- Most cases has positive values
- submetering in columns 7,8,9 has a skew distribution, with really long tails for every case, particular the submetering in 9 has a range between 16 and 20 with more cases, can be considered with a bimodal distribution.
- Global intensity (6) follow similar rule
- Voltage for some reason has a lot outliers [-20,0) around 26k
- Glabal active and reactive power have similar distribution, but second one have several outliers
- Correletion heatmap show some strong correlation between: a. Global activity and global intensity b. Global reactive power and global Voltage c. Global activity and sub metering 3 but is a) who shows the more linear behaviour trend expected.
- Year 2006 is more noisy for first variable
- searching for average the voltage show clear trends for months and in july show a clear down trend, probably is the hot season (summer)
- The missing data is represented by -1 in data frame by all row
- Every year is increasing the amount of data missing
- there is no big difference in hour
- high presence of missing values on weekend if compare with the rest of week.
This can lead to a reason where there is more presence of people in the house but is just a theory for now