Skip to main content
EPJ Data Science

Table 3 Mean daily energy consumption features

From: Energy consumption prediction using people dynamics derived from cellular network data

Rank

Feature

Decrease in MSE

Decrease in node impurity

1

Number of consumers per grid powerline

19.64

69,109.19

2

Variance of real numbers part of Fourier transform of area codes

4.43

6,534.18

3

Variance of real numbers part of Fourier transform of outgoing SMS activity

3.89

6,811.37

4

Variance of calling direction area codes

3.56

7,964.44

5

Variance of entropy of outgoing call activity in time domain

3.47

11,568.62

6

Entropy of first harmonic of outgoing calls

3.47

3,671.82

7

Entropy of Internet activity summed in time domain

3.21

2,750.14

8

Kurtosis of entropy distribution of outgoing calls

3.14

1,472.87

9

Variance of skewness of temporal distribution of outgoing calls

3.10

2,111.95

10

Entropy of fundamental frequency (first harmonic) of Internet activity

3.03

1,668.04

11

Standard deviation of frequencies distribution skewness of incoming SMS

3.03

2,848.24

12

Entropy of sum in time domain of outgoing calls

2.98

3,651.77

13

Variance of the kurtosis of outgoing calls in time domain

2.98

2,136.39

14

Kurtosis of time entropy of mobile Internet activity

2.96

4,409.30

15

Median of outgoing calls variance in frequency domain

2.92

1,317.35

16

Sum of 4 harmonic of incoming calls

2.88

5,035.26

17

Sum of outgoing calls skewness in frequency domain

2.88

825.13

18

Median of outgoing SMS temporal distribution kurtosis

2.88

1,260.23

19

Kurtosis of outgoing calls 32 harmonic

2.81

947.15

20

Median of 11 harmonic of Internet activity

2.72

185.14

21

Variance of 29 harmonic of outgoing SMS

2.71

1,243.93

22

Variance of outgoing SMS fundamental frequency

2.69

5,418.83

23

Sum of intertemporal mean of outgoing SMS

2.67

182.93

24

Sum of calling direction area codes

2.63

4,075.28

25

Kurtosis of skewness of Internet activity frequencies

2.62

2,290.58

26

Median of temporal distribution kurtosis of Internet activity

2.61

1,510.62

27

Sum of 5 harmonic of outgoing calls

2.60

4,752.84

28

Sum of incoming calls temporal entropy

2.59

2,494.76

29

Variance of Internet activity temporal entropy

2.58

3,569.25

30

Sum of 7 harmonic of outgoing calls

2.58

201.58

31

Skewness of 13 harmonic of incoming calls

2.55

751.57

32

Sum of 11 harmonic of incoming calls

2.55

1,431.18

  1. For each variable from CDRs we computed the mathematical functions, which characterize the distributions and measure the information theoretic and statistical properties of such variables, e.g. mean, median, standard deviation, min and max values and Shannon entropy. Moreover, the same computations were repeated on sliding windows of variable length (1-hour, 4-hour and 1 day), producing second-order features that capture temporal relationships.
Back to article page