Machine Learning - Customer Satisfaction Prediction
Model 1
Model 1: Logistic Regression
Using this model, we could get the coefficient of each factor. If the coefficient is negative, then improving that factor will not improve customer satisfaction, which means even if the customers are not satisfied with that factor, airline companies do not have to improve it. If the coefficient is positive, then the larger the value is, the more important it is for airline companies to improve that factor. It's because changing one unit of that factor could improve more customer satisfaction.
Separate the whole dataset into 6 subdatasets:
1) Business class & Long flight
2) Business class & Short flight
3) Eco plus class & Long flight
4) Eco plus class & Short flight
5) Eco class & Long flight
6) Eco class & Short flight
——————————————————————————————————————————————————————————————————————————————————
business class & long flight: 23830 samples
business class & short flight: 38330 samples
eco plus class & long flight: 488 samples
eco plus class & short flight: 8923 samples
eco class & long flight: 2563 samples
eco class & short flight: 55746 samples——————————————————————————————————————————————————————————————————————————————————
business class & long flight:
accuracy: 0.8964890194432787
precision recall f1-score support
0 0.80 0.75 0.78 1704
1 0.92 0.94 0.93 5445
accuracy 0.90 7149
macro avg 0.86 0.85 0.85 7149
weighted avg 0.89 0.90 0.90 7149parameters of the variables:
Seat Comfort: -0.27170151986045044
Time Convenience: 0.05032427608714323
Food & Drink: -0.22877304436012874
Gate Location: 0.27003568989034993
Inflight Wifi Service: 0.11426041261155742
Inflight Entertainment: 0.5865139826576083
Online Support: 0.29025289153079203
Ease Of Online Booking: 0.3179517922586324
On Board Service: 0.4735721970872351
Leg Room: 0.4241832735470077
Baggage Handling: 0.21886740210182465
Check-in Service: 0.5371673664554574
Cleanliness: 0.15766920131347115
Online Boarding: 0.4759343712688616factors having negative influence on passenger satisfaction:
Seat Comfort, Food & Drink
Conclusion:
Improving these factors will not improve customer satisfaction for business-class passenger having long flight. Considering the absolute value of the parameters, Inflight Entertainment and On Board Service have highest positive influence on customer satisfaction.——————————————————————————————————————————————————————————————————————————————————
business class & short flight:
accuracy: 0.8244195147404122
precision recall f1-score support
0 0.77 0.71 0.74 3998
1 0.85 0.88 0.87 7501
accuracy 0.82 11499
macro avg 0.81 0.80 0.80 11499
weighted avg 0.82 0.82 0.82 11499
parameters of the variables:
Seat Comfort: 0.03268704378839648
Time Convenience: -0.19824093769104942
Food & Drink: -0.042345439944953815
Gate Location: 0.1224663224115965
Inflight Wifi Service: -0.11405756235715651
Inflight Entertainment: 0.8919897813554252
Online Support: 0.15783953849204227
Ease Of Online Booking: 0.23453956061482695
On Board Service: 0.3255073027247243
Leg Room: 0.43606782383683623
Baggage Handling: 0.06558173091974656
Check-in Service: 0.27285880356287096
Cleanliness: 0.0027456636455543116
Online Boarding: 0.17166366249574985factors having negative influence on passenger satisfaction:
Time Convenience, Food & Drink, Inflight Wifi Service
Conclusion:
Improving these factors will not improve customer satisfaction for business-class passenger having short flight. Considering the absolute value of the parameters, Inflight Entertainment and Leg Room have highest positive influence on customer satisfaction.——————————————————————————————————————————————————————————————————————————————————
eco plus class & long flight:
accuracy: 0.891156462585034
precision recall f1-score support
0 0.90 0.97 0.94 119
1 0.83 0.54 0.65 28
accuracy 0.89 147
macro avg 0.87 0.76 0.79 147
weighted avg 0.89 0.89 0.88 147
parameters of the variables:
Seat Comfort: 0.9282359486880141
Time Convenience: -0.143001745910316
Food & Drink: -0.3937478870432294
Gate Location: 0.036985734226757584
