctorate -0.06062482 -0.043666086 0.004242611
## Education_Level_Graduate 0.59788110 0.573743969 0.056813175
## Education_Level_High School -0.27902010 -0.314964137 0.041218032
## Education_Level_Post-Graduate -0.01778466 -0.089634858 0.072376302
## Education_Level_Uneducated -0.22224984 -0.056704358 -0.065748883
## Education_Level_Unknown -0.11544485 -0.188968978 -0.102969462
## Marital_Status_Divorced -0.02137404 0.012848235 0.033203519
## Marital_Status_Married 0.03995022 -0.070870852 0.006221875
## Marital_Status_Single -0.03607840 0.082890741 -0.014349066
## Marital_Status_Unknown 0.01246608 -0.032253389 -0.018302586
## Income_Category_$40K - $60K 0.08572588 -0.136438725 0.734077625
## Income_Category_$60K - $80K 0.08865545 -0.077504469 0.062202222
## Income_Category_$80K - $120K -0.01252546 0.108829575 -0.248231448
## Income_Category_$120K + -0.06450217 -0.020313601 -0.178653732
## Income_Category_Less than $40K 0.12865975 -0.067686509 -0.472721598
## Income_Category_Unknown -0.33139538 0.247399566 0.189816987
## Card_Category_Blue -0.11104710 0.134914611 -0.121375625
## Card_Category_Gold 0.03367939 -0.075707427 -0.013790893
## Card_Category_Platinum 0.01518487 0.013319676 -0.005984434
## Card_Category_Silver 0.10431618 -0.116668449 0.142082446
## PC10 PC11 PC12
## Customer_Age -0.023887483 -0.0375348089 -3.807232e-02
## Dependent_count 0.121909790 -0.0083398450 2.454218e-02
## Months_on_book -0.028999572 -0.0393308764 -4.986042e-02
## Total_Relationship_Count -0.018561680 -0.0235615607 -4.594087e-02
## Months_Inactive_12_mon -0.012540239 -0.0013706112 -1.011301e-01
## Contacts_Count_12_mon 0.003772090 0.0022648757 1.355260e-02
## Credit_Limit 0.133579110 0.0453632150 4.621317e-02
## Total_Revolving_Bal 0.372081021 0.0945571332 1.438395e-02
## Avg_Open_To_Buy 0.100193697 0.0368765666 4.491394e-02
## Total_Amt_Chng_Q4_Q1 -0.255030489 -0.0794835388 1.713273e-02
## Total_Trans_Amt -0.008204847 -0.0008039531 1.713456e-03
## Total_Trans_Ct 0.017181356 0.0029799793 -6.095045e-03
## Total_Ct_Chng_Q4_Q1 -0.217597424 -0.0772250040 -5.816621e-05
## Avg_Utilization_Ratio 0.216176106 0.0371202721 -9.359374e-03
## Gender_F 0.046024757 0.0273077248 3.044897e-02
## Gender_M -0.046024757 -0.0273077248 -3.044897e-02
## Education_Level_College 0.021306318 -0.1600044233 2.440679e-02
## Education_Level_Doctorate 0.039409654 -0.0151677577 -1.879981e-02
## Education_Level_Graduate 0.031491886 0.1485308774 -1.989404e-03
## Education_Level_High School -0.152550043 0.7052984729 -8.087349e-02
## Education_Level_Post-Graduate 0.023220419 -0.0983960551 -1.595789e-02
## Education_Level_Uneducated 0.210994839 -0.3764480062 -4.511000e-01
## Education_Level_Unknown -0.134369635 -0.4035127644 5.402949e-01
## Marital_Status_Divorced 0.217180866 -0.1681771923 -4.059675e-01
## Marital_Status_Married -0.105266686 0.0249520173 1.334302e-01
## Marital_Status_Single -0.091623344 0.0768742408 1.710238e-01
## Marital_Status_Unknown 0.154221170 -0.0226955219 -1.671455e-01
## Income_Category_$40K - $60K 0.108071982 -0.1284467039 1.743917e-01
## Income_Category_$60K - $80K -0.440442868 -0.0231697758 -3.730729e-01
## Income_Category_$80K - $120K 0.227239778 0.0498123505 1.830153e-01
## Income_Category_$120K + 0.188203522 0.0022667258 1.177377e-01
## Income_Category_Less than $40K -0.230667615 -0.0495692381 -6.830731e-02
## Income_Category_Unknown 0.290916777 0.1990088092 -3.473403e-03
## Card_Category_Blue 0.140714586 0.0456475063 6.310396e-02
## Card_Category_Gold -0.063947660 0.0636351666 2.788938e-02
## Card_Category_Platinum 0.035229132 -0.0066209603 2.193321e-03
## Card_Category_Silver -0.132865947 -0.0790316257 -8.337767e-02
## PC13 PC14 PC15
## Customer_Age -0.061163984 0.063176059 