-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathMNN_with_NARX_2.R
183 lines (137 loc) · 11.3 KB
/
MNN_with_NARX_2.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
library(readxl) # readxl package used to import excel files
library(dplyr)
library(neuralnet)
library(Metrics) # For MAE evaluation
uow_consumptions <- read_excel("D:\\Coding Area\\University Projects\\Courseworks\\R-Machine-Learning-MNN---MLCW-IIT\\uow_consumption.xlsx")
# Function to normalize data
normalize <- function(x) {
return((x - min(x)) / (max(x) - min(x)))
}
# Function to un-normalize data
unnormalize <- function(x, min, max) {
return( (max - min)*x + min )
}
# MAPE calculator
calculate_mape <- function(actual, predicted) {
return (mean(abs((actual-predicted)/actual)) * 100)
}
# RMSE calculator
calculate_rmse <- function(actual, predicted) {
sqrt(mean((actual - predicted)^2))
}
# ----------------- Data Pre-processing -----------------
# Get hourly data separately
uow_consumptions_inputs_18th <- uow_consumptions[-c(1, 3, 4)]
uow_consumptions_inputs_19th <- uow_consumptions[-c(1, 2, 4)]
uow_consumptions_inputs_20th <- uow_consumptions[-c(1: 3)]
# Normalize data sets
uow_consumptions_inputs_18th_norm <- as.data.frame(lapply(uow_consumptions_inputs_18th, normalize))
uow_consumptions_inputs_19th_norm <- as.data.frame(lapply(uow_consumptions_inputs_19th, normalize))
uow_consumptions_inputs_20th_norm <- as.data.frame(lapply(uow_consumptions_inputs_20th, normalize))
# Not normalized 20th hour data sets (For evaluations)
original_train_data_18th_hour <- uow_consumptions_inputs_18th[1: 380,]
original_test_data_18th_hour <- uow_consumptions_inputs_18th[381: 470,]
original_data_18th_hour_min <- min(uow_consumptions_inputs_18th)
original_data_18th_hour_max <- max(uow_consumptions_inputs_18th)
original_train_data_19th_hour <- uow_consumptions_inputs_19th[1: 380,]
original_test_data_19th_hour <- uow_consumptions_inputs_19th[381: 470,]
original_data_19th_hour_min <- min(uow_consumptions_inputs_19th)
original_data_19th_hour_max <- max(uow_consumptions_inputs_19th)
# Creating lags for the data sets
lag_18_1 = lag(uow_consumptions_inputs_18th_norm, 1)
lag_18_2 = lag(uow_consumptions_inputs_18th_norm, 2)
lag_18_3 = lag(uow_consumptions_inputs_18th_norm, 3)
lag_18_4 = lag(uow_consumptions_inputs_18th_norm, 4)
lag_18_7 = lag(uow_consumptions_inputs_18th_norm, 7)
lag_19_1 = lag(uow_consumptions_inputs_19th_norm, 1)
lag_19_2 = lag(uow_consumptions_inputs_19th_norm, 2)
lag_19_3 = lag(uow_consumptions_inputs_19th_norm, 3)
lag_19_4 = lag(uow_consumptions_inputs_19th_norm, 4)
lag_19_7 = lag(uow_consumptions_inputs_19th_norm, 7)
lag_20_1 = lag(uow_consumptions_inputs_20th_norm, 1)
lag_20_2 = lag(uow_consumptions_inputs_20th_norm, 2)
lag_20_3 = lag(uow_consumptions_inputs_20th_norm, 3)
lag_20_4 = lag(uow_consumptions_inputs_20th_norm, 4)
lag_20_7 = lag(uow_consumptions_inputs_20th_norm, 7)
# Create time-delayed I/O metrics for the data set
uow_consumptions_inputs_norm_io_18_1 <- cbind(uow_consumptions_inputs_18th_norm, lag_18_1, lag_18_2, lag_18_3, lag_18_4, lag_18_7)
colnames(uow_consumptions_inputs_norm_io_18_1) <- c("original", "t18_1", "t18_2", "t18_3", "t18_4", "t18_7")
uow_consumptions_inputs_norm_io_19_1 <- cbind(uow_consumptions_inputs_19th_norm, lag_19_1, lag_19_2, lag_19_3, lag_19_4, lag_19_7)
colnames(uow_consumptions_inputs_norm_io_19_1) <- c("original", "t19_1", "t19_2", "t19_3", "t19_4", "t19_7")
uow_consumptions_inputs_norm_io_20_1 <- cbind(uow_consumptions_inputs_20th_norm, lag_20_1, lag_20_2, lag_20_3, lag_20_4, lag_20_7)
colnames(uow_consumptions_inputs_norm_io_20_1) <- c("original", "t20_1", "t20_2", "t20_3", "t20_4", "t20_7")
uow_consumptions_inputs_norm_io_20_2 <- cbind(uow_consumptions_inputs_20th_norm, lag_20_1, lag_20_4, lag_20_7)
colnames(uow_consumptions_inputs_norm_io_20_2) <- c("original", "t20_1", "t20_4", "t20_7")
