-
Notifications
You must be signed in to change notification settings - Fork 1
/
assign2_lara2.R
282 lines (202 loc) · 11 KB
/
assign2_lara2.R
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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
library(tseries)
library(lubridate)
#READ DATA
radio <- read.csv("/Users/laradaoud/retail_assign2/Assignment 2/sales_radio_2.csv")
tv <- read.csv('/Users/laradaoud/retail_assign2/Assignment 2/sales_tv_2.csv')
plot(radio$sales,type = "o", col="blue",
ylab = "", xlab = "Time",main = "Sales")
sales_ts <- ts(radio$sales, frequency = 7)
plot(stl(sales_ts, 'periodic'))
#CHECK FOR UNIT ROOTS
adf.test(radio$sales)
pp.test(radio$sales)
kpss.test(radio$sales)
adf.test(radio$visits)
pp.test(radio$visits)
kpss.test(radio$visits)
#no unit root
#ADD LAGGED SALES AND VISITS
radio$lag_sales <- c(NA, head(radio$sales, -1))
radio$lag_visits <- c(NA, head(radio$visits, -1))
radio$lag2_sales <- c(NA, NA, head(radio$sales, -2))
radio$lag2_visits <- c(NA, NA, head(radio$visits, -2))
radio$lag3_sales <- c(NA, NA, NA, head(radio$sales, -3))
radio$lag3_visits <- c(NA, NA, NA, head(radio$visits, -3))
radio$BUSINESS_DATE <- format(as.Date(radio$BUSINESS_DATE, "%m/%d/%y"))
radio$month <- as.character(month(as.Date(radio$BUSINESS_DATE)))
tv$lag1_sales <- c(NA, head(tv$sales, -1))
tv$lag1_visits <- c(NA, head(tv$visits, -1))
tv$lag2_sales <- c(NA, NA, head(tv$sales, -2))
tv$lag2_visits <- c(NA, NA, head(tv$visits, -2))
tv$lag3_sales <- c(NA, NA, NA, head(tv$sales, -3))
tv$lag3_visits <- c(NA, NA, NA, head(tv$visits, -3))
tv$BUSINESS_DATE <- format(as.Date(tv$BUSINESS_DATE, "%m/%d/%y"))
tv$month <- as.character(month(as.Date(tv$BUSINESS_DATE)))
#AGGREGATE RADIO 15 SECOND ADS
radio["X15.secondi"] <- 0
radio$X15.secondi <- rowSums(radio[,grep("X15.secondi", names(radio))])
#AGGREGATE RADIO 20 SECOND ADS
radio["X20.secondi"] <- 0
radio$X20.secondi <- rowSums(radio[,grep("X20.secondi", names(radio))])
#AGGREGATE RADIO 30 SECOND ADS
radio["X30.secondi"] <- 0
radio$X30.secondi <- rowSums(radio[,grep("X30.secondi", names(radio))])
#AGGREGATE TV ADS
tv$X10.secondi <- 0
tv$X15.secondi <- 0
tv$X20.secondi <- 0
tv$X30.secondi <- 0
tv$X45.secondi <- 0
tv$X10.secondi <- rowSums(tv[ ,grep("X10.secondi", names(tv))])
tv$X15.secondi <- rowSums(tv[ ,grep("X15.secondi", names(tv))])
tv$X20.secondi <- rowSums(tv[ ,grep("X20.secondi", names(tv))])
tv$X30.secondi <- rowSums(tv[ ,grep("X30.secondi", names(tv))])
tv$X45.secondi <- rowSums(tv[ ,grep("X45.secondi", names(tv))])
#ADD LOG AND SQR FOR EACH FORMAT
radio$log.15.secondi <- log(radio$X15.secondi)
radio$sqr.15.secondi <- radio$X15.secondi^2
radio$log.20.secondi <- log(radio$X20.secondi)
radio$sqr.20.secondi <- radio$X20.secondi^2
radio$log.30.secondi <- log(radio$X30.secondi)
radio$sqr.30.secondi <- radio$X30.secondi^2