0.0064948431
## Dependent_count -0.026389358 0.116420168 0.0016736072
## Months_on_book -0.078965996 0.062895401 0.0028303063
## Total_Relationship_Count -0.022549609 0.047412293 0.0487814012
## Months_Inactive_12_mon -0.011089981 -0.026357799 -0.0495525093
## Contacts_Count_12_mon 0.074693689 -0.100881187 0.0724740773
## Credit_Limit 0.044046467 0.016109699 -0.0071859771
## Total_Revolving_Bal 0.153593279 0.070310645 -0.1033196081
## Avg_Open_To_Buy 0.030267456 0.009802910 0.0020782186
## Total_Amt_Chng_Q4_Q1 -0.075680809 -0.021373231 0.0510257380
## Total_Trans_Amt 0.017618170 -0.048788365 0.0259664638
## Total_Trans_Ct 0.016129173 -0.027736366 0.0052639284
## Total_Ct_Chng_Q4_Q1 -0.098040486 0.010668876 0.0567337499
## Avg_Utilization_Ratio 0.086800353 0.025473868 -0.0662771727
## Gender_F 0.001589306 0.020910812 -0.0004545378
## Gender_M -0.001589306 -0.020910812 0.0004545378
## Education_Level_College -0.316110988 -0.138549707 -0.7430049560
## Education_Level_Doctorate 0.136610188 0.068790444 -0.2285129758
## Education_Level_Graduate 0.056504862 0.042640927 0.0662496412
## Education_Level_High School -0.126385707 0.005381515 0.1905506292
## Education_Level_Post-Graduate -0.073813589 0.023485936 -0.0374425208
## Education_Level_Uneducated 0.171165227 -0.411251305 0.3207080577
## Education_Level_Unknown 0.130636883 0.408679029 0.1628156043
## Marital_Status_Divorced -0.111981873 0.270098061 -0.0798117535
## Marital_Status_Married 0.133768025 -0.204663293 -0.0275176301
## Marital_Status_Single 0.123027511 -0.210766377 -0.0596786780
## Marital_Status_Unknown -0.372168811 0.512694944 0.2433819710
## Income_Category_$40K - $60K -0.125992271 -0.137852787 0.1527616458
## Income_Category_$60K - $80K 0.288212031 0.255853247 -0.1892007085
## Income_Category_$80K - $120K -0.463352798 -0.165797095 0.0468355335
## Income_Category_$120K + 0.334388218 -0.023627185 0.0771084498
## Income_Category_Less than $40K -0.116041471 -0.071192617 0.0617878484
## Income_Category_Unknown 0.267954392 0.203973168 -0.1891345923
## Card_Category_Blue 0.040171121 0.009809729 0.0064236817
## Card_Category_Gold 0.080244792 0.065038445 -0.1153709541
## Card_Category_Platinum 0.166425328 0.062287897 -0.0058996636
## Card_Category_Silver -0.114485430 -0.053426125 0.0479749451
## PC16 PC17 PC18
## Customer_Age 0.009436493 0.0046260131 0.014744864
## Dependent_count 0.055360500 0.1146010390 0.154788127
## Months_on_book 0.003837683 0.0001919797 0.012622455
## Total_Relationship_Count -0.041637320 -0.0408216142 0.016862859
## Months_Inactive_12_mon 0.051869876 0.0824836888 -0.014483388
## Contacts_Count_12_mon -0.005302043 0.0407699668 -0.027262253
## Credit_Limit -0.035246631 -0.0090690595 0.008332373
## Total_Revolving_Bal 0.082337506 0.0314775853 -0.043426637
## Avg_Open_To_Buy -0.042620855 -0.0118891465 0.012223857
## Total_Amt_Chng_Q4_Q1 -0.114400317 -0.0377113911 0.058518589
## Total_Trans_Amt 0.005331819 -0.0250995871 -0.038653751
## Total_Trans_Ct -0.006877850 -0.0187466983 -0.034472390
## Total_Ct_Chng_Q4_Q1 -0.085508181 0.0072578249 0.063946518
## Avg_Utilization_Ratio 0.057254894 0.0230179777 -0.018297788
## Gender_F 0.001036912 0.0091368264 0.008978226
## Gender_M -0.001036912 -0.0091368264 -0.008978226
## Education_Level_College -0.176478050 0.2811322063 0.073969070
## Education_Level_Doctorate 0.078759206 -0.3233753320 0.141512911
## Education_Level_Graduate -0.060667954 0.0573014053 -0.037206940
## Education_Level_High School -0.130376033 -0.0114860519 -0.114380679
## Education_Level_Post-Graduate 0.142026280 -0.6208536012 0.475490015
## Education_Level_Uneducated 0.170822496 0.2232478403 0.024520732
## Education_Level_Unknown 0.070215927 0.0503370409 -0.285064713