# Formatting time-delayed I/O metrics training data set removing N/As
uow_consumptions_inputs_norm_io_18_1 <- uow_consumptions_inputs_norm_io_18_1[complete.cases(uow_consumptions_inputs_norm_io_18_1),]
uow_consumptions_inputs_norm_io_19_1 <- uow_consumptions_inputs_norm_io_19_1[complete.cases(uow_consumptions_inputs_norm_io_19_1),]
uow_consumptions_inputs_norm_io_20_1 <- uow_consumptions_inputs_norm_io_20_1[complete.cases(uow_consumptions_inputs_norm_io_20_1),]
uow_consumptions_inputs_norm_io_20_2 <- uow_consumptions_inputs_norm_io_20_2[complete.cases(uow_consumptions_inputs_norm_io_20_2),]
# Divide testing and training data sets
uow_consumptions_inputs_norm_io_18_1_train <- uow_consumptions_inputs_norm_io_18_1[1: 373, ]
uow_consumptions_inputs_norm_io_18_1_test <- uow_consumptions_inputs_norm_io_18_1[-(1: 373), ]
uow_consumptions_inputs_norm_io_19_1_train <- uow_consumptions_inputs_norm_io_19_1[1: 373, ]
uow_consumptions_inputs_norm_io_19_1_test <- uow_consumptions_inputs_norm_io_19_1[-(1: 373), ]
uow_consumptions_inputs_norm_io_20_1_train <- uow_consumptions_inputs_norm_io_20_1[1: 373, ]
uow_consumptions_inputs_norm_io_20_1_test <- uow_consumptions_inputs_norm_io_20_1[-(1: 373), ]
uow_consumptions_inputs_norm_io_20_2_train <- uow_consumptions_inputs_norm_io_20_2[1: 373, ]
uow_consumptions_inputs_norm_io_20_2_test <- uow_consumptions_inputs_norm_io_20_2[-(1: 373), ]
# ----------------- NN Implementation -----------------
# ---- Train NN models ----
# Create the comparison table
columns = c("Model Name", "RMSE", "MAE", "MAPE", "sMAPE", "Training data set","Hidden layers", "Activation function", "Linear", "Algorithm")
comparison_table <- data.frame(matrix(nrow = 0, ncol = length(columns)))
colnames(comparison_table) <- columns
get_predicted_data_from_nn_model <- function(model, testing_data, data_min, data_max) {
# Test NN model and un-normalize predicted data
model_result <- neuralnet::compute(model, testing_data)
predicted_data <- unnormalize(model_result$net.result, data_min, data_max)
return(predicted_data)
}
# Update comparison table
insert_comparison_table_row <- function(model, model_name_str, training_data_set, hidden_layer_count, act_func, isLinear, algorithm, testing_data, actual_data, data_min, data_max) {
predicted_data <- get_predicted_data_from_nn_model(model, testing_data, data_min, data_max)
# RMSE evaluation https://www.r-bloggers.com/2021/07/how-to-calculate-root-mean-square-error-rmse-in-r/
rmse_value = calculate_rmse(data.matrix(actual_data), predicted_data)
# MAE evaluation https://www.r-bloggers.com/2021/07/how-to-calculate-mean-absolute-error-in-r/
mae_value = mae(data.matrix(actual_data), predicted_data)
# MAPE evaluation https://www.r-bloggers.com/2021/08/how-to-calculate-mean-absolute-percentage-error-mape-in-r/
mape_value = calculate_mape(data.matrix(actual_data), predicted_data)
# sMAPE evaluation https://www.r-bloggers.com/2021/08/how-to-calculate-smape-in-r/
smape_value = smape(data.matrix(actual_data), predicted_data) * 100
comparison_table[nrow(comparison_table) + 1,] <<- c(model_name_str, rmse_value, mae_value, mape_value, smape_value, training_data_set, hidden_layer_count, act_func, isLinear, algorithm)
}
# NN model - 1 (18th hour data set)
uow_consumptions_inputs_norm_io_18_1_train_model_1 <- neuralnet("original ~ t18_1 + t18_2 + t18_3 + t18_4 + t18_7", hidden = c(4, 2), data = uow_consumptions_inputs_norm_io_18_1_train, linear.output = TRUE)
insert_comparison_table_row(uow_consumptions_inputs_norm_io_18_1_train_model_1, "uow_consumptions_inputs_norm_io_18_1_train_model_1", "Set 1", 2, "logistic", FALSE, "Default", uow_consumptions_inputs_norm_io_18_1_test, original_test_data_18th_hour, original_data_18th_hour_min, original_data_18th_hour_max)
predicted_data_18th_hour <- get_predicted_data_from_nn_model(uow_consumptions_inputs_norm_io_18_1_train_model_1, uow_consumptions_inputs_norm_io_18_1_test, original_data_18th_hour_min, original_data_18th_hour_max)
