A simple long term predictive LightGBM model can be found in this notebook.

• Features
• Parameters
• Results

Based on the exploratory data analysis a simple feature engineering was performed. Based on EDA of meter readings:

• Healthcare and Utility usages shows the highest meter reading values
• Steam meter shows the highest meter reading values
• Monthly behaviour (meter-reading median):
  • Utility usage peaks in April-March
  • Chilledwater meter shows higher values in warm season
  • Steam meter shows lower values in April-October
• Hourly behaviour (meter-reading median):
  • Higher values from 6 hs to 19 hs
  • Utility usage shows oposite tendency
  • Steam meter pikes from 5 hs to 8 hs
• Weekday behaviour: lowers during weekends

In the following section can be found the features selected, transformed and created.

the following features were selected from each data set:

• Building metadata
  • Building ID*
  • Site ID*
  • Primary space usage
  • Building size (sqft)
• Weather data
  • Timestamp*
  • Site ID*
  • Air temperature
• Meter reading data
  • Timestamp*
  • Building ID*
  • meter
  • meter reading (target)

The following features were transformed:

• primaryspaceusage categories (16) were reduced to healthcare, utility and other
• meter categories (8) were preserved

The following features were created:

• month
• day of the week
• hour of the day

• Timestamp*
• Site ID
• Building ID
• Month
• Hour
• Day of the week
• Usage (3 levels: healthcare, utility, other)
• Building size (sqft)
• Air temperature
• Meter (8 levels)
• Meter reading / target

Parameters for this model were not tuned, but were manually modified to perform better than default.

• "objective": "regression"
• "metric": "rmse"
• "random_state": 55
• "learning_rate": 0.01, (default 0.1)
• "max_bin": 761 (default 255)
• "num_leaves": 2197 (default 31)