## Marital_Status_Divorced -0.378507416 -0.3371320083 -0.484402888
## Marital_Status_Married 0.082503901 0.0075712712 -0.008370833
## Marital_Status_Single 0.086885500 0.0031526986 0.043437903
## Marital_Status_Unknown 0.059202335 0.3166257032 0.419121584
## Income_Category_$40K - $60K -0.176092973 0.0317640687 0.051307221
## Income_Category_$60K - $80K 0.210520849 0.0897480023 -0.016855971
## Income_Category_$80K - $120K 0.325132497 -0.1189858661 -0.180565700
## Income_Category_$120K + -0.612244381 0.0332329831 0.246594426
## Income_Category_Less than $40K -0.126658973 -0.0716059210 0.045418353
## Income_Category_Unknown 0.308307390 0.0805019013 -0.109834498
## Card_Category_Blue -0.056252178 -0.0156373224 0.018349150
## Card_Category_Gold -0.062146510 0.2343788364 0.163351732
## Card_Category_Platinum 0.073561700 -0.1988542231 0.211928028
## Card_Category_Silver 0.077026353 -0.0534683834 -0.138041607
## PC19 PC20 PC21
## Customer_Age 0.017839119 -0.007356427 -0.0655883337
## Dependent_count -0.058970425 0.074815354 -0.5079683342
## Months_on_book 0.021953939 0.015122347 -0.0684529664
## Total_Relationship_Count 0.013482133 -0.018687486 -0.1058093787
## Months_Inactive_12_mon 0.110595550 0.277638094 0.2158482520
## Contacts_Count_12_mon 0.065626205 0.146390167 0.5203790465
## Credit_Limit 0.029909019 0.004536405 -0.0190005846
## Total_Revolving_Bal 0.032376866 -0.017011422 0.0318339084
## Avg_Open_To_Buy 0.027000127 0.006060540 -0.0218505310
## Total_Amt_Chng_Q4_Q1 -0.023550215 0.084626380 0.1340621474
## Total_Trans_Amt 0.020907382 -0.006301436 0.0946399533
## Total_Trans_Ct 0.017355748 -0.024749760 0.0294383735
## Total_Ct_Chng_Q4_Q1 -0.058632129 0.041378237 -0.0209592084
## Avg_Utilization_Ratio 0.001294720 -0.003687047 0.0358501172
## Gender_F 0.022388941 0.016355059 -0.0360462392
## Gender_M -0.022388941 -0.016355059 0.0360462392
## Education_Level_College 0.112951731 -0.171014835 0.0957997392
## Education_Level_Doctorate -0.834392611 0.115103181 0.0619659894
## Education_Level_Graduate 0.022962014 -0.006898671 -0.0159230533
## Education_Level_High School 0.029925479 -0.011723531 0.0108438741
## Education_Level_Post-Graduate 0.431635241 0.069397212 -0.1529273384
## Education_Level_Uneducated -0.013896247 0.035752198 -0.0421521071
## Education_Level_Unknown 0.071911693 0.021057918 0.0281508297
## Marital_Status_Divorced 0.072306962 0.184913173 -0.0007937136
## Marital_Status_Married 0.009071313 -0.009299559 -0.0133691503
## Marital_Status_Single 0.011678124 -0.020447851 -0.0848604860
## Marital_Status_Unknown -0.111303917 -0.128983813 0.1843741114
## Income_Category_$40K - $60K -0.080570568 0.100573644 0.0059548893
## Income_Category_$60K - $80K 0.066297259 -0.061665088 -0.0584517467
## Income_Category_$80K - $120K -0.053735148 0.134427472 0.0434374965
## Income_Category_$120K + 0.001943384 -0.175477924 -0.0377880933
## Income_Category_Less than $40K -0.029349186 0.007605949 0.0248822493
## Income_Category_Unknown 0.129922247 -0.075514161 0.0006786573
## Card_Category_Blue 0.032977612 -0.026253661 -0.0132296939
## Card_Category_Gold 0.051975261 0.778517841 -0.0648669859
## Card_Category_Platinum 0.095282505 -0.092621087 0.5240837791
## Card_Category_Silver -0.079421413 -0.316836679 -0.0572455070
## PC22 PC23 PC24
## Customer_Age -0.0635120941 -0.010646478 0.1024420656
## Dependent_count 0.0200129806 -0.505646941 0.5719209556
## Months_on_book -0.0427591174 -0.026572200 0.1239861759
## Total_Relationship_Count -0.0365115735 0.030682392 -0.1217489426
## Months_Inactive_12_mon 0.7809691560 -0.401650465 -0.2026134929
## Contacts_Count_12_mon 0.0921055343 0.274494668 0.7165355288
## Credit_Limit 0.0011512860 0.004717453 -0.0022020304
## Total_Revolving_Bal 