# 2nd data set of 19th hour data
uow_consumptions_inputs_norm_io_19_1 <- cbind(uow_consumptions_inputs_19th_norm, lag_19_1, lag_19_2, lag_19_3, lag_19_4, lag_19_7)
colnames(uow_consumptions_inputs_norm_io_19_1) <- c("original", "t19_1", "t19_2", "t19_3", "t19_4", "t19_7")
# NN model - 2 (19th hour data set)
uow_consumptions_inputs_norm_io_19_1_train_model_1 <- neuralnet("original ~ t19_1 + t19_2 + t19_3 + t19_4 + t19_7", hidden = c(4, 2), data = uow_consumptions_inputs_norm_io_19_1_train, linear.output = TRUE)
insert_comparison_table_row(uow_consumptions_inputs_norm_io_19_1_train_model_1, "uow_consumptions_inputs_norm_io_19_1_train_model_1", "Set 1", 2, "logistic", FALSE, "Default", uow_consumptions_inputs_norm_io_19_1_test, original_test_data_19th_hour, original_data_19th_hour_min, original_data_19th_hour_max)
View(comparison_table)
# NN model - 2 (19th hour data set)
uow_consumptions_inputs_norm_io_19_1_train_model_1 <- neuralnet("original ~ t19_1 + t19_2 + t19_3 + t19_4 + t19_7", hidden = c(4, 2), data = uow_consumptions_inputs_norm_io_19_1_train, linear.output = TRUE)
# NN model - 2 (1st train set)
uow_consumptions_inputs_norm_io_1_train_model_2 <- neuralnet("original ~ t120 + t220 + t320 + t420 + t720 + t19 + t18", hidden = c(5, 3), data = uow_consumptions_inputs_norm_io_1_train, act.fct = "logistic", stepmax=1e7, linear.output = FALSE)
insert_comparison_table_row(uow_consumptions_inputs_norm_io_1_train_model_2, "uow_consumptions_inputs_norm_io_1_train_model_2", "Set 1", 2, "logistic", FALSE, "Default", uow_consumptions_inputs_norm_io_1_test, original_test_data)
# NN model - 3 (2nd train set)
uow_consumptions_inputs_norm_io_2_train_model_1 <- neuralnet("original ~ t120 + t420 + t720 + t19 + t18", hidden = c(4, 2), data = uow_consumptions_inputs_norm_io_2_train, linear.output = TRUE)
insert_comparison_table_row(uow_consumptions_inputs_norm_io_2_train_model_1, "uow_consumptions_inputs_norm_io_2_train_model_1", "Set 2", 2, "None", TRUE, "Default", uow_consumptions_inputs_norm_io_2_test, original_test_data)
# NN model - 4 (2nd train set)
uow_consumptions_inputs_norm_io_2_train_model_2 <- neuralnet("original ~ t120 + t420 + t720 + t19 + t18", hidden = c(5), data = uow_consumptions_inputs_norm_io_2_train, act.fct = "tanh", stepmax=1e7, linear.output = FALSE)
insert_comparison_table_row(uow_consumptions_inputs_norm_io_2_train_model_2, "uow_consumptions_inputs_norm_io_2_train_model_2", "Set 2", 1, "tanh", FALSE, "Default", uow_consumptions_inputs_norm_io_2_test, original_test_data)
# NN model - 5 (3rd train set)
uow_consumptions_inputs_norm_io_3_train_model_1 <- neuralnet("original ~ t120 + t720 + t19 + t18", hidden = c(3, 2), data = uow_consumptions_inputs_norm_io_3_train, linear.output = TRUE)
insert_comparison_table_row(uow_consumptions_inputs_norm_io_3_train_model_1, "uow_consumptions_inputs_norm_io_3_train_model_1", "Set 3", 2, "None", TRUE, "Default", uow_consumptions_inputs_norm_io_3_test, original_test_data)
# NN model - 6 (3rd train set)
uow_consumptions_inputs_norm_io_3_train_model_2 <- neuralnet("original ~ t120 + t720 + t19 + t18", hidden = c(5), data = uow_consumptions_inputs_norm_io_3_train, act.fct = "logistic", stepmax=1e7, linear.output = FALSE)
insert_comparison_table_row(uow_consumptions_inputs_norm_io_3_train_model_2, "uow_consumptions_inputs_norm_io_3_train_model_2", "Set 3", 1, "logistic", FALSE, "Default", uow_consumptions_inputs_norm_io_3_test, original_test_data)
View(comparison_table)
# Graphical representation - Best performed NN model
par(mfrow=c(1,1))
plot(data.matrix(original_test_data), get_predicted_data_from_nn_model(uow_consumptions_inputs_norm_io_1_train_model_1, uow_consumptions_inputs_norm_io_1_test), col='red', main='Real vs predicted NN', pch = 18, cex = 0.7)
abline(a=0, b=1, h=90, v=90)
plot(uow_consumptions_inputs_norm_io_2_train_model_1)