ml_r/SPP_ML_1_Airfare Predictions.R at main · drdataSpp/ml_r · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
#Spp Machine Learning in R - Project 1
#Predicting Airfare On New Routes
#---------------------------------------------

airfares.df <- read.csv("Airfares.csv")
head(airfares.df)
summary(airfares.df)

#removing the first 4 variables, as it's not important for the model
airfares.df <- airfares.df[, c(5:18)]
head(airfares.df)
t(t(names(airfares.df)))

#Creating correlation matrix and scatter plot between FARE & Numerical predictors.
corr.airfares.df <- airfares.df[, c(1,2,5,6,7,8,9,12,13,14)]
corr.airfares <- cor(corr.airfares.df)

#Creating scatter plots to understand the relationship between the features and
#label.
library(gplots)
heatmap.2(corr.airfares, Rowv = FALSE, Colv = FALSE, dendrogram = "none",
          cellnote = round(corr.airfares,2),
          notecol = "black", key = FALSE, trace = 'none', margins = c(7,7))

library(ggplot2)
ggplot(data = corr.airfares.df) + geom_point(aes(x=COUPON, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. COUPON") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=NEW, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. NEW") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=HI, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. HI") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=S_INCOME, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. S_INCOME") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=E_INCOME, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. E_INCOME") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=S_POP, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. S_POP") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=E_POP, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. E_POP") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=DISTANCE, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. DISTANCE") +
  theme(plot.title = element_text(hjust = 0.5))

ggplot(data = corr.airfares.df) + geom_point(aes(x=PAX, y=FARE),
                                             color='darkred', size=2) +
  labs(title = "FARE Vs. PAX") +
  theme(plot.title = element_text(hjust = 0.5))

#Converting categorical variables to dummy/binary & creating partition sets.
airfares.df$VACATION <- ifelse(airfares.df$VACATION == 'Yes', 1, 0)
airfares.df$SW <- ifelse(airfares.df$SW == 'Yes', 1, 0)
airfares.df$SLOT <- ifelse(airfares.df$SLOT == 'Controlled', 1, 0)
airfares.df$GATE <- ifelse(airfares.df$VACATION == 'Constrained', 1, 0)

head(airfares.df)

#Partitioning the data set.
airfare.train.rows <- sample(rownames(airfares.df), dim(airfares.df)[1]*0.6)
airfares.train.df <- airfares.df[airfare.train.rows, ]

airfare.validate.rows <- sample(setdiff(rownames(airfares.df), airfare.train.rows))
airfare.validate.df <- airfares.df[airfare.validate.rows, ]

#Building a Regression model with all 13 predictors/features
airfares.reg <- lm(FARE ~., data = airfares.train.df)
options(scipen = 999, digits = 4)
summary(airfares.reg)

#Using exhaustive search to reduce the no.of predictors
library(leaps)
best.airfares.lm.model <- regsubsets(FARE ~ ., data = airfares.train.df, method = "exhaustive")
new.model.summary <- summary(best.airfares.lm.model)

new.model.summary$adjr2
new.model.summary$rsq
new.model.summary$cp
new.model.summary$which

#Building a new model with reduce features.
adjreg.airfares.model <- lm(FARE ~ VACATION + SW + HI + E_INCOME + S_POP + E_POP
                            + SLOT + DISTANCE + PAX, data = airfares.train.df)
summary(adjreg.airfares.model)
summary(airfares.reg)

#Comparing predictive accuracy of the 2 different models.
library(forecast)
lm.with.13.variables <- predict(airfares.reg, airfare.validate.df)
lm.with.9.variables <- predict(adjreg.airfares.model, airfare.validate.df)
options(scipen=999, digits = 4)

accuracy(lm.with.13.variables, airfare.validate.df$FARE)
accuracy(lm.with.9.variables, airfare.validate.df$FARE)

#Checking the new model with custom values.
custom.values1 <- data.frame("VACATION" = 0, "SW" = 0,
                             "HI" = 4442.141, "E_INCOME" = 27664,
                             "S_POP" = 4557004,"E_POP" = 3195503, "SLOT" = 0,
                             "DISTANCE"=1976, "PAX"=12782)

adjreg.model.prediction <- predict(adjreg.airfares.model, custom.values1)
adjreg.model.prediction

#If SW = 1
custom.values2 <- data.frame("VACATION" = 0, "SW" = 1,
                             "HI" = 4442.141, "E_INCOME" = 27664,
                             "S_POP" = 4557004,"E_POP" = 3195503, "SLOT" = 0,
                             "DISTANCE"=1976, "PAX"=12782)

adjreg.model.prediction2 <- predict(adjreg.airfares.model, custom.values2)
adjreg.model.prediction2

#Building a LM model before flights began to operate on a new route.
lm.model.before.airport <- regsubsets(FARE ~ VACATION + SW +
                                        S_INCOME + E_INCOME +
                                        S_POP + E_POP
                                      + SLOT + GATE + DISTANCE
                                      , data = airfares.train.df, method = "exhaustive")
no.airport.model.summary <- summary(lm.model.before.airport)

no.airport.model.summary$adjr2
no.airport.model.summary$cp
no.airport.model.summary$rsq
no.airport.model.summary$which

no.airport.airfares.model <- lm(FARE ~ VACATION + SW + E_INCOME
                                + S_POP + E_POP + SLOT + DISTANCE,
                                data = airfares.train.df)
summary(adjreg.airfares.model)
summary(no.airport.airfares.model)

#Trying custom values on the above
custom.values3 <- data.frame("VACATION" = 0, "SW" = 0,
                             "E_INCOME" = 27664, "S_POP" = 4557004,
                             "E_POP" = 3195503, "SLOT" = 0,
                             "DISTANCE"=1976)

no.airpot.predictions <- predict(no.airport.airfares.model, custom.values3)
no.airpot.predictions

summary(adjreg.airfares.model)
summary(no.airport.airfares.model)

lm.model.before.airpot <- predict(no.airport.airfares.model, airfare.validate.df)

accuracy(lm.model.before.airpot, airfare.validate.df$FARE)
accuracy(lm.with.9.variables, airfare.validate.df$FARE)
##------------------------------------------------------------------------------