0.0515680517 0.026597577 0.0461699403
## Avg_Open_To_Buy -0.0034720736 0.002331968 -0.0063407408
## Total_Amt_Chng_Q4_Q1 0.0591480451 -0.017335674 0.0971976407
## Total_Trans_Amt 0.0575260139 0.102080591 0.0529468968
## Total_Trans_Ct 0.0509677727 0.082037901 0.0334475666
## Total_Ct_Chng_Q4_Q1 0.0516100296 -0.087256174 0.0987317233
## Avg_Utilization_Ratio 0.0454243011 -0.007964092 0.0236114964
## Gender_F -0.0153108649 -0.014645755 0.0291509344
## Gender_M 0.0153108649 0.014645755 -0.0291509344
## Education_Level_College -0.0518197633 -0.020232043 0.0339583145
## Education_Level_Doctorate 0.0949800591 0.052175125 0.0078561779
## Education_Level_Graduate -0.0075923961 0.005373742 -0.0001288012
## Education_Level_High School -0.0234451057 -0.093441864 0.0218381184
## Education_Level_Post-Graduate 0.1389208791 0.190469428 0.0328137326
## Education_Level_Uneducated -0.0915630923 -0.008386306 -0.0520310712
## Education_Level_Unknown 0.0299075626 -0.024647130 -0.0259471307
## Marital_Status_Divorced -0.0776791592 -0.065570087 0.0558229027
## Marital_Status_Married 0.0131178217 0.011870066 -0.0271034675
## Marital_Status_Single 0.0009131781 -0.052776874 0.0331149200
## Marital_Status_Unknown 0.0509373523 0.141246611 -0.0658485167
## Income_Category_$40K - $60K -0.0191400679 -0.107852750 -0.0234909204
## Income_Category_$60K - $80K -0.0263249858 0.009106933 0.0665576208
## Income_Category_$80K - $120K -0.0720407582 -0.031990320 -0.0157525933
## Income_Category_$120K + 0.1488695158 0.039554606 -0.0261164055
## Income_Category_Less than $40K -0.0334813027 0.006286889 0.0057202115
## Income_Category_Unknown 0.0632829483 0.115964878 -0.0139656819
## Card_Category_Blue 0.0184858720 0.035518394 0.0295226752
## Card_Category_Gold -0.2379112293 0.136589919 -0.1383986467
## Card_Category_Platinum -0.4312532374 -0.596399052 -0.0795837952
## Card_Category_Silver 0.1748757390 0.013125450 0.0475240334
## PC25 PC26 PC27
## Customer_Age -0.0204351527 0.0233675064 -0.018369091
## Dependent_count -0.0669788538 0.1036475314 -0.073862975
## Months_on_book -0.0366332380 0.0551419417 -0.015721295
## Total_Relationship_Count -0.8595542301 0.0983712090 0.012247228
## Months_Inactive_12_mon -0.0626953676 -0.0100011833 0.032628165
## Contacts_Count_12_mon -0.1133172891 -0.0839821092 -0.008385573
## Credit_Limit 0.0039256059 -0.0371181015 0.443276941
## Total_Revolving_Bal 0.0450753731 0.0044223470 0.250511060
## Avg_Open_To_Buy -0.0001162621 -0.0375067752 0.420725351
## Total_Amt_Chng_Q4_Q1 0.1879939478 0.6726205077 0.031686498
## Total_Trans_Amt -0.2186772541 0.0858158852 0.007445360
## Total_Trans_Ct -0.3587886997 0.0395023320 -0.049171353
## Total_Ct_Chng_Q4_Q1 0.0193112184 -0.7073090297 -0.062965780
## Avg_Utilization_Ratio 0.0301852735 0.0144075899 -0.122653345
## Gender_F 0.0022939619 -0.0077918529 0.219078537
## Gender_M -0.0022939619 0.0077918529 -0.219078537
## Education_Level_College -0.0514030788 0.0016143908 0.011132820
## Education_Level_Doctorate -0.0335030226 0.0015129714 0.020449022
## Education_Level_Graduate 0.0171609325 0.0028502686 0.002270450
## Education_Level_High School -0.0091883952 0.0006689748 -0.004009590
## Education_Level_Post-Graduate 0.0483801465 -0.0277039896 -0.019143581
## Education_Level_Uneducated 0.0310516158 0.0130128589 -0.003384217
## Education_Level_Unknown -0.0099627834 -0.0025037360 -0.004480537
## Marital_Status_Divorced 0.0458682609 -0.0111046265 -0.040580692
## Marital_Status_Married -0.0643513974 -0.0225223942 0.020842855
## Marital_Status_Single 0.0309011170 0.0213273882 -0.002536524
## Marital_Status_Unknown 0.0191911399 0.0142719735 0.005570973
## Income_Category_$40K - $60K -0.0312638370 -0.0186732490 0.209963928
## Income_Category_$60K - $80K -0.0073987912 -0.0092679589 0.209332420
## Income_Category_$80K - $120K -0.0087083205 0.0138837268 -0.129221741
## Income_Category_$120K + 0.0241946636 0.0108523102 -0.274286150
## Income_Category_Less than $40K 0.0125864991 -0.0077806924 0.105333142
## Income_Category_Unknown 0.0171079374 0.0200189219 -0.273598935
## Card_Category_Blue -0.0103199890 -0.0098764714 0.288894786
## Card_Category_Gold -0.0080312103 -0.0058678082 -0.170765294
## Card_Category_Platinum -0.0683504772 -0.0148153613 -0.073944557
## Card_Category_Silver 0.0285207852 0.0165751479 -0.225791406
## PC28 PC29 PC30
## Customer_Age 1.445410e-01 -0.6355155995 2.615395e-01
## Dependent_count -6.476359e-02 -0.0260860527 3.226946e-02
## Months_on_book -1.656818e-01 0.6364612109 -2.343061e-01
## Total_Relationship_Count -2.977542e-02 -0.0391477751 -2.596917e-02
## Months_Inactive_12_mon 3.588938e-03 -0.0196577624 -5.034916e-03
## Contacts_Count_12_mon 2.473470e-02 0.0015697110 3.087043e-03
## Credit_Limit 1.866537e-02 0.0797214793 2.068137e-01
## Total_Revolving_Bal -2.833686e-01 -0.1886168103 -4.088080e-01
## Avg_Open_To_Buy 4.406568e-02 0.0966143915 2.434203e-01
## Total_Amt_Chng_Q4_Q1 1.955655e-02 -0.0020066827 1.947556e-02
## Total_Trans_Amt -1.468859e-01 -0.1976948001 -1.408634e-01
## Total_Trans_Ct 1.681964e-01 0.1836003210 1.512690e-01
## Total_Ct_Chng_Q4_Q1 -1.341042e-02 -0.0025419594 -1.695475e-02
## Avg_Utilization_Ratio 3.421800e-01 0.2504618557 5.313303e-01
## Gender_F 3.252925e-01 -0.0077437800 -2.105569e-01
## Gender_M -3.252925e-01 0.0077437800 2.105569e-01
## Education_Level_College -7.878981e-03 -0.0062368678 2.840905e-03
## Education_Level_Doctorate -2.025649e-03 -0.0004670800 8.626003e-05
## Education_Level_Graduate 4.313880e-03 -0.0020435177 -2.336165e-03
## Education_Level_High School 2.297664e-03 0.0005813492 4.924471e-03
## Education_Level_Post-Graduate 1.901083e-03 -0.0031950825 -5.391779e-04
## Education_Level_Uneducated 1.852245e-03 0.0023957380 -6.210991e-03
## Education_Level_Unknown -3.365971e-03 0.0070980352 1.570699e-03
## Marital_Status_Divorced 4.772085e-03 -0.0120571068 5.712439e-04
## Marital_Status_Married 9.264557e-05 0.0166336837 1.343018e-02
## Marital_Status_Single -5.034506e-03 -0.0117525636 -1.411913e-02
## Marital_Status_Unknown 4.434467e-03 0.0022554545 1.476420e-04
## Income_Category_$40K - $60K -7.054356e-02 0.0040842006 7.252694e-02
## Income_Category_$60K - $80K 3.032757e-01 0.0087968907 -1.507979e-01
## Income_Category_$80K - $120K 3.349033e-01 -0.0088669199 -2.152424e-01
## Income_Category_$120K + 2.495079e-01 -0.0117007061 -1.974813e-01
## Income_Category_Less than $40K -3.927259e-01 -0.0028712229 2.152282e-01
## Income_Category_Unknown -2.393050e-01 0.0095155348 1.593219e-01
## Card_Category_Blue -2.416119e-02 0.0141091152 5.418619e-02
## Card_Category_Gold 7.107344e-03 -0.0112720742 -2.666170e-02
## Card_Category_Platinum 1.134564e-03 -0.0005754741 -1.900828e-02
## Card_Category_Silver 2.322294e-02 -0.0102485204 -4.385746e-02
## PC31 PC32 PC33
## Customer_Age -0.0987750530 -9.835442e-16 0.000000e+00
## Dependent_count 0.0151794297 1.799684e-16 1.538879e-16
## Months_on_book 0.0913927754 9.473709e-17 -3.153204e-16
## Total_Relationship_Count 0.0841916368 -2.278656e-16 -9.234225e-17
## Months_Inactive_12_mon -0.0106667566 1.616785e-16 -5.829859e-17
## Contacts_Count_12_mon -0.0241074922 -1.136391e-16 3.232837e-18
## Credit_Limit 0.0442614425 -2.275309e-02 6.511029e-02
## Total_Revolving_Bal -0.2278670827 2.040262e-03 -5.838416e-03
## Avg_Open_To_Buy 0.0646806187 2.275787e-02 -6.512396e-02
## Total_Amt_Chng_Q4_Q1 -0.0419554121 3.471735e-16 4.039525e-16
## Total_Trans_Amt 0.6575921876 1.908735e-16 -5.130727e-16
## Total_Trans_Ct -0.6190960363 -7.491127e-16 5.624206e-16
## Total_Ct_Chng_Q4_Q1 0.0265892400 5.895447e-17 -4.922604e-17
## Avg_Utilization_Ratio 0.2972450231 -2.232494e-16 2.104148e-16
## Gender_F 0.0395933681 -5.742222e-03 8.143242e-03
## Gender_M -0.0395933681 -5.742222e-03 8.143242e-03
## Education_Level_College 0.0125293675 -1.942735e-02 5.260492e-02
## Education_Level_Doctorate 0.0015230198 -1.335643e-02 3.616624e-02
## Education_Level_Graduate -0.0017171667 -2.991615e-02 8.100625e-02
## Education_Level_High School 0.0003882791 -2.584007e-02 6.996914e-02
## Education_Level_Post-Graduate -0.0055983669 -1.423847e-02 3.855461e-02
## Education_Level_Uneducated -0.0041144862 -2.291747e-02 6.205539e-02
## Education_Level_Unknown -0.0020939113 -2.311981e-02 6.260329e-02
## Marital_Status_Divorced 0.0071782566 -2.745413e-01 -1.836317e-01
## Marital_Status_Married -0.0226423944 -5.233902e-01 -3.500786e-01
## Marital_Status_Single 0.0185362598 -5.118312e-01 -3.423472e-01
## Marital_Status_Unknown 0.0014292971 -2.747101e-01 -1.837446e-01
## Income_Category_$40K - $60K -0.0079783205 2.360289e-01 -3.342751e-01
## Income_Category_$60K - $80K 0.0288883971 2.136931e-01 -3.026421e-01
## Income_Category_$80K - $120K 0.0224204022 2.218887e-01 -3.142490e-01
## Income_Category_$120K + 0.0246681275 1.597221e-01 -2.262058e-01
## Income_Category_Less than $40K -0.0490032017 2.954406e-01 -4.184167e-01
## Income_Category_Unknown 0.0065780371 1.934501e-01 -2.739730e-01
## Card_Category_Blue 0.0164132467 -6.262752e-02 1.662485e-01
## Card_Category_Gold -0.0084468490 -2.643017e-02 7.016048e-02
## Card_Category_Platinum -0.0082165822 -1.102703e-02 2.927190e-02
## Card_Category_Silver -0.0126315772 -5.653014e-02 1.500627e-01
## PC34 PC35 PC36
## Customer_Age 0.000000e+00 0.000000e+00 0.000000e+00
## Dependent_count -2.480264e-16 -1.191855e-16 -2.143204e-16
## Months_on_book 1.061491e-16 5.956080e-17 -7.922438e-18
## Total_Relationship_Count 3.528702e-17 -1.381530e-16 -3.714609e-17
## Months_Inactive_12_mon -5.262477e-17 -1.598883e-16 1.316486e-16
## Contacts_Count_12_mon 6.862563e-17 1.173288e-16 -5.213234e-17
## Credit_Limit 6.020853e-02 6.210231e-01 -2.713703e-01
## Total_Revolving_Bal -5.398878e-03 -5.568692e-02 2.433368e-02
## Avg_Open_To_Buy -6.022118e-02 -6.211535e-01 2.714273e-01
## Total_Amt_Chng_Q4_Q1 -2.124368e-16 -1.565931e-16 -1.700961e-17
## Total_Trans_Amt -3.540520e-16 1.912286e-16 2.482142e-16
## Total_Trans_Ct 6.892297e-17 2.659650e-17 -1.591570e-16
## Total_Ct_Chng_Q4_Q1 3.102295e-17 2.024023e-17 5.042840e-17
## Avg_Utilization_Ratio -3.302595e-16 2.472077e-16 -3.264259e-16
## Gender_F -1.267843e-02 -1.170858e-01 1.653820e-01
## Gender_M -1.267843e-02 -1.170858e-01 1.653820e-01
## Education_Level_College 1.135369e-02 1.357383e-01 2.957994e-01
## Education_Level_Doctorate 7.805739e-03 9.332097e-02 2.033641e-01
## Education_Level_Graduate 1.748353e-02 2.090232e-01 4.555010e-01
## Education_Level_High School 1.510140e-02 1.805437e-01 3.934390e-01
## Education_Level_Post-Graduate 8.321219e-03 9.948376e-02 2.167939e-01
## Education_Level_Uneducated 1.339338e-02 1.601236e-01 3.489397e-01
## Education_Level_Unknown 1.351163e-02 1.615374e-01 3.520206e-01
## Marital_Status_Divorced -1.917182e-02 1.480013e-02 8.568116e-03
## Marital_Status_Married -3.654949e-02 2.821522e-02 1.633440e-02
## Marital_Status_Single -3.574230e-02 2.759209e-02 1.597366e-02
## Marital_Status_Unknown -1.918361e-02 1.480923e-02 8.573384e-03
## Income_Category_$40K - $60K -1.015261e-01 7.238445e-02 4.651016e-02
## Income_Category_$60K - $80K -9.191852e-02 6.553459e-02 4.210883e-02
## Income_Category_$80K - $120K -9.544378e-02 6.804797e-02 4.372379e-02
## Income_Category_$120K + -6.870328e-02 4.898296e-02 