Inflight Wifi Service: 0.2292189034813514
Inflight Entertainment: 0.21332055344928785
Online Support: 0.09141016580449653
Ease Of Online Booking: 0.018822425569141626
On Board Service: 0.12451995346456242
Leg Room: 0.0004446526670066196
Baggage Handling: -0.06060118758047381
Check-in Service: 0.11749095692630343
Cleanliness: -0.33095547386967333
Online Boarding: -0.3047262909403945factors having negative influence on passenger satisfaction:
Time Convenience, Food & Drink, Leg Room, Baggage Handling, Cleanliness, Online Boarding
Conclusion:
Improving these factors will not improve customer satisfaction for ecoplus-class passenger having long flight. Considering the absolute value of the parameters, Seat Comfort, Inflight Wifi Service, and Inflight Entertainment have highest positive influence on customer satisfaction.——————————————————————————————————————————————————————————————————————————————————
eco class & long flight:
accuracy: 0.9024707412223667
precision recall f1-score support
0 0.90 0.99 0.94 650
1 0.88 0.43 0.58 119
accuracy 0.90 769
macro avg 0.89 0.71 0.76 769
weighted avg 0.90 0.90 0.89 769
parameters of the variables:
Seat Comfort: 1.1827035001358428
Time Convenience: -0.09622230932267573
Food & Drink: -0.36968587222918
Gate Location: 0.05396189391300705
Inflight Wifi Service: 0.25514383793005924
Inflight Entertainment: 0.3898444523480091
Online Support: -0.013097585910497858
Ease Of Online Booking: -0.07136016379340993
On Board Service: 0.06922101047701557
Leg Room: 0.01147441688290121
Baggage Handling: 0.13312259741520793
Check-in Service: -0.021436063438705334
Cleanliness: -0.03786757684519453
Online Boarding: -0.16696395328393512factors having negative influence on passenger satisfaction:
Time Convenience, Food & Drink, Online Support, Ease Of Online Booking, Check-in Service, Cleanliness, Online Boarding
Conclusion:
Improving these factors will not improve customer satisfaction for eco-class passenger having long flight. Considering the absolute value of the parameters, Inflight Entertainment and Seat Comfort have highest positive influence on customer satisfaction.——————————————————————————————————————————————————————————————————————————————————
eco class & short flight:
accuracy: 0.8939248983496771
precision recall f1-score support
0 0.89 0.99 0.94 13557
1 0.90 0.49 0.64 3167
accuracy 0.89 16724
macro avg 0.90 0.74 0.79 16724
weighted avg 0.89 0.89 0.88 16724
parameters of the variables:
Seat Comfort: 1.3916256536043898
Time Convenience: -0.05890369432713001
Food & Drink: -0.2740047536632685
Gate Location: 0.0835197266089381
Inflight Wifi Service: -0.042687622586944966
Inflight Entertainment: 0.07452871039210476
Online Support: 0.00010923739269644444
Ease Of Online Booking: 0.24416408331787257
On Board Service: 0.09887483395301677
Leg Room: 0.0022120901855860937
Baggage Handling: -0.07285954877687027
Check-in Service: 0.10579706813116935
Cleanliness: -0.0835310524764779
Online Boarding: -0.04817812037790174
factors having negative influence on passenger satisfaction:
Time Convenience, Food & Drink, Inflight Wifi Service, Baggage Handling, Cleanliness, Online Boarding
Conclusion:
Improving these factors will not improve customer satisfaction for eco-class passenger having short flight. Considering the absolute value of the parameters, Seat Comfort and Ease Of Online Booking have highest positive influence on customer satisfaction.——————————————————————————————————————————————————————————————————————————————————
Conclusion from Logistic Regression:
The accuracy of the models is high, so the results are valid to make recommendations for airline companies. For all groups of customers, improving time convenience of the flight or food & drink will not improve customer satisfaction. For long-flight customers, improving inflight entertainment would be the most effective method to improve customer satisfaction. For ecoplus-class customers and eco-class customers, improving seat comfort would be the most effective method to improve customer satisfaction.