Airline Revenue Prediction
Hawaiian Airlines ( HA )
1 - Look for direct correlation between specific ppm (price per mile) on selected ‘future-departing day’ and ‘Revenue’
2 - Box charts future-departing days price per mile variation
3 - Building revenue predictive model based on ppm for all Days
# Model A:
modelA <- lm( REVENUE ~ PPM_DAY0+PPM_DAY1+PPM_DAY2+PPM_DAY3+PPM_DAY4+PPM_DAY5+PPM_DAY6+PPM_DAY7+PPM_DAY15+PPM_DAY21+PPM_DAY30+PPM_DAY60+PPM_DAY90+PPM_DAY180 , data = dfa )
# Stats summary:
summary( modelA )##
## Call:
## lm(formula = REVENUE ~ PPM_DAY0 + PPM_DAY1 + PPM_DAY2 + PPM_DAY3 +
## PPM_DAY4 + PPM_DAY5 + PPM_DAY6 + PPM_DAY7 + PPM_DAY15 + PPM_DAY21 +
## PPM_DAY30 + PPM_DAY60 + PPM_DAY90 + PPM_DAY180, data = dfa)
##
## Residuals:
## Min 1Q Median 3Q Max
## -26.984 -7.954 1.286 7.703 21.131
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 337.43 52.09 6.478 7.63e-06 ***
## PPM_DAY0 -54.75 75.16 -0.728 0.4768
## PPM_DAY1 133.71 146.11 0.915 0.3737
## PPM_DAY2 -16.62 16.03 -1.037 0.3154
## PPM_DAY3 -132.51 71.60 -1.851 0.0828 .
## PPM_DAY4 231.54 256.50 0.903 0.3801
## PPM_DAY5 55.61 117.46 0.473 0.6423
## PPM_DAY6 -45.39 86.58 -0.524 0.6073
## PPM_DAY7 -251.30 517.33 -0.486 0.6337
## PPM_DAY15 1004.92 819.57 1.226 0.2379
## PPM_DAY21 -1039.69 733.35 -1.418 0.1755
## PPM_DAY30 119.98 322.32 0.372 0.7146
## PPM_DAY60 -108.51 194.64 -0.557 0.5849
## PPM_DAY90 -569.21 247.45 -2.300 0.0352 *
## PPM_DAY180 -22.95 43.71 -0.525 0.6067
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.21 on 16 degrees of freedom
## Multiple R-squared: 0.6245, Adjusted R-squared: 0.2959
## F-statistic: 1.9 on 14 and 16 DF, p-value: 0.1093dfa$modelA <- predict( modelA , dfa)
# Plot scattered chart: Model Prediction vs Actual
plot( data.frame(cbind(actuals=dfa$REVENUE, predicteds=dfa$modelA )) )
# Plot Hexa Binning chart: Model Prediction vs Actual
plot(hexbin(dfa$REVENUE, dfa$modelA , xbins=5) , main="Hexagonal Binning")
4 - Building revenue predictive model based on ppm for future depart days 0, 3, 6, 30 and 180
#modelB <- lm( REVENUE ~ PPM_DAY0+PPM_DAY3+PPM_DAY4+PPM_DAY15+PPM_DAY21+PPM_DAY30 , data = dfa )
# Model B:
modelB <- lm( REVENUE ~ PPM_DAY0+PPM_DAY3+PPM_DAY6+PPM_DAY30+PPM_DAY180 , data = dfa )
summary( modelB )##
## Call:
## lm(formula = REVENUE ~ PPM_DAY0 + PPM_DAY3 + PPM_DAY6 + PPM_DAY30 +
## PPM_DAY180, data = dfa)
##
## Residuals:
## Min 1Q Median 3Q Max
## -33.731 -10.107 -1.997 13.748 31.650
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 233.129 24.945 9.346 1.23e-09 ***
## PPM_DAY0 -30.007 55.170 -0.544 0.5913
## PPM_DAY3 -125.481 60.462 -2.075 0.0484 *
## PPM_DAY6 75.223 69.240 1.086 0.2877
## PPM_DAY30 3.275 127.044 0.026 0.9796
## PPM_DAY180 -6.268 25.069 -0.250 0.8046
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 18.17 on 25 degrees of freedom
## Multiple R-squared: 0.1621, Adjusted R-squared: -0.005498
## F-statistic: 0.9672 on 5 and 25 DF, p-value: 0.4566dfa$modelB <- predict( modelB , dfa)
# Plot scattered chart: Model Prediction vs Actual
plot( data.frame(cbind(actuals=dfa$REVENUE, predicteds=dfa$modelB )) )
# Plot Hexa Binning chart: Model Prediction vs Actual
plot(hexbin(dfa$REVENUE, dfa$modelB , xbins=5) , main="Hexagonal Binning")
5 - Line Chart showing ppm variation around the mean for each ppm range, revenue and both predictor models
Chart below is representing normalized (around their means) the fields: revenue, prediction, day 0 to 180