3.147368e-02
## Income_Category_Less than $40K -1.270816e-01 9.060461e-02 5.821741e-02
## Income_Category_Unknown -8.321116e-02 5.932655e-02 3.811989e-02
## Card_Category_Blue -6.786905e-01 3.611279e-02 -2.469439e-02
## Card_Category_Gold -2.864221e-01 1.524038e-02 -1.042157e-02
## Card_Category_Platinum -1.194992e-01 6.358492e-03 -4.348018e-03
## Card_Category_Silver -6.126136e-01 3.259687e-02 -2.229015e-02
## PC37
## Customer_Age 0.000000e+00
## Dependent_count -1.870400e-17
## Months_on_book -1.284589e-17
## Total_Relationship_Count 2.121911e-16
## Months_Inactive_12_mon -4.156710e-17
## Contacts_Count_12_mon -3.473322e-17
## Credit_Limit -1.737817e-01
## Total_Revolving_Bal 1.558294e-02
## Avg_Open_To_Buy 1.738182e-01
## Total_Amt_Chng_Q4_Q1 1.883076e-16
## Total_Trans_Amt 4.385981e-17
## Total_Trans_Ct -1.897359e-16
## Total_Ct_Chng_Q4_Q1 -4.768376e-17
## Avg_Utilization_Ratio -1.023139e-16
## Gender_F -6.772589e-01
## Gender_M -6.772589e-01
## Education_Level_College 4.935020e-02
## Education_Level_Doctorate 3.392860e-02
## Education_Level_Graduate 7.599431e-02
## Education_Level_High School 6.564008e-02
## Education_Level_Post-Graduate 3.616920e-02
## Education_Level_Uneducated 5.821596e-02
## Education_Level_Unknown 5.872996e-02
## Marital_Status_Divorced 1.227884e-05
## Marital_Status_Married 2.340859e-05
## Marital_Status_Single 2.289162e-05
## Marital_Status_Unknown 1.228639e-05
## Income_Category_$40K - $60K -5.276363e-03
## Income_Category_$60K - $80K -4.777052e-03
## Income_Category_$80K - $120K -4.960262e-03
## Income_Category_$120K + -3.570544e-03
## Income_Category_Less than $40K -6.604496e-03
## Income_Category_Unknown -4.324526e-03
## Card_Category_Blue 2.961726e-03
## Card_Category_Gold 1.249913e-03
## Card_Category_Platinum 5.214805e-04
## Card_Category_Silver 2.673374e-03
head(pr_churn$x)
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## [1,] 1.4991047 2.3816917 1.2061287 0.8965607 1.5324615 -0.7452821 -2.9891210
## [2,] -1.3634485 -0.6525061 1.5190587 1.4605792 0.9063029 -3.5810956 -4.9434915
## [3,] 0.9430949 2.2516219 2.3751092 2.2864274 2.4988199 -1.3576559 -4.4120845
## [4,] -2.5046912 -0.2083894 2.3492849 1.3856533 1.7282835 -3.1940012 -3.8275982
## [5,] 0.8411591 2.1366291 3.8216460 0.5588208 2.9101460 -1.3432339 -5.6052130
## [6,] -0.1145821 2.2172736 0.9176018 0.7205818 1.5060464 -0.7306351 0.3933491
## PC8 PC9 PC10 PC11 PC12 PC13
## [1,] 1.567650 0.5894549 -3.4650349 1.0845553 -0.9353536 0.03695702
## [2,] 8.011098 -0.1858311 -4.0692093 -1.0322340 0.3241670 -1.75684897
## [3,] 7.613060 0.4016536 -3.8732515 -1.0277678 0.8400688 -2.83108730
## [4,] 2.319265 -0.8169938 -1.2438990 1.0305130 -1.0588062 -2.66128310
## [5,] 5.198788 0.7194805 -4.8022156 -2.3977562 -1.9266328 -0.03912897
## [6,] 1.869738 2.3727741 -0.6822284 -0.3334129 0.7702467 -0.28100631
## PC14 PC15 PC16 PC17 PC18 PC19
## [1,] 0.5662105 0.4140922 -0.4327398 0.068779578 0.2660919 -0.1957914
## [2,] 0.2698623 1.1873703 -1.8234877 0.007137635 1.6569228 -1.0645323
## [3,] -0.4653215 1.1153690 -0.8125954 -0.648642503 0.6624920 -1.1526232
## [4,] 2.4200909 1.5256875 -0.8660267 1.375442440 2.2290439 -0.8703956
## [5,] -0.6327165 1.1519419 -0.3991568 0.492429597 1.1702627 -0.9454862
## [6,] -0.5265218 0.6013645 -0.8558381 0.015422879 0.2995518 -0.5281009
## PC20 PC21 PC22 PC23 PC24 PC25
## [1,] -0.10791987 -0.367440380 -0.8836780 -0.24567744 1.5008726 0.2390350
## [2,] 0.56404982 -1.931416490 -0.5501501 -2.04357111 2.7475257 0.4437751
## [3,] 0.70139821 -0.802962463 -0.8229375 -1.32181396 0.1654824 2.5271509
## [4,] 0.42961227 0.128351692 1.8633014 -2.20794121 0.2604017 2.2280835
## [5,] 0.09595182 -1.324047088 -0.8745157 -1.18859514 -0.3194066 1.7258709