tv$log.10.secondi <- log(tv$X10.secondi)
tv$sqr.10.secondi <- tv$X10.secondi^2
tv$log.15.secondi <- log(tv$X15.secondi)
tv$sqr.15.secondi <- tv$X15.secondi^2
tv$log.20.secondi <- log(tv$X20.secondi)
tv$sqr.20.secondi <- tv$X20.secondi^2
tv$log.30.secondi <- log(tv$X30.secondi)
tv$sqr.30.secondi <- tv$X30.secondi^2
tv$log.45.secondi <- log(tv$X45.secondi)
tv$sqr.45.secondi <- tv$X45.secondi^2
#ADD ADSTOCK AND LAGGED ADDSTOCK
#for linear model
radio$X15.secondi.adstock.1 <- 0 + radio$X15.secondi
radio$X20.secondi.adstock.1 <- 0 + radio$X20.secondi
radio$X30.secondi.adstock.1 <- 0 + radio$X30.secondi
radio$lag.adstock.15.1 <- c(NA, head(radio$X15.secondi.adstock.1, -1))
radio$lag.adstock.20.1 <- c(NA, head(radio$X20.secondi.adstock.1, -1))
radio$lag.adstock.30.1 <- c(NA, head(radio$X30.secondi.adstock.1, -1))
tv$X10.secondi.adstock.1 <- 0 + tv$X10.secondi
tv$X15.secondi.adstock.1 <- 0 + tv$X15.secondi
tv$X20.secondi.adstock.1 <- 0 + tv$X20.secondi
tv$X30.secondi.adstock.1 <- 0 + tv$X30.secondi
tv$X40.secondi.adstock.1 <- 0 + tv$X45.secondi
tv$lag.adstock.10.1 <- c(NA, head(tv$X10.secondi.adstock.1, -1))
tv$lag.adstock.15.1 <- c(NA, head(tv$X15.secondi.adstock.1, -1))
tv$lag.adstock.20.1 <- c(NA, head(tv$X20.secondi.adstock.1, -1))
tv$lag.adstock.30.1 <- c(NA, head(tv$X30.secondi.adstock.1, -1))
tv$lag.adstock.45.1 <- c(NA, head(tv$X45.secondi.adstock.1, -1))
#for concave model
radio$X15.secondi.adstock.2 <- 0 + radio$X15.secondi
radio$X20.secondi.adstock.2 <- 0 + radio$X20.secondi
radio$X30.secondi.adstock.2 <- 0 + radio$X30.secondi
radio$lag.adstock.15.2 <- c(NA, head(radio$X15.secondi.adstock.2, -1))
radio$lag.adstock.20.2 <- c(NA, head(radio$X20.secondi.adstock.2, -1))
radio$lag.adstock.30.2 <- c(NA, head(radio$X30.secondi.adstock.2, -1))
#for concave qudratic model
radio$X15.secondi.adstock.3 <- 0 + radio$X15.secondi
radio$X20.secondi.adstock.3 <- 0 + radio$X20.secondi
radio$X30.secondi.adstock.3 <- 0 + radio$X30.secondi
radio$lag.adstock.15.3 <- c(NA, head(radio$X15.secondi.adstock.3, -1))
radio$lag.adstock.20.3 <- c(NA, head(radio$X20.secondi.adstock.3, -1))
radio$lag.adstock.30.3 <- c(NA, head(radio$X30.secondi.adstock.3, -1))
tv$X10.secondi.adstock.3 <- 0 + tv$X10.secondi
tv$X15.secondi.adstock.3 <- 0 + tv$X15.secondi
tv$X20.secondi.adstock.3 <- 0 + tv$X20.secondi
tv$X30.secondi.adstock.3 <- 0 + tv$X30.secondi
tv$X40.secondi.adstock.3 <- 0 + tv$X45.secondi
tv$lag.adstock.10.3 <- c(NA, head(tv$X10.secondi.adstock.3, -1))
tv$lag.adstock.15.3 <- c(NA, head(tv$X15.secondi.adstock.3, -1))
tv$lag.adstock.20.3 <- c(NA, head(tv$X20.secondi.adstock.3, -1))
tv$lag.adstock.30.3 <- c(NA, head(tv$X30.secondi.adstock.3, -1))
tv$lag.adstock.45.3 <- c(NA, head(tv$X45.secondi.adstock.3, -1))
#complete models
radio <- radio[!is.na(radio$lag3_sales),]
tv <- tv[!is.na(tv$lag3_sales),]
#RADIO
#Without carryover