## [6,] 0.05399539 0.007003597 -0.9675322 0.02289997 -0.0740152 1.8153898
## PC26 PC27 PC28 PC29 PC30 PC31
## [1,] -1.0546456 0.47296426 -0.004411802 0.38876167 -0.1271152 -0.004201762
## [2,] -6.1827731 -0.10603680 -0.758221526 0.10725396 -0.5180684 0.583206267
## [3,] 0.9837203 -1.48617141 0.194312158 -0.64272935 -0.2934515 0.480992995
## [4,] -2.7440947 0.12634688 -0.419938441 0.11856959 -0.3122945 0.732234493
## [5,] -0.9296224 -0.47207806 0.293953216 -0.77778703 0.0332695 0.208334166
## [6,] 1.2708393 -0.07502082 -0.948163483 0.09269952 0.2923200 0.166142469
## PC32 PC33 PC34 PC35 PC36
## [1,] -4.803438e-15 2.840301e-15 -1.899910e-15 -1.213371e-15 -1.353816e-15
## [2,] 5.993481e-15 2.632135e-15 -1.261532e-15 3.563024e-17 9.568352e-16
## [3,] 1.524833e-15 -2.683230e-16 -9.839758e-16 6.046195e-16 1.768686e-15
## [4,] 9.046594e-15 2.743157e-15 -1.513085e-16 -7.137703e-16 -6.807438e-16
## [5,] -2.138903e-15 4.394614e-15 -1.483576e-15 -1.712971e-15 -2.186484e-15
## [6,] -5.290797e-16 -1.045479e-15 -1.317043e-15 1.506676e-15 1.171941e-15
## PC37
## [1,] 4.519022e-16
## [2,] 1.357428e-15
## [3,] 1.431154e-15
## [4,] 1.136311e-16
## [5,] 3.165938e-16
## [6,] 8.578275e-16
Looking at the principal components side by side, we can see in PC2 our findings from the plot are verified with total transaction amount and total transaction count as the most significant variables.
rownames_to_column(as.data.frame(pr_churn$rotation)) %>%
select(1:5) %>%
filter(abs(PC1) >= 0.25 | abs(PC2) >= 0.25 | abs(PC3) >= 0.25 | abs(PC4) >= 0.25)
## rowname PC1 PC2 PC3
## 1 Customer_Age -0.010813997 0.07575479 -0.433019360
## 2 Months_on_book -0.005699747 0.07134628 -0.427853027
## 3 Credit_Limit 0.413014059 -0.11846478 -0.115635228
## 4 Total_Revolving_Bal -0.034971577 -0.00101243 0.216509638
## 5 Avg_Open_To_Buy 0.416062573 -0.11834914 -0.135021215
## 6 Total_Trans_Amt 0.100869534 -0.36992793 0.289883802
## 7 Total_Trans_Ct 0.044317988 -0.38057169 0.303259402
## 8 Avg_Utilization_Ratio -0.268875657 0.03487461 0.211170707
## 9 Gender_F -0.355263483 -0.32165703 -0.190336830
## 10 Gender_M 0.355263483 0.32165703 0.190336830
## 11 Marital_Status_Married -0.030806901 0.17389456 -0.049881916
## 12 Marital_Status_Single 0.017599981 -0.14781526 0.008531959
## 13 Income_Category_Less than $40K -0.284878123 -0.20900967 -0.089170360
## 14 Card_Category_Blue -0.257857151 0.35138834 0.159134939
## 15 Card_Category_Silver 0.223199987 -0.30357645 -0.153834541
## PC4
## 1 0.443499234
## 2 0.439072869
## 3 -0.003755843
## 4 0.334555958
## 5 -0.033748268
## 6 0.262444253
## 7 0.225273171
## 8 0.257176086
## 9 -0.067697739
## 10 0.067697739
## 11 0.320106969
## 12 -0.266356421
## 13 -0.005177428
## 14 -0.022247279
## 15 0.005039861
Here we renamed and graphed the factors based on our findings in the prior table. From this we determined PC2 (or “modest users”) to show a slightly higher likelihood to churn. This made intuitive sense to us, because the less people use their credit cards, the less incentive they have to stay. Low usage will generally increase the likelihood to leave because, well… they don’t use it.
prc <- bind_cols(select(banko,Churn),as.data.frame(pr_churn$x)) %>%
select(1:5) %>%
rename("rich_dudes"=PC1, "modest_users"=PC2, "young_spenders"=PC3, "veterans"=PC4)
prc %>%
select(Churn, rich_dudes,modest_users) %>%
pivot_longer(cols = -Churn,names_to = "component",values_to = "loading") %>%
ggplot(aes(loading, fill=Churn))+
geom_density(alpha=0.5)+
facet_grid(.~component)
prc %>%
select(Churn, young_spenders, modest_users) %>%
pivot_longer(cols = -Churn,names_to = "component",values_to = "loading") %>%
ggplot(aes(loading, fill=Churn))+
geom_density(alpha=0.5)+
facet_grid(.~component)