#linear
radio.1.sales <- lm(sales ~ weekday + month + X15.secondi + X20.secondi + X30.secondi, data = radio)
radio.1.visits <- lm(visits ~ weekday + month + X15.secondi + X20.secondi + X30.secondi, data = radio)
#concave
radio.2.sales <- lm(sales ~ weekday + month + X15.secondi + X20.secondi + X30.secondi, data = radio)
radio.2.visits <- lm(visits ~ weekday + month + X15.secondi + X20.secondi + X30.secondi, data = radio)
#concave qudratic
radio.3.sales <- lm(sales ~ weekday + month + X15.secondi + X20.secondi + X30.secondi + sqr.15.secondi + sqr.20.secondi + sqr.30.secondi, data = radio)
radio.3.visits <- lm(visits ~ weekday + month + X15.secondi + X20.secondi + X30.secondi + + sqr.15.secondi + sqr.20.secondi + sqr.30.secondi, data = radio)
#With carryover
#linear
radio.4.sales <- lm(sales ~ lag_sales + weekday + month + X15.secondi + X20.secondi + X30.secondi, data = radio)
radio.4.visits <- lm(visits ~ lag_visits + weekday + month + X15.secondi + X20.secondi + X30.secondi, data = radio)
#concave
radio.5.sales <- lm(sales ~ lag_sales + lag2_sales + lag3_sales + weekday + month + X15.secondi + X20.secondi + X30.secondi, data = radio)
radio.5.visits <- lm(visits ~ lag_visits + lag2_visits + lag3_visits + weekday + month + log.15.secondi + log.20.secondi + log.30.secondi, data = radio)
#concave qudratic
summary(radio.6.sales <- lm(sales ~ lag_sales + lag2_sales + lag3_sales + weekday + month + X15.secondi + X20.secondi + X30.secondi + sqr.15.secondi + sqr.20.secondi + sqr.30.secondi, data = radio))
summary(radio.6.visits <- lm(visits ~ lag_visits + lag2_visits + lag3_visits + weekday + month +X15.secondi + X20.secondi + X30.secondi + sqr.15.secondi + sqr.20.secondi + sqr.30.secondi, data = radio))
#TV
#With carryover
#linear
summary(tv.4.sales <- lm(sales ~ lag1_sales + lag2_sales + lag3_sales + weekday + month + X10.secondi + X15.secondi + X20.secondi + X30.secondi + X45.secondi, data = tv))
#significant vars for sales: lag1 (t-1), lag2 (t-2); days: mon, sat, thurs, wed; month: feb, dec; X.20.secondi
summary(tv.4.visits <- lm(visits ~ lag1_visits + lag2_visits + lag3_visits + weekday + month + X10.secondi + X15.secondi + X20.secondi + X30.secondi + X45.secondi, data = tv))
#significant vars for visits: lag1 (t-1), lag2 (t-2); days: all; month: july, sep, dec; X.20.secondi, X15.secondi
#concave qudratic
summary(tv.6.sales <- lm(sales ~ lag1_sales + lag2_sales + lag3_sales + weekday + month + X10.secondi + X15.secondi + X20.secondi + X30.secondi + X45.secondi + sqr.10.secondi + sqr.15.secondi + sqr.20.secondi + sqr.30.secondi + sqr.45.secondi, data = tv))
#significant vars for sales: lag1 (t-1), lag2 (t-2); days: mon, sat, thurs, wed; month: feb, dec; X.20.secondi, sqr.20.secondi
summary(tv.6.visits <- lm(visits ~ lag1_visits + lag2_visits + lag3_visits + weekday + month + X10.secondi + X15.secondi + X20.secondi + X30.secondi + X45.secondi + sqr.10.secondi + sqr.15.secondi + sqr.20.secondi + sqr.30.secondi + sqr.45.secondi, data = tv))
#significant vars for visits: lag1 (t-1), lag2 (t-2); days: all; month: july, sep, dec; X.20.secondi, sqr.20.secondi
#Adstock
alpha.values <- c(0,0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9)
#model 1
grid.model1 <- list()
rmse <- function(error)
{
sqrt(mean(error^2))
}
i <- 1
for (value in alpha.values){
radio$X15.secondi.adstock.1 <- value * radio$lag.adstock.15.1 + radio$X15.secondi
radio$lag.adstock.15.1 <- c(NA, head(radio$X15.secondi.adstock.1, -1))
radio$X20.secondi.adstock.1 <- value * radio$lag.adstock.20.1 + radio$X20.secondi
radio$lag.adstock.20.1 <- c(NA, head(radio$X20.secondi.adstock.1, -1))
radio$X30.secondi.adstock.1 <- value * radio$lag.adstock.30.1 + radio$X30.secondi
radio$lag.adstock.30.1 <- c(NA, head(radio$X30.secondi.adstock.1, -1))
radio.1.2.adstock <- lm(sales ~ X15.secondi + X20.secondi + X30.secondi + X15.secondi.adstock.1 + X20.secondi.adstock.1 + X30.secondi.adstock.1, data = radio)
grid.model1[i] <- rmse(fitted(radio.1.2.adstock) - radio$sales)
i <- i+1
}
print(grid.model1)
#alpha 0.7
#model 2
grid.model2 <- list()
k <- 1
for (value in alpha.values){
radio$X15.secondi.adstock.2 <- value * radio$lag.adstock.15.2 + radio$X15.secondi
radio$lag.adstock.15.2 <- c(NA, head(radio$X15.secondi.adstock.2, -1))
radio$X20.secondi.adstock.2 <- value * radio$lag.adstock.20.2 + radio$X20.secondi
radio$lag.adstock.20.2 <- c(NA, head(radio$X20.secondi.adstock.2, -1))
radio$X30.secondi.adstock.2 <- value * radio$lag.adstock.30.2 + radio$X30.secondi
radio$lag.adstock.30.2 <- c(NA, head(radio$X30.secondi.adstock.2, -1))
radio.2.2.adstock <- lm(sales ~ log.15.secondi + log.20.secondi + log.30.secondi + X15.secondi.adstock.2 + X20.secondi.adstock.2 + X30.secondi.adstock.2, data = radio)
grid.model2[k] <- rmse(fitted(radio.2.2.adstock) - radio$sales)
k <- k+1
}
print(grid.model2)
#model 3
grid.model3 <- list()
j <- 1
for (value in alpha.values){
radio$X15.secondi.adstock.3 <- value * radio$lag.adstock.15.3 + radio$X15.secondi
radio$lag.adstock.15.3 <- c(NA, head(radio$X15.secondi.adstock.3, -1))
radio$X20.secondi.adstock.3 <- value * radio$lag.adstock.20.3 + radio$X20.secondi
radio$lag.adstock.20.3 <- c(NA, head(radio$X20.secondi.adstock.3, -1))
radio$X30.secondi.adstock.3 <- value * radio$lag.adstock.30.3 + radio$X30.secondi
radio$lag.adstock.30.3 <- c(NA, head(radio$X30.secondi.adstock.3, -1))
radio.3.2.adstock <- lm(sales ~ X15.secondi + X20.secondi + X30.secondi + sqr.15.secondi + sqr.20.secondi + sqr.30.secondi + X15.secondi.adstock.3 + X20.secondi.adstock.3 + X30.secondi.adstock.3, data = radio)
grid.model3[j] <- rmse(fitted(radio.3.2.adstock) - radio$sales)
j <- j+1
}
print(grid.model3)
#alhpa 0.7